AI Search Measurement and Attribution: Beyond Rankings and Traffic

2. Why Traditional SEO Metrics Are Insufficient

Traditional search engine optimisation has developed around a relatively established collection of performance indicators. These include keyword rankings, search impressions, organic clicks, website sessions, engagement, conversions and revenue.

These measurements remain important because they provide information about how websites perform within conventional search environments and how visitors interact with organisational content.

However, AI-assisted discovery introduces interactions that may not be adequately represented by conventional search reporting systems.

A user can ask an AI assistant to explain a subject, compare products or identify potential service providers. The response may contain information about organisations without producing a measurable website visit.

This creates a distinction between visibility within an AI-generated response and visibility recorded through traditional website analytics.

The central issue is not that traditional SEO metrics have become irrelevant. It is that their measurement boundaries do not encompass every form of AI-assisted discovery.

2.1 The Historical Relationship Between Rankings and Traffic

Keyword rankings have traditionally been used to assess how prominently webpages appear for selected search queries.

Rankings can help organisations understand their competitive position within defined search results.

However, ranking measurements are affected by several variables, including location, device, query interpretation, search features and changes in search systems.

A ranking position therefore represents an observation under particular conditions rather than a permanent characteristic of a webpage.

The relationship between rankings and traffic is also more complex than a direct progression from higher position to greater commercial value.

Search result features, user intent and the relevance of competing results can influence whether a search impression produces a click.

Consequently, rankings should be considered alongside impressions, clicks, engagement and commercial outcomes.

2.2 Search Impressions Do Not Establish User Engagement

Search impressions describe exposure within the reporting definitions of a particular search platform.

They can help identify whether a website appears for relevant queries.

However, an impression does not establish that a user read the result, recognised the organisation or considered its services.

Nor does it establish that the impression contributed to a subsequent commercial decision.

This limitation becomes especially relevant when comparing conventional search exposure with AI-generated responses.

A website impression and an organisation appearing within an AI-generated answer are not equivalent measurement events.

Their definitions, observation methods and implications differ.

2.3 Organic Clicks Represent Only Identifiable Interactions

Organic clicks provide information about users selecting links within relevant search environments.

They remain useful because they represent identifiable interactions with search results.

However, not every search interaction produces a click.

Users may obtain information directly from search features, maps, product information or generated explanations.

AI-assisted interfaces can extend this behaviour by providing detailed answers to complex questions.

A user researching a subject may receive sufficient information without visiting the original sources.

Alternatively, the user may encounter an organisation during the interaction and investigate it later.

In either case, the initial discovery interaction may not produce an identifiable organic click.

This creates a measurement gap between information exposure and website engagement.

2.4 Zero-Click Discovery and AI-Generated Answers

Zero-click search describes interactions in which a user obtains information or completes a search activity without selecting an external result.

This behaviour existed before the widespread availability of generative AI.

Search engines have long provided direct answers, knowledge panels, maps and other features within their interfaces.

Generative AI introduces additional ways to deliver information directly.

An AI-generated answer may explain a subject, summarise available information or provide an initial comparison of alternatives.

The user may consider the answer sufficient for the immediate task.

However, zero-click behaviour should not automatically be interpreted as evidence of commercial influence.

The absence of a click does not establish that an organisation benefited from the interaction.

Equally, the absence of an identifiable referral does not prove that no influence occurred.

These outcomes must remain analytically distinct.

2.5 AI Visibility Can Occur Without Website Traffic

An organisation may appear within an AI-generated response in several ways.

Its research may be cited as an information source.

Its brand may be mentioned in an explanation.

Its services may appear within a provider comparison.

It may be presented as a potentially suitable organisation for a particular requirement.

None of these outcomes necessarily produces a website visit.

For example, a user may ask an AI assistant to identify the main considerations when selecting an enterprise SEO agency.

The response may mention several organisations and explain their apparent areas of expertise.

The user may then continue researching without visiting any of the named websites.

A conventional analytics system would not necessarily record the original interaction.

This does not make the interaction commercially valuable by definition.

It means that website traffic alone cannot describe every observable appearance within AI-assisted discovery.

2.6 The Difference Between Citations and Recommendations

AI source citations and provider recommendations represent different types of visibility.

A citation identifies information associated with a generated response.

A recommendation presents an organisation as potentially relevant to a user's requirements.

An organisation may be cited frequently without appearing in commercial recommendations.

Another organisation may appear in provider comparisons without being cited as an information source.

These distinctions matter because the commercial implications are different.

A citation may indicate that a source appeared in an informational context.

A provider recommendation may be more directly related to a commercial research task.

However, neither outcome independently establishes customer engagement or revenue.

Measurement systems should therefore classify these events separately.

2.7 Brand Mentions Are Not Equivalent to Brand Preference

A brand mention occurs when an organisation is named within an AI-generated response.

The mention may be positive, neutral, comparative or critical.

Its presence does not establish that the system prefers the organisation.

Nor does it demonstrate that the user has developed a favourable opinion.

For example, an AI assistant may mention several competing software providers while explaining the differences between their products.

The appearance of each brand represents an observable mention.

It does not establish equivalent suitability, recommendation strength or commercial influence.

A credible measurement approach should examine the context of each mention.

Relevant classifications may include informational, comparative, recommended, excluded or otherwise qualified appearances.

2.8 Conversational Search Creates Multi-Stage Discovery Journeys

Traditional keyword reporting commonly examines individual search queries.

Conversational AI allows users to refine requirements through a sequence of questions.

An initial informational question may develop into a commercial investigation.

Consider a user researching AI search measurement.

The conversation may begin with:

"How can a business measure visibility in AI-generated search results?"

The user may then ask:

"What is the difference between AI citations and recommendations?"

A subsequent question might be:

"Which agencies specialise in measuring AI search visibility for enterprise organisations?"

These questions represent different stages of a developing research process.

The user may receive information about multiple organisations before making any identifiable website visit.

Conventional keyword rankings cannot fully describe this sequence.

However, external researchers generally cannot observe every private conversation.

Consequently, structured prompt testing should be treated as a research sample, not a complete record of actual user behaviour.

2.9 Branded Search Can Follow Unobservable Discovery

A customer may first encounter an organisation through an AI-generated response.

The customer might later search for the organisation by name.

The eventual website visit may appear within analytics as organic branded traffic.

However, the earlier AI interaction may remain unobservable.

This creates a potential attribution gap.

An increase in branded search demand could have several explanations.

These include advertising, public relations, referrals, existing customer awareness and other marketing activities.

AI-assisted discovery may be one possible contributing factor.

Without supporting evidence, the relationship should not be presented as established.

Branded search trends may therefore provide useful contextual information, but they are not a standalone measure of AI-generated commercial influence.

2.10 Direct Traffic Does Not Identify Its Original Discovery Source

Direct traffic is frequently misunderstood within marketing attribution.

Analytics systems may classify visits as direct when a suitable referring source is unavailable.

This does not necessarily mean that the user manually entered the website address.

A visitor may arrive through a bookmark, an application or another interaction where referral information is absent.

AI-assisted discovery may contribute to some visits whose original source cannot be identified.

However, it would be incorrect to classify all unexplained direct traffic as AI-generated.

Such an approach would substantially overstate the available evidence.

Direct traffic should remain a separate reporting category unless reliable information supports a more specific attribution.

2.11 Referral Traffic Provides Partial Visibility

Some AI-assisted interfaces may generate website referrals that can be identified through available analytics data.

These referrals can provide useful information about engagement.

Organisations may examine:

• Identifiable referral sources
• Landing pages
• Engagement behaviour
• Qualified enquiries
• Recorded conversions
• Commercial outcomes where available

However, referral data provides only a partial view of AI-assisted discovery.

It cannot capture interactions that produce no website visit.

Referral information may also be incomplete or classified differently between analytics environments.

Consequently, identifiable AI referrals should be interpreted as a measurable subset of relevant activity rather than the complete scale of AI influence.

2.12 Conventional Conversion Attribution Has Important Limitations

Conversion attribution attempts to connect marketing interactions with subsequent business outcomes.

Common approaches include last-click attribution and other rules-based or modelled methods.

These methods depend on the information available to the measurement system.

If an AI-assisted discovery interaction is not recorded, the attribution model may be unable to include it.

This does not establish that the interaction was commercially important.

It establishes a limitation in the available measurement data.

Organisations should therefore distinguish between the attribution produced by a reporting model and the broader question of causal commercial influence.

A recorded attribution relationship does not automatically establish that a particular interaction caused the conversion.

2.13 The Risk of Overstating AI Search Performance

Emerging measurement disciplines can create pressure to demonstrate performance through simplified indicators.

For AI-assisted discovery, this may involve reporting citation counts, mention frequency or recommendation appearances as evidence of commercial value.

These indicators can be useful when clearly defined.

However, their interpretation requires caution.

A citation count does not establish how many users encountered the citation.

A brand mention does not establish customer preference.

A provider recommendation does not establish that the user contacted or selected the organisation.

An increase in branded traffic does not independently establish that AI discovery caused the increase.

These distinctions are essential for credible reporting.

2.14 The Need for a Broader Measurement Model

Traditional SEO metrics should remain part of organisational search reporting.

However, AI-assisted discovery requires additional observational and attribution categories.

CGO Media proposes examining search performance through four complementary dimensions.

Dimension One — Conventional Search Visibility

Rankings, impressions and other indicators describing performance within conventional search environments.

Dimension Two — AI Answer Visibility

Documented appearances within AI-generated responses, including citations, mentions, comparisons and recommendations.

Dimension Three — Identifiable Engagement

Website visits, enquiries and other interactions supported by available measurement data.

Dimension Four — Commercial Outcomes

Qualified opportunities, conversions, revenue and other relevant business results.

These dimensions should be analysed together while retaining their individual definitions.

The objective is not to produce a single universal AI visibility score.

It is to establish a more complete and transparent understanding of different discovery outcomes.

