How to Measure AI Search Visibility

AI search visibility is not one metric

Traditional SEO reporting follows a familiar path: a page ranks, earns an impression, receives a click and contributes to a conversion.

AI search is less linear. A generated answer can use several sources, mention a company without linking to it or link to a page that did not rank first in classic search. Google also explains that AI Overviews and AI Mode may use query fan-out, running several related searches across subtopics before producing an answer.

This creates three distinct measurement layers:

‍

Combining these into one visibility score usually makes the report easier to present but harder to act on. An impression, a citation and a qualified enquiry are not interchangeable.

Google measures appearances in AI Overviews and AI Mode

Google introduced dedicated generative AI performance reports in Search Console in June 2026. The Search report includes impressions from AI Overviews and AI Mode and can be analysed by page, country, device and date.

An impression means that a link to the site was shown in a supported generative AI feature. If two results from the same property appear in one AI experience, Google counts one property-level impression in the chart. Page-level totals can differ because the table uses a different aggregation method.

Start with three views:

1. Total impressions over time. Compare monthly periods to see whether exposure is growing, stable or declining.
2. Pages receiving impressions. Check whether visibility relies on one informational article or extends to product, service, comparison and proof pages.
3. Country and device. Confirm that exposure comes from markets and browsing contexts that matter to the business.

Use longer reporting periods when volumes are small. Daily movement can reflect low volume, reporting delays or changing demand rather than a meaningful change in visibility.

The report's limits matter. It currently focuses on impressions. It does not provide query, click, click-through-rate or conversion data inside the dedicated generative AI view. It can tell you that a page appeared, but not what the buyer asked, how prominent the link was or what happened next.

Google says it had rolled out the reports to all websites worldwide by 31 August 2026. If the report does not appear for a property, the website may not have received enough impressions in Google’s generative AI features.

Google also added multimodal search reporting in September 2026. A separate filter now covers searches made through Lens, Circle to Search, image uploads and Chrome’s “Search this image” feature. This data appears in both the standard Search results report and the Generative AI performance report.

Microsoft measures citations, not Google-style impressions

Bing Webmaster Tools' AI Performance report covers supported experiences including Microsoft Copilot, AI-generated summaries in Bing and selected partner integrations.

It reports total citations, average cited pages, sampled grounding queries, page-level citation activity and trends. A citation means that a page was displayed as a source in an AI-generated answer. Microsoft warns that the count does not reveal ranking, authority, placement or the role a page played in the answer.

Microsoft Clarity’s Citations dashboard reports page citations, grounding queries, Share of Authority and AI referral traffic. Share of Authority includes only query-days on which the domain received a citation, so it should not be read as overall market share across AI search.

A Google impression and a Microsoft citation are different events. Adding them together produces a larger number with less meaning. Compare each platform against its own earlier baseline.

In October 2026, Bing added a Topics Portfolio beta to its AI Performance report. It groups grounding queries into broader topics and shows approximate topic demand, competing sources and queries where other websites are cited but yours is not. Microsoft says these figures should be treated as directional planning signals rather than precise measurements.

A controlled query sample reveals the missing context

First-party reports show exposure and citations, but they do not fully explain how the company appears inside an answer. A controlled query sample fills that gap.

Choose 20 to 40 questions based on real customer research. Use questions from sales calls, support conversations, Search Console and product research. Include different stages of the buying process, such as problem discovery, solution evaluation, commercial comparison, provider discovery and branded due diligence.

For each question, record:

• platform, market, date and device
• whether an AI answer appeared
• whether the brand was mentioned or linked
• which page was used as a source
• whether the description was accurate
• which competitors or alternative sources appeared

Keep the wording, location and account state as consistent as possible. Repeat the sample monthly. Generated answers can vary by platform, model, wording, location and time, so one screenshot is an observation, not a trend.

This review also prevents a common reporting error: treating every citation as positive. A page may be cited only to define a general term. A company may appear in a comparison but be described inaccurately. Visibility is commercially useful only when the system connects the brand to the right problem, capability and audience.

