Google and Bing both give site owners a free report on how often their pages appear in AI answers. Neither report shows why a page is missing. That question has to be answered somewhere else.
Both reports are worth setting up, and both are easy to over-read.
What the two free reports show
Google's generative AI performance report in Search Console launched on 3 June 2026 and has been open to all websites worldwide since 31 August 2026. It counts impressions: how often links to your site appeared in AI Overviews and AI Mode. You can split the count by page, country, device and date. Google's documentation lists no clicks and no queries for it.
Bing Webmaster Tools added AI Performance in public preview on 10 February 2026. It covers Microsoft Copilot, AI-generated summaries in Bing and select partner integrations that Microsoft does not name. It counts citations and shows which pages earned them. It also lists the phrases the AI searched for to find your content, which Microsoft calls grounding queries and describes as a sample of the whole.
On 16 June 2026 Microsoft added four preview views: Intents, Topics, Citation Share and Compare. Citation Share gives your percentage of the citations for a query, out of every site cited for that query.
Neither report names ChatGPT, Claude or Perplexity as a source of its data. For those, you ask the buying questions yourself and log what comes back. We cover tools that do the logging in what marketing leaders use to monitor brand visibility in ChatGPT and Perplexity.
What a report cannot see
A report counts outcomes. Google's counts impressions and Bing's counts citations, so each tells you whether a page appeared. Neither shows what a model already knows about your firm or how other sites describe it.
Microsoft does not claim otherwise. Its June announcement does not say why a particular page is left out. It says citation activity can be influenced by many factors, including evolving models, competing content, freshness signals and shifts in user demand, and it does not say which applies to a given page.
Where the causes sit
Two studies point at the same place: what a model already knows about a brand, and who else names it.
Chen and Koudas at the University of Toronto replayed questions through ChatGPT's API and tested which brands the model knew with search switched off. ChatGPT wrote a brand it knew into its own searches 67% of the time. For a brand it did not know, it did so 11% of the time.
The paper is a preprint and has not been peer reviewed. It used small models and US English questions, and it shows an association.
Ahrefs studied 75,000 brands and found that branded web mentions correlate at 0.664 with a brand's visibility in Google's AI Overviews. Backlinks correlate at 0.218. Ahrefs says correlation is not causation. The study covers AI Overviews only and is limited to established sites. It leaves out the roughly 26% of brands with no AI Overview mentions, and it comes from a company that sells SEO tools.
Neither figure appears in either dashboard. Both describe what happens before a page gets a chance to be cited.
One site, three kinds of question
In a citation study published on 29 October 2026, Ahrefs tracked 22 questions about its own brand between 28 August and 4 October. On Google's AI tools, its own site was cited on 97-100% of days for factual questions such as pricing. For review questions the figure was 5%. For comparison questions it was 0-5%.
ChatGPT behaved differently. It cited Ahrefs' site on 97% of days for reviews and 71% for comparisons.
On Google's tools the site did not change between those questions. The domain was the same and so was its authority. What changed was how much of the answer depended on other people. A report that counts citations per page records the pricing page doing well and the comparison pages missing. It does not record who writes the reviews and comparisons.
The explanation is my reading of the numbers, not a finding in the study. The split rests on 22 questions about one brand, and Ahrefs sells SEO tools.
What upstream means
A model names a firm from what it can retrieve and what it already holds. Both depend on how the firm is described. Positional clarity means a buyer, or a model, can say in one sentence what the firm is known for. Consistency of meaning means the firm's own site and the pages other people write about it say the same thing.
Structural retrievability sits on your own pages. It is whether a model that goes looking can find a plain statement of what the firm does, and for whom.
The reports earn their place. They give you a free baseline and show which pages already get shown. They cannot tell you whether the cause sits on your pages or elsewhere.
Where this argument is weaker
The cause is not only on your own pages. Both studies above point to what others say about a brand.
Both reports are young. Google's launched in June and Microsoft's in February, and Microsoft calls its four June additions early previews. The studies behind the figures cover US English questions, and the Toronto one used small models that can behave differently from the apps people use.
A zero in either report is a question about how the market describes your firm. The report only tells you the question exists.
A few common follow-up questions
Does Google's report show clicks from AI Overviews?
The report shows impressions, with breakdowns by page, country, device and date. Google's documentation for it lists no clicks. Google's announcement says the same data also stays in the overall Performance report.
Does Bing's report cover ChatGPT?
Microsoft's announcement names Microsoft's own AI answers and unnamed partner integrations. ChatGPT, Claude and Perplexity are not named.
Do I need both reports?
They measure different things. Google counts impressions and Bing counts citations, so the two numbers cannot be added together.
Is GEO different from AEO?
In my work, both terms name the same job: getting a firm named in AI answers. Microsoft calls its report an early step toward GEO tooling. I use AEO, and the label does not change what the reports can tell you.