2.15 Implications for SEO and GEO Reporting

Organisations reporting on both SEO and GEO should avoid combining unrelated metrics into a single performance figure without explaining the methodology.

A useful reporting structure should distinguish:

• Traditional search performance
• Observed AI citation visibility
• Observed brand and entity representation
• Observed provider comparison visibility
• Observed recommendation visibility
• Identifiable AI referral engagement
• Commercial outcomes supported by evidence
• Potential influence that remains unattributed

Reports should also explain the observation period, platforms examined, questions tested and relevant limitations.

This provides a more defensible basis for evaluating AI search performance than relying on isolated visibility claims.

Section 2 — Research Interpretation

The development of AI-assisted discovery exposes important limitations in conventional search measurement.

Rankings, impressions, clicks and website traffic remain useful, but they do not independently describe how organisations appear within AI-generated answers.

AI citations, brand mentions and provider recommendations introduce additional forms of observable visibility.

However, these appearances should not be confused with commercial engagement or proven revenue.

The principal measurement challenge is to connect different types of visibility with identifiable outcomes while recognising the limits of attribution.

Supporting CGO Media Research

This section connects with CGO Media's wider research into AI visibility, search authority and commercial outcomes.

Research Observations Library:
AI Search Visibility, AI Search ROI and AI Search Conversion Research

Research Library:
AI Search, GEO and Search Authority Research

The State of Search:
The State of Search — Search Measurement and Commercial Impact

Research Methodology:
CGO Media Research Methodology

Section 2 — Key Finding

Traditional SEO metrics are necessary but insufficient for measuring the full range of AI-assisted discovery outcomes.

A credible measurement approach must distinguish conventional search visibility, AI answer visibility, identifiable engagement and commercial performance.

These categories should be connected through transparent analysis rather than combined into unsupported claims about AI-generated revenue.

4. Measuring AI Visibility in Practice

The development of AI search visibility metrics creates an important methodological challenge. Although researchers can observe how organisations appear within AI-generated responses, the reliability of their conclusions depends on how those observations are collected, classified and interpreted.

A single AI-generated response provides evidence of what occurred during one interaction. It does not establish how frequently the same organisation will appear for other users, across different platforms or at different times.

Similarly, repeated appearances may indicate consistent visibility within a defined research sample without establishing the internal reasons why a system selected particular information.

CGO Media therefore proposes a structured observational methodology designed to support transparent, repeatable and appropriately qualified AI search measurement.

The objective is not to reverse-engineer proprietary algorithms. It is to document observable outcomes, evaluate their consistency and identify meaningful differences between discovery environments.

4.1 Defining the Research Question

Every AI visibility investigation should begin with a clearly defined research question.

Broad questions such as "How visible is our brand in AI search?" are insufficient unless visibility is operationally defined.

A more precise question might be:

This question establishes several important research boundaries:

• The organisation being examined.
• The type of visibility being measured.
• The relevant user requirements.
• The AI platforms included.
• The observation period.
• The research sample used for comparison.

These boundaries help prevent the results from being interpreted more broadly than the evidence supports.

4.2 Defining the Research Population

The research population describes the wider group of interactions about which the investigation intends to draw conclusions.

For AI search research, this population may be difficult to observe directly because researchers generally cannot access all private conversations occurring within AI platforms.

Consequently, most external AI visibility investigations rely on selected prompts intended to represent relevant discovery tasks.

These prompts constitute a research sample rather than a complete record of actual user behaviour.

Possible Research Populations

• UK businesses researching enterprise SEO agencies.
• European manufacturers evaluating industrial suppliers.
• Healthcare organisations researching digital service providers.
• SaaS companies comparing technical solutions.
• Consumers investigating products within a defined category.

The research population should be described as precisely as possible, including relevant geography, industry and commercial requirements.

Researchers should avoid describing a convenience sample of prompts as representative of all AI search users.

4.3 Designing a Structured Prompt Sample

Prompt selection is one of the most consequential decisions in an AI visibility study.

The wording of a question can influence which information appears in the response.

A question asking for a general explanation may produce different organisational references from one asking for a shortlist of commercial providers.

CGO Media recommends organising prompts into clearly defined categories.

The proportion of prompts assigned to each category should reflect the research objectives.

Equal weighting may be appropriate for some comparative studies, but it should not be assumed to reflect real-world search behaviour.

4.4 Avoiding Prompt Selection Bias

Prompt selection bias occurs when the questions included in a study systematically favour particular outcomes.

For example, questions containing an organisation's name are more likely to produce responses discussing that organisation.

Such questions may be useful for evaluating entity representation accuracy.

However, they are generally unsuitable for measuring unaided provider discovery.

Branded and Unbranded Prompts

Branded prompts explicitly identify an organisation.

Unbranded prompts describe a requirement without naming the organisation being investigated.

Both categories can be useful, but they answer different research questions.

Researchers should report them separately.

Additional Bias Controls

• Define prompts before collecting results.
• Avoid selecting only favourable questions.
• Include relevant competing solutions where appropriate.
• Record excluded prompts and the reasons for exclusion.
• Retain unsuccessful or negative observations.
• Document any changes to the prompt sample.

These controls reduce the risk of overstating visibility through selective reporting.

4.5 Determining Sample Size

There is no universally appropriate sample size for every AI visibility investigation.

The number of observations required depends on the research question, expected variability, desired precision and available resources.

A small exploratory sample may help identify possible patterns.

However, it may provide insufficient evidence for strong comparative conclusions.

Larger samples can improve the precision of estimates under suitable sampling assumptions, but sample size alone does not eliminate bias.

Illustrative Research Design

A hypothetical investigation might use:

• 30 predefined research prompts.
• Three selected AI platforms.
• Five repeated observations per prompt and platform.

This would produce a planned total of 450 response observations.

The example is illustrative and does not represent a completed CGO Media experiment.

Repeated responses to the same prompt should not automatically be treated as independent observations for statistical inference.

Appropriate analysis should account for clustering by prompt and other relevant features of the research design.

4.6 Controlling Conversation Context

Conversational AI systems may incorporate information from earlier interactions.

This means that identical questions can produce different responses when submitted within different conversation histories.

Researchers should therefore define how conversation context will be handled.

Independent Prompt Testing

Each question is submitted within a separate conversation where the platform permits this approach.

This can help reduce the influence of preceding research questions.

Sequential Conversation Testing

Several related questions are submitted as part of a defined conversational journey.

This approach can help investigate how responses change as user requirements become more specific.

These two methods should not be combined without appropriate classification.

A recommendation appearing after several contextual questions is not directly equivalent to one appearing in response to an independent prompt.

4.7 Platform and Model Identification

AI-assisted discovery environments differ in their available capabilities, information sources and response formats.

Researchers should identify the platform examined and, where available, the relevant model or system configuration.

Important details may include:

• Platform name.
• Model or system identifier, if available.
• Interface or access method.
• Whether web retrieval was enabled.
• Relevant account or configuration settings.
• Observation date.

Where system details are unavailable, researchers should state this limitation.

Platform comparisons should not assume that all systems use identical retrieval or citation methods.

4.8 Geographic and Language Controls

Search and AI-assisted discovery can be influenced by geographic and linguistic context.

A question concerning UK financial services providers may produce different results from one concerning the European market.

Similarly, questions submitted in different languages may produce different organisational references.

Researchers should document:

• Prompt language.
• Requested geographic market.
• Relevant platform location settings.
• Any known regional restrictions.
• Whether geographic context was explicitly stated.

Researchers should distinguish a geographic requirement included in the prompt from the actual location context available to the platform.

These may not be equivalent.

4.9 Repeated Observations and Temporal Variability

AI-generated responses may change between repeated interactions.

Variation can arise from system behaviour, available information, contextual differences and platform updates.

Repeated observations help describe the consistency of an outcome within a defined testing environment.

Within-Period Repetition

The same prompt is tested repeatedly during a defined observation period.

This helps examine short-term response variability.

Longitudinal Observation

The same research design is repeated at different points in time.

This can help identify changes in observed visibility.

However, longitudinal changes may reflect multiple factors, including platform updates, source changes and differences in testing conditions.

A change in visibility should not automatically be attributed to an SEO or GEO intervention.

4.10 Recording Complete Responses

Reliable research requires adequate documentation of observed results.

Recording only the organisation names appearing in a response may remove important context.

For example, an organisation may be named as a recommended provider, an unsuitable option or a historical example.

These appearances have different meanings.

Recommended Observation Record

• Unique observation identifier.
• Research prompt identifier.
• Exact prompt text.
• Platform and available configuration details.
• Date and time of observation.
• Relevant conversation context.
• Complete response record.
• Source links or citations.
• Organisations mentioned.
• Recommendation classifications.
• Researcher notes.

Records should be retained in accordance with applicable platform terms, privacy requirements and research governance procedures.

4.11 Classifying AI Source Citations

Citation classification should distinguish source attribution from general references to organisations.

A qualifying citation may identify a webpage, publication or other source associated with the generated response.

Researchers should define whether citations are attributed at the domain, organisation or individual resource level.

Domain-Level Citation

A citation associated with a defined website domain.

Publication-Level Citation

A citation associated with a specific research paper, article or resource.

Organisation-Level Citation

A citation attributed to a defined organisational entity.

These classifications should not be combined without explaining how duplicate and overlapping citations are handled.

Researchers should also consider whether the cited source supports the relevant claim.

Citation presence and citation accuracy are separate research questions.

4.12 Classifying Brand Mentions

Brand mentions should be assessed in context.

The appearance of an organisation's name does not necessarily indicate endorsement.

CGO Media proposes the following classifications.

Researchers should document the criteria used to assign each classification.

Where multiple categories apply, the reporting method should explain whether multiple labels are permitted.

4.13 Classifying Provider Recommendations

Recommendation visibility requires more careful interpretation than simple name detection.

A response may mention an organisation without recommending it.