Analytics shows what happens after the click

ChatGPT visits can also be identified when a user clicks through to the website. OpenAI adds utm_source=chatgpt.com to referral URLs, allowing this traffic to be separated in tools such as Google Analytics. This still captures only visits. It does not reveal how often the company was mentioned or cited without a click.

Search Console measures visibility in Google Search. Website analytics measures what visitors do after they arrive.

Google states that appearances in AI Overviews and AI Mode remain included in the standard Search Console Performance report under the Web search type. That report provides wider context through clicks, queries, click-through rate and average position, but it does not isolate every AI-driven interaction.

Create a page group containing URLs that receive generative AI impressions. Then review their wider organic performance and on-site behaviour:

• engaged sessions
• progression to relevant product, service or case-study pages
• qualified form submissions or account creation
• assisted conversions
• lead quality reported by sales

Treat this as contextual evidence, not exact attribution. A click to a visible page is not automatically an AI Mode click. Visibility without a click will not appear in analytics, and a buyer may return later through branded search or direct traffic.

For longer sales cycles, combine analytics with CRM notes and a simple question such as “How did you first hear about us?” This will not create perfect attribution, but it can reveal patterns that last-click reporting misses.

Report a scorecard instead of a universal visibility score

A useful monthly report can stay compact:

‍

Do not combine these measures unless the weighting reflects an actual business decision. Ten extra impressions and one qualified enquiry should not be treated as equivalent units.

The data should change the website backlog

Reporting is useful only when it leads to a decision.

If a commercially important page receives visibility but visitors do not continue, review the page as a user journey. Does it answer the question that brought the visitor there? Is the offer explicit? Can the reader verify the claims? Is the next useful action clear?

If an important topic receives no visibility, check the foundations before rewriting it. Confirm that the page can be crawled and indexed, that important information is available as text, that internal links make it discoverable and that the subject is covered clearly enough to be useful.

If a page is cited for the wrong idea, reduce ambiguity. Clarify the company, audience, capability and supporting evidence. Use consistent language across product pages, case studies and credible external profiles.

Google confirms that there are no special technical requirements, AI text files or dedicated schema needed for AI Overviews and AI Mode. The same search foundations still matter: crawlability, indexability, internal links, useful text and structured data that matches visible content.

Our guide to the B2B website as a sales funnel explains how each page can support the buyer’s next decision.

What this means for your team

Marketing should own the reporting baseline, buyer questions and commercial interpretation. Content teams should examine which topics and page types earn visibility, then improve missing evidence or unclear explanations. Web and SEO teams should protect access, indexing, internal linking and measurement. Sales should report whether AI-informed prospects arrive with different questions or stronger prior knowledge.

Start with the first-party reports already available. Build a baseline before paying for a complex visibility platform. Once the team knows which decisions the data must support, it can judge whether another tool adds useful coverage or simply produces a polished score.

AI search visibility is measurable, but only in pieces. Keep those pieces separate long enough to understand them, then connect them through the website journey and the outcomes that matter to the business.

If your team can see AI visibility but cannot turn it into clear website priorities, Clustr Studio can audit the content structure, user journey and technical foundations, then translate the findings into a focused optimisation or redesign roadmap.

AI search visibility cannot be measured with one number. The useful approach is to track three layers separately: exposure in AI features, how the brand is represented in generated answers and what visitors do after reaching the website.

Google Search Console now reports impressions from AI Overviews and AI Mode. Microsoft reports citations across supported AI experiences. Analytics can show visits and conversions. None of these signals tells the whole story alone.

AI search visibility is not one metric

Traditional SEO reporting follows a familiar path: a page ranks, earns an impression, receives a click and contributes to a conversion.

AI search is less linear. A generated answer can use several sources, mention a company without linking to it or link to a page that did not rank first in classic search. Google also explains that AI Overviews and AI Mode may use query fan-out, running several related searches across subtopics before producing an answer.