It may also recommend an organisation only under particular conditions.

Recommended Classification

• Explicit recommendation.
• Conditional recommendation.
• Specialist or niche recommendation.
• Neutral comparison inclusion.
• Alternative provider.
• Explicitly unsuitable provider.
• Insufficient information to classify.

These classifications should be applied consistently across the research sample.

The researcher should evaluate the relationship between the recommendation and the user's stated requirements.

An organisation should not be counted as recommended solely because its name appears in a list.

4.14 Entity Accuracy Verification

AI-generated responses may contain information about an organisation's identity, services, locations and capabilities.

These claims should be evaluated against appropriate evidence.

Relevant sources may include official organisational information, public records, professional registers and other suitable documentation.

Researchers should record the source used to verify each material claim.

Unsupported claims should not automatically be classified as false.

Where evidence is insufficient, the appropriate classification may be unverified or not assessable.

This distinction is particularly important for organisations operating in regulated industries.

4.15 Researcher Agreement and Quality Assurance

Some AI visibility classifications require human judgement.

Two researchers may disagree about whether a response constitutes a recommendation or a neutral comparison.

Research quality can be improved through predefined coding rules and independent review.

Recommended Quality Controls

• Prepare a written classification guide.
• Define examples for each category.
• Use independent coding where feasible.
• Record disagreements.
• Apply a documented resolution procedure.
• Review ambiguous observations separately.

Where the study design supports it, researchers may calculate an appropriate inter-rater agreement statistic.

However, agreement statistics should be interpreted in relation to the classification task and sample characteristics.

4.16 Statistical Interpretation

AI visibility studies often report percentages describing observed appearances.

Such percentages are useful descriptive measures.

However, their precision and generalisability depend on the research design.

A result obtained from a small convenience sample should not be interpreted as a reliable estimate of visibility across all real-world users.

Researchers should consider:

• Sample size.
• Prompt selection method.
• Repeated-measure dependence.
• Platform differences.
• Temporal variability.
• Classification uncertainty.
• Missing or excluded observations.

Where appropriate, uncertainty intervals may be calculated using methods suitable for the sampling structure.

Statistical significance should not be claimed without a valid inferential design.

4.17 Comparing AI Platforms

Platform comparisons can help identify differences in observed visibility.

However, comparisons require consistent research definitions.

The same prompt may produce different responses because platforms differ in their capabilities, retrieval methods and information sources.

Researchers should distinguish between differences caused by the platform and differences caused by the research conditions.

Comparison Requirements

• Comparable prompt samples.
• Clearly identified platforms.
• Documented observation periods.
• Consistent classification criteria.
• Transparent handling of missing responses.
• Recognition of platform-specific limitations.

A platform with more observed citations is not necessarily more commercially valuable.

Citation frequency and commercial performance remain separate questions.

4.18 Longitudinal AI Visibility Research

Repeated measurement over time can help organisations examine changes in AI visibility.

A longitudinal study may repeat a defined prompt sample at scheduled intervals.

Researchers can then compare observed citation, mention and recommendation rates across periods.

However, changes may reflect several factors.

• Platform updates.
• Changes in available sources.
• Changes in organisational information.
• Changes in competing organisations.
• Response variability.
• Changes in research conditions.

Consequently, a visibility increase following an optimisation activity does not independently establish causation.

Stronger causal claims require an appropriate evaluation design and consideration of alternative explanations.

4.19 Research Reporting Standards

AI visibility findings should be accompanied by sufficient methodological information to support responsible interpretation.

Where commercial outcomes are discussed, the report should also explain the available attribution evidence.

Visibility observations should not be presented as confirmed commercial performance without supporting data.

4.20 Ethical and Privacy Considerations

AI search research should respect applicable privacy requirements, platform terms and information security responsibilities.

Researchers should avoid submitting confidential customer information or unnecessary personal data into public AI interfaces.

Research records should be retained securely and access should be appropriately controlled.

Where customer interviews or surveys are used, suitable consent and data protection procedures should be followed.

These considerations become especially important when investigating commercial attribution and customer behaviour.

Section 4 — Research Interpretation

Measuring AI visibility requires more than collecting generated responses and counting organisation names.

Reliable research depends on defined questions, structured prompt samples, documented observations and consistent classification rules.

Repeated testing can help describe variability, while longitudinal observation can identify changes across time.

However, neither method automatically establishes causal relationships between optimisation activities and observed outcomes.

The credibility of an AI visibility measurement depends on the quality of the research design, not simply the number of responses collected.

Supporting CGO Media Research

Research Methodology:
CGO Media Research Methodology

Research Observations Library:
AI Search Visibility and Citation Authority Research

Framework Library:
CGO Media Framework Library

The State of Search:
Search Measurement and Commercial Impact

Section 4 — Key Finding

AI search visibility can be investigated through structured observation, but reliable findings require carefully defined samples, consistent classification and transparent reporting.

Researchers should distinguish descriptive observations from generalisable estimates and causal conclusions.

The objective is to produce evidence that organisations can interpret responsibly, rather than unsupported claims of universal AI search visibility.

5. Attribution Beyond Website Traffic

The measurement of AI search visibility and the attribution of commercial outcomes represent two related but distinct research challenges.

Visibility research investigates whether an organisation appears within AI-generated responses. Attribution research investigates whether there is sufficient evidence to connect a discovery interaction with a subsequent customer action or commercial outcome.

These questions require different evidence.

An organisation may be cited in an AI-generated answer without receiving a website visit. It may appear in a provider comparison without receiving an enquiry. It may also be discovered through an AI assistant before a customer contacts it through another channel.

Consequently, website analytics alone may provide an incomplete account of AI-assisted discovery.

However, the absence of identifiable referral data does not justify attributing unexplained commercial activity to AI search.

CGO Media proposes that AI search attribution should distinguish documented interactions, supported attribution relationships, modelled estimates and potential influence that remains unverified.

5.1 The Difference Between Visibility and Attribution

Visibility concerns an organisation's observed presence within a defined discovery environment.

Attribution concerns the evidence connecting an interaction or channel with a subsequent outcome.

These concepts should not be treated as interchangeable.

Both observations are useful, but they answer different questions.

5.2 The AI Search Attribution Gap

The AI search attribution gap describes the difference between potentially relevant discovery interactions and the subset of those interactions that can be identified through available measurement systems.

The gap may arise because users receive information without visiting an external website.

It may also arise when users continue their research through other channels after encountering an organisation within an AI-generated response.

Consider a hypothetical customer researching specialist manufacturing suppliers.

The customer asks an AI assistant to explain which suppliers may meet a particular technical requirement.

Several organisations are mentioned.

The customer later searches for one supplier by name, visits its website and submits a quotation request.

The website analytics system may identify the branded search visit but may have no record of the earlier AI interaction.

This creates a possible attribution gap.

However, without additional evidence, the organisation cannot establish that the AI interaction occurred or influenced the customer's decision.

5.3 Four Categories of AI Search Attribution Evidence

CGO Media proposes four categories for classifying evidence associated with AI-assisted commercial discovery.

5.4 Direct Referral Attribution

Identifiable AI referral traffic provides one of the clearest opportunities to connect AI-assisted discovery with website engagement.

Where referral information is available, organisations may examine visits originating from recognised AI-related sources.

Relevant measurements include:

• Identifiable AI referral sessions.
• Landing pages receiving those referrals.
• Engagement events.
• Enquiry form submissions.
• Recorded purchases.
• Qualified leads.
• CRM opportunities linked to identifiable visits.
• Revenue associated with recorded transactions.

However, attribution depends on the reliability of referral information.

Some visits may not retain sufficient source information for classification.

Organisations should therefore avoid assuming that identifiable referrals represent all AI-assisted discovery activity.

5.5 AI Referral Conversion Rate

Where suitable analytics data exists, organisations can calculate a conversion rate for identifiable AI referral sessions.

Consider a hypothetical reporting period containing 200 eligible AI referral sessions.

If 10 sessions contain the defined conversion, the observed session conversion rate is 5%.

This example is illustrative and does not represent an empirical CGO Media finding.

Researchers should define the conversion event and denominator consistently.

Session conversion rates should not be confused with user conversion rates, lead qualification rates or revenue attribution.

5.6 Assisted Attribution and Multi-Channel Journeys

Customer journeys frequently involve more than one interaction.

A prospective customer may discover an organisation through one channel, investigate alternatives through another and convert through a third.

AI-assisted discovery may form part of such a sequence.

For example, a customer might:

1. Research a business problem using an AI assistant.
2. Encounter information about a potential provider.
3. Visit the provider's website through an identifiable referral.
4. Return several days later through branded search.
5. Submit a commercial enquiry.

If the relevant interactions can be reliably connected, an attribution system may recognise multiple contributing touchpoints.

However, attribution models assign credit according to predefined rules or statistical assumptions.

The allocation of attribution credit does not independently establish the causal contribution of each interaction.

5.7 First-Touch, Last-Touch and Multi-Touch Attribution

Different attribution approaches may assign different commercial credit to the same customer journey.

First-Touch Attribution

Assigns credit to the first recorded qualifying interaction.

If an identifiable AI referral is the first recorded touchpoint, it may receive attribution credit under this model.

Last-Touch Attribution

Assigns credit to the final recorded qualifying interaction before conversion.

If the customer returns through branded search, the AI referral may receive no credit under a strict last-touch model.

Multi-Touch Attribution

Distributes credit across multiple recorded interactions.

The allocation may depend on fixed rules or a statistical model.

None of these approaches can include an interaction that was never observed unless additional modelling assumptions are introduced.

5.8 Customer-Reported Discovery

Customer-reported discovery can provide useful evidence about interactions that are not visible in website analytics.

Organisations may ask customers how they first became aware of the business.

This information can be collected through enquiry forms, sales conversations or customer surveys.

Where a customer identifies AI-assisted discovery, a follow-up question may establish which platform was used.

However, questions should avoid encouraging customers to select AI search as the expected answer.

Customer-reported attribution remains subject to recall errors and incomplete knowledge of the discovery process.

5.9 The Importance of Neutral Survey Design

Survey wording can influence the information customers provide.

A question such as "Did ChatGPT recommend our company?" may encourage a particular response.

A more neutral approach asks how the customer discovered the organisation before requesting additional details where appropriate.

Researchers should distinguish between prompted recognition and unaided recall.

They should also record whether customers identify a specific AI interaction or simply believe that AI search may have played a role.

These distinctions affect the strength of the attribution evidence.

5.10 Integrating AI Discovery into CRM Systems

Customer relationship management systems can provide a useful connection between marketing interactions and commercial outcomes.

However, the quality of CRM attribution depends on the information recorded.

CGO Media recommends separating observed referral data from customer-reported discovery.

Suggested CRM Fields

• First recorded acquisition source.
• Latest recorded acquisition source.
• Customer-reported discovery source.
• AI platform identified by the customer, if known.
• Discovery evidence classification.
• Enquiry date.
• Lead qualification status.
• Opportunity value.
• Opportunity outcome.
• Recognised revenue where applicable.

These fields can support more consistent reporting.

However, they should not be interpreted as proof that the recorded source caused the commercial outcome.

5.11 Branded Search as a Supporting Indicator

Branded search demand may provide contextual information about organisational awareness.

A customer who encounters an organisation through AI-assisted discovery may later search for its name.

However, branded search can also be influenced by advertising, public relations, referrals, offline activity and existing customer relationships.

Consequently, branded search growth should not be automatically attributed to AI visibility.

It may be examined as one supporting indicator within a broader research design.

5.12 Direct Traffic and Unidentified Sources

Direct traffic is another category that requires careful interpretation.

Analytics systems may classify visits as direct when suitable referral information is unavailable.

Some of these visits may follow earlier AI-assisted discovery.

Others may originate from bookmarks, applications, offline activity or other unidentified sources.

Therefore, direct traffic should not be reclassified as AI search traffic without supporting evidence.

Unexplained traffic is not equivalent to AI-attributed traffic.

5.13 Dark Social, Private Sharing and Unobservable Discovery

Some discovery interactions occur through channels that do not provide complete referral information.

Private messaging, copied links and certain application environments may contribute to this measurement problem.

AI-assisted discovery introduces additional situations where a user may encounter information without producing a measurable external referral.

However, these different sources of attribution loss should not be combined into a single AI influence estimate without a defensible methodology.

5.14 Attribution Windows

Attribution systems commonly use defined time windows to determine which recorded interactions are eligible for credit.

The appropriate window may depend on the customer journey.

A relatively simple consumer purchase may involve a shorter decision process than a complex enterprise procurement project.

Organisations should therefore explain the attribution windows used in their reporting.

They should also recognise that a longer attribution window does not necessarily provide stronger evidence of causal influence.

5.15 Attribution in Long B2B Sales Cycles

Business-to-business purchasing decisions may involve multiple stakeholders and extended evaluation periods.

An AI-generated response may be encountered by one member of a purchasing organisation.

Another stakeholder may subsequently visit the provider's website.

A third individual may submit the eventual enquiry.

These interactions may not be identifiable as part of the same buying process.

Consequently, individual session attribution may provide only a partial account of organisational purchasing journeys.

Account-level CRM information and customer interviews may provide additional context where collected appropriately.

5.16 Lead Quality and Commercial Qualification

Not every enquiry has equivalent commercial value.

Organisations should distinguish raw enquiries from qualified commercial opportunities.

A high volume of low-quality referrals may produce less commercial value than a smaller number of relevant enquiries.

Relevant qualification indicators may include:

• Customer requirement alignment.
• Geographic suitability.
• Available budget.
• Decision-making authority.
• Expected purchasing timeframe.
• Commercial opportunity value.

These criteria should be adapted to the organisation's business model and documented consistently.

5.17 AI-Attributed Qualified Lead Rate

Where an organisation has suitable lead qualification records, it may calculate the proportion of AI-attributed leads meeting its qualification criteria.

This metric should be calculated separately for directly observed, customer-reported and modelled attribution categories.

Combining them without distinction may obscure differences in evidence quality.

5.18 From Enquiries to Revenue

Commercial attribution becomes more complex when organisations attempt to connect discovery activity with revenue.

A recorded enquiry may develop into a qualified opportunity.

That opportunity may subsequently result in a contract, purchase or other commercial outcome.

However, each stage introduces additional factors influencing the eventual result.

These may include pricing, product suitability, sales performance, customer requirements and competitive alternatives.

Attribution systems may assign revenue credit to an identifiable acquisition source.

This does not independently establish that the source caused the revenue to occur.

5.19 Attribution Versus Incrementality

Attribution and incrementality answer different questions.

Attribution asks how credit should be assigned among observed or modelled interactions.

Incrementality asks whether an activity produced outcomes that would not otherwise have occurred.

This distinction is central to evaluating commercial AI search performance.

Incrementality generally requires a suitable causal evaluation design.

Observational attribution reports alone are insufficient to establish incremental revenue.

5.20 Approaches to Incrementality Research

Where appropriate, organisations may investigate incremental effects using experimental or quasi-experimental research methods.

Potential approaches include controlled interventions, matched comparisons and carefully designed time-series analysis.

However, AI search introduces practical challenges.

Organisations cannot necessarily control how AI platforms represent their content.

They may also be unable to isolate AI visibility changes from broader marketing and search activity.

Consequently, causal evaluation should be attempted only where the research design supports meaningful conclusions.

Otherwise, results should remain classified as observational associations.

5.21 Attribution Confidence Classifications

CGO Media recommends that commercial reports classify attribution evidence according to its source and reliability.

These classifications describe different evidence types.

They should not automatically be interpreted as a universal ranking of causal certainty.

A directly observed interaction may provide strong evidence that an event occurred, while providing limited evidence of its incremental commercial effect.

5.22 The Risk of Double Counting

Organisations should take particular care when combining different attribution sources.

A customer may arrive through an identifiable AI referral and also report using an AI assistant.

These records may describe the same discovery journey.

Counting them as two separate acquisitions would overstate commercial performance.

Attribution systems should therefore distinguish between unique commercial outcomes and multiple pieces of supporting evidence.

Where appropriate, deduplication rules should be documented.

5.23 Building an AI Attribution Evidence Register

CGO Media proposes maintaining an attribution evidence register for organisations undertaking systematic AI search measurement.

The register should record the evidence supporting each attribution classification.

Suggested Register Fields

• Unique commercial event identifier.
• Event date.
• Recorded acquisition source.
• Customer-reported discovery source.
• Relevant AI platform, if known.
• Attribution evidence category.
• Supporting record reference.
• Lead qualification status.
• Opportunity value.
• Commercial outcome.
• Revenue recognition status.
• Known limitations.

The register should be maintained in accordance with applicable privacy and data protection requirements.

Its purpose is to make attribution claims auditable rather than to maximise the volume of revenue assigned to AI search.

5.24 A Practical Attribution Reporting Structure

Organisations can organise AI search attribution reporting into four separate reporting categories.

These categories should be reported separately unless an explicit reconciliation methodology explains how overlapping evidence is handled.

5.25 What AI Attribution Cannot Reliably Establish

Even sophisticated attribution systems have important limitations.

They generally cannot identify every private AI-assisted interaction.

They cannot reliably reconstruct all unrecorded customer research activity.

They cannot assume that unexplained branded or direct traffic originated from AI discovery.

They cannot establish incremental revenue merely by assigning attribution credit.

They also cannot determine how much an individual AI recommendation influenced a customer's decision without appropriate supporting evidence.

These limitations should be acknowledged rather than concealed through overly precise reporting.

Section 5 — Research Interpretation

AI-assisted discovery introduces important challenges for commercial attribution.

Some interactions produce identifiable referrals and recorded conversions.

Others may contribute to customer research without generating directly observable website engagement.

Customer-reported discovery and CRM records can provide additional evidence, but these sources have their own limitations.

Attribution models can allocate credit among recorded interactions.

However, attribution should not be confused with causal evidence of incremental commercial value.

Credible AI search attribution requires organisations to distinguish documented interactions, supported customer reports, modelled estimates and unverified influence.

Supporting CGO Media Research

AI Search ROI Research:
AI Search ROI Research UK 2026

AI Search Conversion Research:
AI Search Conversion Research UK 2026

Research Observations Library:
CGO Media Research Observations

The State of Search:
The State of Search

Research Methodology:
CGO Media Research Methodology

Section 5 — Key Finding

AI search attribution should be based on identifiable evidence rather than assumptions about unobserved customer behaviour.

Website referrals, customer-reported discovery, CRM records and statistical models can each contribute useful information.

However, they should remain clearly classified according to their evidential limitations.

The absence of a measurable referral does not prove that AI-assisted discovery had no influence.

Equally, the possibility of influence does not justify claiming commercial outcomes that cannot be supported by evidence.

6. Commercial Measurement, Investment and ROI

The commercial evaluation of AI search and Generative Engine Optimisation requires organisations to move beyond visibility indicators and examine whether relevant activities contribute to measurable business outcomes.

Citation frequency, brand mentions, provider comparisons and recommendation visibility can describe an organisation's observed presence within AI-generated responses.

However, these indicators do not independently establish customer acquisition, revenue generation or profitability.

A commercially meaningful measurement framework must distinguish between the cost of AI search activities, the outcomes recorded through available measurement systems and the evidence supporting any claimed financial contribution.

CGO Media proposes evaluating AI search commercial performance through six connected dimensions: investment, visibility, engagement, lead quality, revenue attribution and incremental economic value.

These dimensions should be examined together while retaining their individual definitions and evidential limitations.

6.1 The Difference Between Visibility and Commercial Value

An organisation may achieve frequent visibility within AI-generated answers without receiving identifiable enquiries.