This creates three distinct measurement layers:

‍

Combining these into one visibility score usually makes the report easier to present but harder to act on. An impression, a citation and a qualified enquiry are not interchangeable.

Google measures appearances in AI Overviews and AI Mode

Google introduced dedicated generative AI performance reports in Search Console in June 2026. The Search report includes impressions from AI Overviews and AI Mode and can be analysed by page, country, device and date.

An impression means that a link to the website was shown in a supported generative AI feature. If two results from the same property appear in one AI experience, Google counts one property-level impression in the chart. Page-level totals can differ because the table uses a different aggregation method.

Start with three views:

1. Total impressions over time. Compare monthly periods to see whether exposure is growing, stable or declining.
2. Pages receiving impressions. Check whether visibility relies on one informational article or extends to product, service, comparison and proof pages.
3. Country and device. Confirm that exposure comes from markets and browsing contexts that matter to the business.

Use longer reporting periods when volumes are small. Daily movement can reflect low volume, reporting delays or changing demand rather than a meaningful change in visibility.

The report's limits matter. It currently focuses on impressions. It does not provide query, click, click-through-rate or conversion data inside the dedicated generative AI view. It can tell you that a page appeared, but not what the buyer asked, how prominent the link was or what happened next.

Google says the reports had rolled out to websites worldwide by 31 August 2026. However, its current help page still lists phased rollout and insufficient impression volume as possible reasons a property may not see the report. In practice, absence of the report should not be treated as proof of zero visibility.

Microsoft measures citations, not Google-style impressions

Bing Webmaster Tools' AI Performance report covers supported experiences including Microsoft Copilot, AI-generated summaries in Bing and selected partner integrations.

It reports total citations, average cited pages, sampled grounding queries, page-level citation activity and trends. A citation means that a page was displayed as a source in an AI-generated answer. Microsoft warns that the count does not reveal ranking, authority, placement or the role a page played in the answer.

Microsoft Clarity's Citations dashboard adds eligible query volume, citation rate, page citations, grounding queries and Share of Authority. Microsoft describes this data as representative rather than exhaustive. Low volumes may be excluded, and the dashboard is intended for trend analysis rather than a precise record of every answer.

A Google impression and a Microsoft citation are different events. Adding them together produces a larger number with less meaning. Compare each platform against its own earlier baseline.

A controlled query sample reveals the missing context

First-party reports show exposure and citations, but they do not fully explain how the company appears inside an answer. A controlled query sample fills that gap.

Choose 20 to 40 questions based on real customer research. Use questions from sales calls, support conversations, Search Console and product research. Include different stages of the buying process, such as problem discovery, solution evaluation, commercial comparison, provider discovery and branded due diligence.

For each question, record:

• platform, market, date and device
• whether an AI answer appeared
• whether the brand was mentioned or linked
• which page was used as a source
• whether the description was accurate
• which competitors or alternative sources appeared

Keep the wording, location and account state as consistent as possible. Repeat the sample monthly. Generated answers can vary by platform, model, wording, location and time, so one screenshot is an observation, not a trend.

This review also prevents a common reporting error: treating every citation as positive. A page may be cited only to define a general term. A company may appear in a comparison but be described inaccurately. Visibility is commercially useful only when the system connects the brand to the right problem, capability and audience.

Analytics shows what happens after the click

Search Console measures visibility in Google Search. Website analytics measures what visitors do after they arrive.

Google states that appearances in AI Overviews and AI Mode remain included in the standard Search Console Performance report under the Web search type. That report provides wider context through clicks, queries, click-through rate and average position, but it does not isolate every AI-driven interaction.

Create a page group containing URLs that receive generative AI impressions. Then review their wider organic performance and on-site behaviour:

• engaged sessions
• progression to relevant product, service or case-study pages
• qualified form submissions or account creation
• assisted conversions
• lead quality reported by sales

Treat this as contextual evidence, not exact attribution. A click to a visible page is not automatically an AI Mode click. Visibility without a click will not appear in analytics, and a buyer may return later through branded search or direct traffic.