Another organisation may appear less frequently but receive a smaller number of commercially valuable referrals.

These possibilities demonstrate why visibility frequency should not be treated as a direct substitute for financial performance.

The third question requires stronger evidence than a conventional attribution report.

It concerns the causal effect of an intervention rather than the allocation of credit among observed marketing interactions.

6.2 The Six Commercial Measurement Dimensions

CGO Media proposes organising commercial AI search measurement into six dimensions.

The six dimensions provide a progression from organisational investment to increasingly demanding forms of commercial evidence.

They do not imply that every organisation will possess sufficient information to measure all six dimensions.

6.3 Establishing the Cost of AI Search Investment

Commercial evaluation begins with understanding the resources invested.

Organisations should avoid calculating return on investment using only a narrow subset of relevant costs.

AI search and GEO programmes may involve several expenditure categories.

Research and Strategy Costs

• Research planning and analysis.
• AI visibility investigations.
• Competitive and sector research.
• Strategic development.
• Research reporting.

Content and Authority Development Costs

• Research publication development.
• Editorial production.
• Technical documentation.
• Content maintenance.
• Digital public relations.
• Supporting authority development.

Technical and Measurement Costs

• Technical SEO implementation.
• Structured data development.
• Analytics configuration.
• Measurement tools.
• Data processing.
• Quality assurance.

Operational Costs

• Internal personnel.
• External agency fees.
• Software subscriptions.
• Research infrastructure.
• Project management.

Cost allocation should be consistent with the organisation's financial reporting requirements.

Where an activity supports both conventional SEO and GEO, the allocation method should be documented rather than arbitrarily assigning the entire cost to one discipline.

6.4 Shared SEO and GEO Investment

Many activities relevant to AI search also support conventional search visibility and broader digital marketing performance.

For example, improving technical accessibility, publishing authoritative research and correcting organisational information may support multiple discovery environments.

This creates a cost allocation problem.

An organisation cannot automatically attribute the full cost or benefit of a shared activity exclusively to AI search.

CGO Media recommends distinguishing:

• Costs exclusively associated with AI search research.
• Costs exclusively associated with conventional SEO.
• Shared technical and content investment.
• General marketing and organisational expenditure.

Any allocation of shared costs should use a stated methodology.

This is particularly important when comparing the apparent profitability of SEO and GEO programmes.

6.5 Measuring Qualified Commercial Enquiries

Enquiry volume provides an initial indication of identifiable commercial engagement.

However, enquiries differ in their potential value.

A specialist enterprise service provider may receive relatively few enquiries, but individual opportunities may involve substantial commercial commitments.

A consumer ecommerce business may receive many transactions with lower individual values.

Consequently, organisations should define qualification criteria appropriate to their business model.

Potential Qualification Criteria

• Relevance of the customer's requirement.
• Geographic eligibility.
• Product or service suitability.
• Available budget.
• Decision-making authority.
• Expected purchasing timeframe.
• Commercial opportunity size.

These criteria should be documented before performance comparisons are undertaken.

6.6 Qualified Lead Conversion Rate

A qualified lead conversion rate can help organisations examine the commercial relevance of identifiable enquiries.

Consider a hypothetical organisation receiving 40 enquiries associated with identifiable AI referrals.

If 12 enquiries meet the organisation's qualification criteria, the observed qualified lead rate is 30%.

This example is illustrative and does not represent an empirical CGO Media finding.

The calculation describes lead qualification within the defined sample.

It does not establish incremental customer acquisition.

6.7 Measuring Customer Acquisition

Customer acquisition measurement examines how many new customers are associated with a defined marketing activity.

For AI search, organisations may identify customers through recorded referrals, customer-reported discovery or other documented attribution evidence.

However, acquisition reporting should distinguish between attribution and causation.

A customer attributed to an AI referral under a particular model may still have purchased without that interaction.

Incremental acquisition requires additional evidence.

6.8 Customer Acquisition Cost

Customer acquisition cost is commonly calculated by dividing relevant acquisition expenditure by the number of acquired customers.

For AI search analysis, the calculation requires careful definition of both the costs and the customers included.

A programme-level acquisition cost should not be presented as an incremental AI-specific acquisition cost unless the evidence supports that distinction.

6.9 Attributed Revenue

Attributed revenue represents revenue assigned to particular interactions under a defined attribution methodology.

This may include transactions associated with identifiable AI referral sessions or customers whose discovery source was recorded.

However, attribution models may assign different amounts of revenue to the same interaction.

The results depend on the selected attribution rules.

Therefore, attributed revenue should always be accompanied by an explanation of the model used.

6.10 Why Revenue Is Not Profit

Revenue represents income generated through commercial activity.

Profit requires consideration of the relevant costs.

An organisation may attribute substantial revenue to a marketing programme while generating a much smaller financial contribution after delivery costs and other expenses.

Consequently, revenue attribution should not be treated as equivalent to profitability.

Commercial Contribution

For the purposes of this framework, commercial contribution refers to revenue less the variable or directly attributable costs associated with delivering the relevant products or services.

The precise definition should be consistent with the organisation's accounting and management reporting.

6.11 Return on Investment

Return on investment evaluates financial return relative to the resources committed.

However, different ROI calculations can produce materially different results.

Organisations should distinguish between attributed programme returns and causally established incremental returns.

The calculation requires a consistent definition of contribution and programme investment.

Programme investment must not be deducted twice when calculating the numerator.

6.12 Illustrative ROI Calculation

Consider a hypothetical organisation undertaking an AI search and GEO programme.

The calculation is:

((£60,000 - £30,000) ÷ £30,000) × 100 = 100%

This example is entirely hypothetical and is included to demonstrate the calculation.

The resulting 100% figure describes attributed programme performance under the stated assumptions.

It does not establish that the programme generated £30,000 of incremental profit.

Some or all of the attributed customers may have purchased without the programme.

6.13 Incremental ROI

Incremental ROI examines the additional financial contribution produced by a defined intervention.

It requires an estimate of what would have happened without the intervention.

This counterfactual cannot generally be observed directly.

It must be estimated through an appropriate research design.

The incremental contribution should reflect the estimated additional financial contribution associated with the intervention, before deducting the programme investment included in the denominator.

This avoids double counting programme expenditure.

6.14 The Counterfactual Problem

The central difficulty in incremental ROI measurement is determining what would have occurred without the relevant AI search activity.

For example, an organisation may observe increased enquiries after publishing a substantial research programme.

However, enquiry growth may also reflect advertising, seasonal demand, market conditions, referrals or conventional SEO improvements.

Without an appropriate comparison, the organisation cannot reliably isolate the contribution of the AI search programme.

This limitation should be recognised whenever financial performance claims are published.

6.15 Approaches to Incrementality Measurement

Organisations may investigate incremental effects through experimental or quasi-experimental research designs.

Controlled Experiments

Where feasible, comparable groups may receive different interventions.

Differences in subsequent outcomes can then be examined under the assumptions of the experimental design.

Matched Comparisons

Organisations may compare groups, locations or business units with similar characteristics.

However, differences between the groups may affect the interpretation of results.

Interrupted Time-Series Analysis

A sufficiently detailed series of observations may be examined before and after a defined intervention.

The analysis must consider underlying trends, seasonality and other changes occurring during the study period.

None of these methods automatically guarantees a valid causal conclusion.

Their suitability depends on the available data and research design.

6.16 Measuring AI Search in Long Sales Cycles

Commercial measurement becomes more complex when customer decisions extend over several months.

Enterprise technology, professional services, manufacturing and financial services may involve extended procurement processes.

A prospect may encounter an organisation through AI-assisted discovery long before a contract is signed.

During that period, the customer may interact with multiple channels and stakeholders.

Consequently, organisations should distinguish short-term lead indicators from eventual commercial outcomes.

Suggested Measurement Stages

1. Initial identifiable discovery.
2. Recorded enquiry.
3. Lead qualification.
4. Commercial opportunity creation.
5. Proposal or quotation.
6. Contract or transaction.
7. Recognised revenue.

Each stage should be recorded according to the organisation's established commercial processes.

6.17 Customer Lifetime Value

Customer lifetime value can provide additional context when evaluating acquisition economics.

Some organisations generate recurring revenue from long-term customer relationships.

Others generate revenue through individual transactions.

These differences affect the interpretation of customer acquisition costs and commercial returns.

However, projected customer lifetime value should not be confused with revenue already realised.

Forecasts should disclose the assumptions concerning retention, margins and future purchasing behaviour.

6.18 The Role of Gross Margin

Gross margin can materially affect the economic value of attributed revenue.

Two organisations generating identical revenue may produce very different financial contributions.

For example, a software business and a physical product retailer may have substantially different cost structures.

Consequently, comparing AI search ROI across industries requires caution.

Revenue alone provides insufficient information for determining economic value.

6.19 Opportunity Value Versus Realised Revenue

Organisations often record the potential value of commercial opportunities before contracts are completed.

Opportunity value can support pipeline management.

However, it should not be presented as realised revenue.

A proposal worth £100,000 does not establish that the organisation has generated £100,000 in revenue.

CGO Media recommends distinguishing:

• Enquiry volume.
• Qualified lead volume.
• Opportunity pipeline value.
• Contracted value.
• Recognised revenue.
• Financial contribution.

These indicators represent different stages of the commercial process.

6.20 Avoiding Double Counting Across Channels

AI-assisted discovery may overlap with conventional search, digital PR, content marketing and direct engagement.

A customer may encounter the same organisation through several channels.

If each channel claims the full transaction value, the combined attribution total may exceed the organisation's actual revenue.

Attribution reports should therefore explain whether channel figures are exclusive, overlapping or fractionally allocated.

Commercial outcomes should be deduplicated when producing organisation-wide totals.

6.21 Attribution Confidence and Financial Reporting

Financial reporting should reflect the quality of the available attribution evidence.

CGO Media recommends separating commercial outcomes into four evidence categories.

These categories should be reconciled to avoid counting the same commercial outcome more than once.

Category D should not be included in confirmed attributed revenue or ROI calculations.

6.22 Reporting Uncertainty

Commercial AI search measurement often involves incomplete information.

Some referral sources may be unidentified.

Customer-reported discovery may be incomplete.

Attribution models may depend on assumptions that cannot be fully verified.

Consequently, organisations should avoid presenting uncertain estimates with unjustified precision.

Where modelled results are used, appropriate uncertainty ranges or sensitivity analyses may be more informative than a single point estimate.

6.23 Sensitivity Analysis

Sensitivity analysis examines how results change when important assumptions are varied.

For example, an organisation may estimate the commercial contribution of customer-reported AI discovery.

The estimate may depend on assumptions about attribution credit, customer retention or delivery costs.

Testing alternative assumptions can reveal how sensitive the estimated ROI is to those choices.

However, sensitivity analysis does not establish that the underlying attribution assumptions are correct.

6.24 Commercial Measurement for Different Business Models

AI search measurement should reflect the commercial characteristics of the organisation.

Ecommerce

Relevant indicators may include transactions, order value, contribution margin and repeat purchasing.

Professional Services

Relevant indicators may include qualified enquiries, proposals, contract value and client retention.

SaaS

Relevant indicators may include qualified trials, subscriptions, recurring revenue and customer lifetime contribution.

Manufacturing

Relevant indicators may include technical enquiries, quotation requests, procurement opportunities and completed contracts.

Healthcare

Relevant indicators may include appropriate enquiries, service engagement and other outcomes consistent with applicable regulatory and privacy requirements.

A single commercial measurement model should not be imposed on every industry without adaptation.

6.25 The Proposed AI Search Commercial Performance Dashboard

CGO Media proposes a reporting dashboard that separates visibility, engagement, commercial outcomes and financial returns.

The dashboard should present each layer separately.

Organisations should avoid combining visibility rates, customer enquiries and revenue into a single score without an explicitly justified methodology.

6.26 Commercial Decision-Making

The purpose of commercial measurement is not simply to demonstrate positive results.

It is to improve organisational decision-making.

A credible measurement programme should help determine:

• Which activities require further investigation.
• Where measurement evidence is incomplete.
• Whether identifiable enquiries are commercially relevant.
• How investment compares with recorded outcomes.
• Whether additional research is justified.
• Whether available evidence supports continued investment.

Negative or inconclusive results may be commercially useful when they prevent unsupported investment decisions.

Section 6 — Research Interpretation

AI search commercial measurement requires a clear distinction between visibility, engagement, attribution and incremental economic value.

Citation frequency and recommendation visibility can describe observed discovery outcomes.

Referral analytics and CRM records can provide evidence of identifiable commercial interactions.

However, attributed revenue does not independently establish incremental revenue.

Similarly, revenue should not be confused with financial contribution or profit.

A credible AI search ROI calculation must define its costs, financial outcomes, attribution assumptions and evidential limitations.

Supporting CGO Media Research

AI Search ROI Research:
AI Search ROI Research UK 2026

AI Search Conversion Research:
AI Search Conversion Research UK 2026

Research Observations Library:
CGO Media Research Observations Library

Research Methodology:
CGO Media Research Methodology

The State of Search:
The State of Search

Section 6 — Key Finding

The commercial value of AI search cannot be established through visibility metrics alone.

Organisations require reliable cost information, identifiable commercial outcomes and transparent attribution methods.

Attributed revenue can support performance reporting, but incremental financial value requires stronger causal evidence.

AI search ROI should therefore be reported with clear definitions, documented assumptions and appropriate recognition of uncertainty.

7. Sector-Specific AI Search Measurement

AI-assisted discovery affects organisations operating across different industries, but the commercial significance of visibility varies according to the characteristics of each sector.

An ecommerce retailer may evaluate identifiable AI referrals through product engagement, completed transactions and contribution margins.

A manufacturing supplier may instead evaluate technical enquiries, supplier qualification, quotation requests and long-term procurement contracts.

A professional services organisation may measure qualified enquiries, consultations, proposals and client acquisition.

These differences demonstrate why AI search measurement requires sector-specific interpretation.

CGO Media proposes that the core measurement framework should remain consistent across industries while the selection and interpretation of commercial indicators reflect each sector's customer journey, regulatory environment and business model.

7.1 The Sector Measurement Principle

The seven visibility dimensions established in Section 3 can be investigated across different industries.

These include citation visibility, brand mentions, entity accuracy, provider comparisons, recommendations, visibility consistency and commercial relevance.

However, the relative importance of these dimensions may vary considerably.

For example, entity accuracy may be particularly consequential in regulated professional environments, while product suitability and availability may be especially relevant to ecommerce discovery.

Sector-specific measurement should therefore answer three questions.

These questions establish the foundation for adapting the framework to individual industries.

7.2 Financial Services

Financial services organisations operate within environments where trust, accuracy, regulatory requirements and commercial suitability are particularly important.

AI-assisted discovery may involve users researching financial products, comparing providers, examining eligibility requirements or seeking explanations of complex financial concepts.

The appearance of a financial services organisation within an AI-generated response should not be interpreted as evidence of regulatory approval, product suitability or customer preference.

Relevant Visibility Indicators

• Accuracy of organisational and regulatory information.
• Visibility within relevant product comparisons.
• Accuracy of product eligibility descriptions.
• Source citations in financial information responses.
• Recommendation qualifications and limitations.
• Consistency of information across repeated observations.

Commercial Measurement

Depending on the business model, relevant outcomes may include qualified enquiries, applications, completed onboarding and acquired customers.

Organisations should distinguish initial interest from completed applications and realised revenue.

Product eligibility, regulatory requirements and customer suitability may substantially affect the commercial journey.

Measurement Limitations

Financial decisions may involve several research stages and multiple providers.

Attribution should therefore avoid assuming that an AI-generated recommendation caused a subsequent application.

Sensitive financial information should be handled according to applicable privacy and regulatory requirements.

Explore CGO Media Financial Services AI & GEO Research

7.3 Ecommerce and Retail

Ecommerce and retail organisations may encounter AI-assisted discovery through product research, comparisons, recommendations and purchasing guidance.

Customer journeys may involve relatively short purchasing decisions or extended evaluation of expensive products.

Relevant Visibility Indicators

• Product mentions within relevant responses.
• Product recommendation visibility.
• Accuracy of product specifications.
• Brand visibility in category comparisons.
• Accuracy of pricing and availability information.
• Source citations supporting product information.

Commercial Measurement

Relevant outcomes may include identifiable product referrals, add-to-cart events, completed transactions, order values and contribution margins.

Where available, organisations may also examine repeat purchasing and customer acquisition costs.

However, revenue attributed to AI-related referrals should not automatically be interpreted as incremental revenue.

Measurement Limitations

Product availability, pricing, promotions and competing marketing activities may influence purchasing behaviour.

These factors should be considered when interpreting changes in commercial performance.

Explore CGO Media Ecommerce & Retail AI & GEO Research

7.4 Education and EdTech

Education and educational technology organisations may appear within AI-generated responses concerning courses, qualifications, institutions, learning platforms and educational suitability.

Discovery journeys may involve students, parents, employers, educators and institutional buyers.

Relevant Visibility Indicators

• Accuracy of course and programme descriptions.
• Institutional entity representation.
• Visibility in relevant educational comparisons.
• Accuracy of accreditation and qualification claims.
• Recommendation relevance for different learner requirements.
• Citation visibility for educational research and resources.

Commercial Measurement

Depending on the organisation, relevant outcomes may include course enquiries, applications, enrolments, qualified demonstrations and institutional contracts.

Applications should not be treated as completed enrolments.

Similarly, a software demonstration does not establish a completed institutional purchase.

Measurement Limitations

Educational decisions may involve long evaluation periods and multiple stakeholders.

Seasonal application cycles may also affect performance.

These factors should be considered when investigating relationships between AI visibility and enrolment.

Explore CGO Media Education & EdTech AI & GEO Research

7.5 Healthcare

Healthcare discovery involves particularly important considerations concerning information accuracy, patient safety, professional qualifications and regulatory compliance.

AI-generated responses may discuss healthcare organisations, clinical services, professional expertise and potential providers.

However, the appearance of a healthcare provider within an AI-generated response does not establish clinical suitability.

Relevant Visibility Indicators

• Accuracy of healthcare provider information.
• Accuracy of professional qualifications.
• Accuracy of service and location information.
• Appropriate citation of healthcare information.
• Presence of relevant qualifications and limitations.
• Consistency of organisational representation.

Commercial Measurement

Depending on the healthcare organisation, relevant operational indicators may include appropriate service enquiries, appointment requests and completed administrative engagement.

These indicators should be interpreted within the organisation's clinical, regulatory and ethical responsibilities.

Measurement Limitations

Healthcare measurement should not encourage unnecessary treatment or equate increased patient acquisition with improved clinical outcomes.

Patient information requires particularly careful protection.

Clinical suitability and patient welfare remain distinct from marketing performance.

Explore CGO Media Healthcare AI & GEO Research

7.6 International Organisations

International organisations may operate across multiple countries, languages and institutional environments.

Their discovery objectives may concern public information, research dissemination, stakeholder engagement, recruitment or institutional credibility.

Commercial revenue may not be an appropriate primary indicator.

Relevant Visibility Indicators

• Accuracy of institutional identity.
• Visibility of official publications.
• Accuracy of programme descriptions.
• Cross-language entity consistency.
• Source citation visibility.
• Geographic and institutional relevance.

Outcome Measurement

Relevant outcomes may include identifiable research engagement, qualified institutional enquiries, programme participation and other mission-aligned activities.

The selected indicators should reflect the organisation's actual objectives.

Measurement Limitations

Visibility should not be equated with institutional influence or policy impact.

Such outcomes require separate evidence.

Explore CGO Media International Organisations AI & GEO Research

7.7 Legal Services

Legal services discovery frequently involves professional expertise, jurisdiction, trust and the suitability of a provider for a particular matter.

AI-generated responses may explain legal subjects or identify potential legal service providers.

These are different forms of visibility.

Relevant Visibility Indicators

• Accuracy of professional and organisational information.
• Jurisdictional relevance.
• Visibility in appropriate practice-area comparisons.
• Accuracy of qualifications and service descriptions.
• Recommendation suitability for defined legal requirements.
• Citation visibility for relevant legal information.

Commercial Measurement

Relevant outcomes may include qualified enquiries, consultations, accepted instructions and completed engagements.

The appropriate measures depend on the firm's practice areas and commercial structure.

Measurement Limitations

Legal enquiries may contain confidential or sensitive information.

Attribution research should respect professional obligations and applicable data protection requirements.

The appearance of a legal provider in an AI response does not establish professional suitability for an individual matter.

Explore CGO Media Legal AI & GEO Research

7.8 Manufacturing

Manufacturing discovery may involve technical specifications, supplier capability, certifications, geographic suitability and procurement requirements.

AI-assisted research may help users investigate potential suppliers or understand industrial processes.

Relevant Visibility Indicators

• Technical capability representation.
• Accuracy of product and process information.
• Visibility within supplier comparisons.
• Accuracy of certifications and standards claims.
• Geographic supply capability.
• Recommendation relevance to technical requirements.

Commercial Measurement

Relevant outcomes may include technical enquiries, requests for quotation, supplier qualification, procurement opportunities and completed contracts.

Organisations should distinguish preliminary supplier research from qualified commercial opportunities.

Measurement Limitations

Industrial procurement can involve long sales cycles and multiple stakeholders.

A supplier may be researched months before receiving a formal quotation request.

Consequently, short-term website traffic may provide only a partial account of the purchasing journey.

Explore CGO Media Manufacturing AI & GEO Research

7.9 Professional Services

Professional services organisations may compete on specialist expertise, reputation, sector experience and commercial suitability.

AI-assisted discovery may involve researching professional problems, identifying relevant expertise and comparing potential advisers.

Relevant Visibility Indicators

• Specialist expertise representation.
• Professional entity accuracy.
• Visibility in relevant provider comparisons.
• Recommendation suitability.
• Citation visibility for published research.
• Consistency of sector-specific descriptions.

Commercial Measurement

Relevant outcomes may include qualified consultations, proposals, new client engagements, contract values and client retention.

The number of enquiries should be interpreted alongside their commercial quality.

Measurement Limitations

Professional services decisions may involve referrals, personal relationships and reputation developed over extended periods.

These influences may be difficult to isolate within conventional attribution systems.

Explore CGO Media Professional Services AI & GEO Research

7.10 Property and Real Estate

Property and real estate discovery may involve geographic requirements, property characteristics, market information and provider comparisons.

AI-generated responses may assist users researching locations, property services or potential providers.

Relevant Visibility Indicators

• Geographic relevance.
• Accuracy of property and service information.
• Visibility within relevant provider comparisons.
• Entity accuracy for agencies and developers.
• Recommendation relevance to customer requirements.
• Accuracy of market-related claims.

Commercial Measurement

Relevant outcomes may include qualified property enquiries, viewing requests, valuation enquiries, instructions and completed transactions.

These outcomes represent different stages of the property journey.

Measurement Limitations

Property transactions may involve substantial time between initial research and completion.

Market conditions, property availability and geographic factors can also affect commercial outcomes.

These influences should be considered when interpreting AI search attribution.

Explore CGO Media Property & Real Estate AI & GEO Research

7.11 SaaS

Software-as-a-Service organisations may encounter AI-assisted discovery through product research, technical comparisons and software recommendation questions.

Buyers may evaluate functionality, integrations, security, pricing and implementation requirements.

Relevant Visibility Indicators

• Product feature accuracy.
• Visibility in software comparisons.
• Recommendation relevance to user requirements.
• Accuracy of integration information.
• Technical documentation citations.
• Consistency of product positioning.

Commercial Measurement

Relevant outcomes may include trial registrations, qualified demonstrations, paid subscriptions, recurring revenue and customer retention.

Organisations should distinguish trial activity from paid customer acquisition.

Measurement Limitations

SaaS purchasing decisions may involve several technical and commercial stakeholders.

Product-led and sales-led acquisition models may require different measurement structures.

Explore CGO Media SaaS AI & GEO Research

7.12 Technology

Technology organisations may be evaluated according to technical capability, interoperability, security, scalability and commercial suitability.

AI-assisted discovery may involve complex technical questions and comparisons between alternative solutions.

Relevant Visibility Indicators

• Accuracy of technical capability descriptions.
• Visibility within solution comparisons.
• Recommendation relevance to technical requirements.
• Accuracy of integration and compatibility information.
• Citation visibility for technical documentation.
• Entity consistency across product categories.

Commercial Measurement

Relevant outcomes may include technical consultations, qualified demonstrations, procurement enquiries, proposals and completed contracts.

Measurement Limitations

Technology procurement may involve multiple decision-makers and extended evaluation processes.

Technical suitability should not be inferred solely from inclusion in an AI-generated recommendation.

Explore CGO Media Technology AI & GEO Research

7.13 Travel and Hospitality

Travel and hospitality discovery may involve destinations, accommodation, activities, availability, pricing and customer preferences.

AI-generated responses may help users compare destinations, accommodation providers or travel experiences.

Relevant Visibility Indicators

• Destination and property visibility.
• Accuracy of accommodation information.
• Recommendation relevance to traveller requirements.
• Geographic and seasonal suitability.
• Accuracy of amenities and service descriptions.
• Visibility in relevant travel comparisons.

Commercial Measurement

Relevant outcomes may include identifiable booking referrals, reservation enquiries, completed bookings, booking values and contribution margins.

Organisations should distinguish booking enquiries from confirmed transactions.

Measurement Limitations

Travel demand is influenced by seasonality, pricing, availability and wider market conditions.

These factors should be considered when interpreting changes in AI search performance.

Explore CGO Media Travel & Hospitality AI & GEO Research

7.14 Comparing Measurement Priorities Across Sectors

The twelve sectors demonstrate that AI search visibility has no single universal commercial interpretation.

Different industries require different definitions of qualified engagement, commercial opportunity and financial value.

7.15 Why Cross-Sector AI Visibility Rankings Can Mislead

Comparing raw AI visibility percentages across unrelated industries can produce misleading conclusions.

A consumer retailer may appear frequently in broad product recommendation questions.

A specialist manufacturing supplier may appear only in a narrow range of highly technical enquiries.

The two organisations operate within fundamentally different discovery environments.

Their visibility percentages should not be compared without accounting for prompt selection, commercial relevance and research scope.

A lower appearance rate within a highly specialised commercial sample may be more relevant to the organisation's objectives than frequent appearances in unrelated informational responses.

However, commercial value must still be established through appropriate evidence.

7.16 Sector-Specific Prompt Libraries

CGO Media recommends developing separate prompt libraries for each industry.

Each library should reflect the relevant customer journey and commercial discovery requirements.

Recommended Prompt Categories

• Informational discovery.
• Problem-oriented research.
• Specialist requirements.
• Product or service comparisons.
• Provider discovery.
• Commercial suitability.
• Geographic relevance.

Prompt libraries should be defined before observations are collected.

This reduces the risk of selecting questions primarily because they produce favourable results.

7.17 Sector-Specific Attribution Windows

Attribution windows should reflect the characteristics of the purchasing process.

A relatively straightforward consumer transaction may involve a shorter decision period than enterprise procurement.

However, no universal attribution window should be assumed for an entire industry.

Organisations should examine their own customer journey data and document the attribution windows selected.

A longer attribution window does not automatically establish stronger causal evidence.

7.18 Sector-Specific Commercial Value

The financial significance of a conversion depends on the organisation's business model.

A completed ecommerce transaction, a SaaS subscription and a manufacturing procurement contract may have substantially different revenue and cost structures.

Consequently, sector comparisons should distinguish between:

• Conversion volume.
• Average transaction or contract value.
• Contribution margin.
• Customer acquisition cost.
• Customer retention.
• Realised revenue.
• Incremental economic contribution.

These indicators should be selected according to the organisation's actual commercial objectives.

7.19 Regulatory and Ethical Differences

Sector-specific measurement must also account for regulatory and ethical responsibilities.

Financial services, healthcare and legal organisations may operate under particularly demanding professional and information governance requirements.

Measurement activities should not compromise customer confidentiality, professional obligations or applicable data protection rules.

Organisations should avoid collecting unnecessary sensitive information merely to improve marketing attribution.

The availability of a measurement technique does not establish that its use is appropriate in every context.

7.20 The Proposed Sector Measurement Matrix

CGO Media proposes a sector measurement matrix containing five analytical components.

The matrix provides a consistent structure for sector-specific research without imposing identical commercial metrics across unrelated industries.

7.21 Implications for Enterprise Reporting

Enterprise organisations operating across several industries or business units may require both centralised and sector-specific reporting.

A central reporting structure can maintain consistent metric definitions and evidence classifications.

Individual business units can then report outcomes relevant to their commercial responsibilities.

This approach supports organisational consistency without obscuring important differences between business models.

Section 7 — Research Interpretation

AI search measurement requires sector-specific interpretation.

Although citations, brand mentions, entity accuracy and recommendation visibility can be investigated across industries, their commercial significance varies.

Customer journeys, purchasing requirements, regulatory obligations and financial structures influence which outcomes should be measured.

A universal commercial performance score would risk concealing these differences.

AI search measurement should retain consistent research principles while adapting commercial indicators to the characteristics of each sector.

Supporting CGO Media Research

Sector Research:
CGO Media Research Library

Framework Library:
CGO Media Framework Library

Research Observations:
CGO Media Research Observations Library

Research Methodology:
CGO Media Research Methodology

Section 7 — Key Finding

The commercial significance of AI search visibility depends on industry context, customer requirements and the nature of the purchasing journey.

Sector-specific research should combine consistent visibility definitions with appropriate commercial measures.

Cross-sector comparisons should account for differences in prompt samples, attribution evidence and business models.

The objective is not to produce identical performance indicators for every industry, but to establish credible measurement within each commercial environment.

8. Conclusions, Research Methodology and Future Research

AI-assisted discovery introduces new challenges for organisations attempting to measure search visibility, customer acquisition and commercial performance.

Traditional search engine optimisation has historically relied on rankings, impressions, clicks, website engagement, conversions and revenue attribution.

These indicators remain important. However, AI-generated responses can introduce organisations, products, services and research publications to users without necessarily producing an identifiable website visit.

Consequently, conventional website analytics cannot independently describe every observable form of AI-assisted discovery.

This research paper has examined that measurement problem and proposed the CGO Media AI Search Measurement and Attribution Framework as a structured analytical approach.

The framework distinguishes observable AI visibility, identifiable commercial attribution and potential influence that remains unverified.

Its central principle is that measurement claims must remain proportionate to the available evidence.

8.1 Principal Research Conclusions

The analysis developed throughout this publication supports eight principal conclusions.

8.2 The CGO Media AI Search Measurement and Attribution Framework

The framework proposed in this publication contains three principal evidence categories.

8.3 The Seven Visibility Dimensions

Within observable AI visibility, the framework identifies seven complementary measurement dimensions.

1. AI Citation Visibility: The presence of qualifying citations within eligible AI-generated responses.
2. Brand Mention Visibility: The presence of identifiable organisational mentions within relevant responses.
3. Entity Representation Accuracy: The accuracy of verifiable claims concerning an organisation.
4. Provider Comparison Visibility: The inclusion of an organisation within relevant provider comparisons.
5. Recommendation Visibility: The presentation of an organisation as potentially suitable for a defined requirement.
6. Visibility Consistency: The repeatability of observed appearances under defined research conditions.
7. Commercial Relevance: The relationship between observed appearances and relevant customer requirements.

These dimensions should normally be reported separately.

A composite score should only be introduced where its weighting, assumptions and interpretation are explicitly documented.

8.4 Research Methodology

This publication is a conceptual and methodological research paper.

It develops a proposed analytical framework for measuring AI-assisted discovery by examining information retrieval evaluation, citation assessment, marketing attribution and causal measurement principles.

The publication does not present a completed large-scale empirical experiment or claim to have measured the behaviour of all AI-assisted discovery platforms.

8.4.1 Research Objectives

• Identify limitations of conventional SEO metrics within AI-assisted discovery.
• Establish distinct categories of observable AI search visibility.
• Propose reproducible observational measurement methods.
• Distinguish visibility from commercial attribution and causation.
• Develop principles for evaluating financial performance and ROI.
• Identify sector-specific measurement requirements.
• Establish priorities for future empirical research.

8.4.2 Analytical Approach

The research applies conceptual analysis to the relationship between search exposure, AI-generated responses, website engagement and commercial outcomes.

Particular attention is given to the distinction between directly observed events and inferred commercial influence.

The framework draws upon established information retrieval evaluation principles and research into marketing attribution and causal measurement [3][4][5].

8.4.3 Hypothetical Examples

Numerical examples throughout this publication are illustrative.

They demonstrate proposed calculations and measurement interpretations.

They do not represent empirical findings from a completed CGO Media experiment.

8.4.4 Evidence Classification

The proposed framework distinguishes:

• Documented observations.
• Supported attribution relationships.
• Modelled estimates.
• Unverified potential influence.

These categories are intended to reduce the risk of unsupported conclusions about the commercial effects of AI-assisted discovery.

8.4.5 Reproducibility

Future empirical applications should document their research conditions.

These include the prompts tested, platforms examined, observation dates, eligibility criteria, classification methods and relevant limitations.

Where practical, research protocols and suitably anonymised datasets should be made available to support independent assessment.

Public release of research records must respect privacy requirements, intellectual property obligations and applicable platform terms.

8.5 Research Limitations

8.5.1 Limited Access to Private AI Interactions

External researchers generally cannot observe every private interaction occurring within AI-assisted discovery environments.

Structured prompt testing therefore provides observations from a defined research sample rather than a complete representation of real-world usage.

8.5.2 Response Variability

AI-generated responses may vary between repeated observations.

Repeated testing can help describe observed variability but cannot eliminate all uncertainty.

8.5.3 Platform Differences

AI platforms differ in their interfaces, retrieval capabilities, citation presentation and available information.

Measurement definitions must account for platform-specific characteristics.

8.5.4 Prompt Selection Bias

The selection and wording of research prompts can influence observed visibility.

Results obtained from selected prompts should not automatically be generalised to all users.

8.5.5 Attribution Data Gaps

Some AI-assisted discovery interactions may not produce identifiable website referrals.

Others may be followed by branded search, direct visits or offline customer contact.

These interactions may remain unobservable within conventional analytics systems.

8.5.6 Causal Uncertainty

A relationship between observed AI visibility and commercial performance does not independently establish causation.

Research into causal impact measurement demonstrates the importance of estimating a counterfactual outcome when evaluating the effects of an intervention [5].

8.5.7 Sector Differences

Different industries have different customer journeys, commercial structures and regulatory requirements.

Measurement approaches may therefore require adaptation between sectors.

8.5.8 Changes in AI Search Systems

AI-assisted discovery technologies continue to evolve.

Changes in platform capabilities, retrieval methods, user interfaces and reporting systems may affect the practical application of the proposed framework.

Measurement procedures should therefore be reviewed periodically.

8.6 Future Research Priorities

8.7 Implications for Organisations

Organisations seeking to evaluate AI search performance should avoid replacing established SEO reporting with a single AI visibility score.

Instead, measurement should be developed as a complementary analytical capability.

Traditional search performance should continue to be evaluated through appropriate rankings, impressions, engagement and conversion data.

AI-assisted discovery should be examined through structured observational research.

Commercial attribution should be based on identifiable evidence and documented models.

Claims concerning incremental financial returns should be supported by suitable causal evaluation methods.

8.8 Implications for SEO and GEO Agencies

SEO and GEO agencies should distinguish between reporting observable visibility and claiming commercial results.

A credible measurement programme should explain which AI platforms were examined, how prompts were selected and how responses were classified.

Reports should identify the observation period, attribution methodology and relevant limitations.

Where commercial outcomes are attributed to AI search, the supporting evidence should be clearly described.

Agencies should avoid presenting estimated influence as confirmed incremental revenue.

8.9 Implications for Researchers

Researchers investigating AI-assisted discovery should distinguish observational, inferential and causal research.

Observational studies can document how organisations appear within defined response samples.

Inferential studies may attempt to estimate broader patterns where appropriate sampling methods are available.

Causal studies require designs capable of evaluating the effects of defined interventions.

These approaches should not be treated as interchangeable.

Transparent methodology, reproducibility and explicit recognition of uncertainty are essential to credible AI search research.

8.10 Final Research Conclusion

The development of AI-assisted discovery requires organisations to reconsider how search performance is measured.

Traditional SEO indicators remain important, but they do not independently describe every form of visibility within AI-generated responses.

Citations, brand mentions, entity descriptions and recommendations introduce additional observable outcomes.

However, these outcomes should not be confused with customer acquisition or financial return.

Commercial measurement requires identifiable engagement, appropriate attribution evidence and a clear distinction between attributed and incremental value.

8.11 External Academic Bibliography

The following external references provide methodological foundations for generative search visibility, citation evaluation, information retrieval measurement, marketing attribution and causal analysis.

The bibliography distinguishes peer-reviewed research from technical documentation.

These publications inform the development of the proposed CGO Media framework but do not constitute independent empirical validation of its specific metrics.

8.11.1 Generative Engine Optimisation

8.11.2 Citation Evaluation and Source Reliability

8.11.3 Information Retrieval Evaluation

8.11.4 Marketing Attribution

8.11.5 Causal Impact and Incrementality

8.11.6 Official Analytics Documentation

8.12 Interpretation of External Evidence

The external publications cited above provide relevant research foundations for AI search visibility, citation evaluation, information retrieval measurement and commercial attribution.

However, they do not independently demonstrate that the proposed CGO Media measurement framework has been empirically validated.

The framework should therefore be understood as a methodological research proposal suitable for further observational and experimental evaluation.

The bibliography includes peer-reviewed academic publications and official technical documentation.

These source types serve different purposes and should not be treated as equivalent evidence.

External research supports the methodological foundations of this publication, while the specific CGO Media measurement categories remain proposed analytical contributions.

8.13 Supporting CGO Media Research

This publication forms part of the wider CGO Media research programme examining SEO, AI-assisted discovery, entity authority, citations, recommendations and commercial performance.

Research Libraries

• CGO Media Research Library
• CGO Media Framework Library
• CGO Media Research Observations Library
• CGO Media Statistics Library

Related Research Publications

• AI Search ROI Research UK 2026
• AI Search Conversion Research UK 2026
• AI Citation Selection in Generative Search
• AI Recommendation Authority in Generative Search
• AI Source Selection in Generative Search
• The State of Search

Methodology and Editorial Resources

• CGO Media Research Methodology
• Latest CGO Media Research
• Press and Media Resources

8.14 Research Disclosure

8.15 About the Author

8.16 Suggested Citation

8.17 Publication Information

Research Title:
AI Search Measurement and Attribution: Beyond Rankings and Traffic

Research Category:
AI Search · Generative Engine Optimisation · Search Measurement · Attribution · AI Visibility · Commercial Performance · Research Methodology

Author:
Roger Wilkinson

Publisher:
CGO Media

Publication Year:
2026

Research Type:
Conceptual and Methodological Research

Last Reviewed:
12 October 2026

8.18 Final Research Statement