What Google and Bing's AI reports measure

First-party AI visibility data is reporting a search platform gives you about your own site's appearances in its AI answers, and as of mid-2026 there are two of them: Google's generative AI performance reports in Search Console and Bing Webmaster Tools' Citation Share. They landed within two weeks of each other, they are both useful, and they answer different questions. Treating them as two views of one number is the fastest way to build a misleading dashboard.

#What Google's report actually contains

Google announced generative AI performance reports for Search and Discover in June 2026. The Search version covers two surfaces, AI Overviews and AI Mode, and it carries exactly one metric: impressions. Per Google's own documentation, an impression here means links to your site were shown to a user in a generative AI feature. There are no clicks, no CTR, no average position, and no query data.

You can group and filter by pages, countries, devices, and dates. Pages are grouped by final URL after redirects, using canonical URLs, which matters if your canonical strategy is loose. Data from Search Labs experiments is excluded because those experiments are still in active development. The normal Search Console constraints carry over too, including the 1,000-row limit and the usual preliminary-data lag on the most recent days.

The absence of query data is the part people keep waiting on. Prompts in AI Mode are long, personal, and often reformulated by the system before retrieval happens, so there is no clean "query" to attribute an impression to in the way a ten-word search string allows. Do not plan a workflow that assumes it is arriving next quarter.

#The number is a subset, not an addition

This is the detail that quietly breaks reports. Google's documentation on AI features and your website states that sites appearing in AI features such as AI Overviews and AI Mode are included in the overall search traffic in Search Console, reported in the Performance report within the "Web" search type.

Those impressions were always in your totals. The new report does not add a stream, it carves out a view of one you already had. Add the generative AI impression count to your Web search impressions and you have double-counted every one of them. Search Engine Journal's read of the launch makes the same warning about blended reporting, and it applies to derived metrics too: you cannot compute a meaningful blended CTR when one half of the numerator does not exist.

The practical framing is a ratio, not a sum. Generative AI impressions divided by Web search impressions tells you what share of your Google visibility now runs through an AI surface. That number moving is a real signal. The two numbers side by side, labelled as a total, is not.

#What Bing's Citation Share adds

Microsoft shipped Citation Share on June 16, 2026, alongside three other previews: Intents, Topics, and Compare. Microsoft defines Citation Share as the percentage of citations attributed to your site out of all citations shown across all sites for that same grounding query.

That definition is doing a lot of work. The denominator is not your site's activity. It is the full set of cited sources for a query, which means Citation Share moves when your competitors' coverage changes even if nothing on your side does. It is the closest thing either platform ships to a competitive metric, and Microsoft is careful about the boundary: the company describes it as an observational metric, not a ranking system or a competitive scoreboard, and says it does not expose competitor domains, represent traffic share, or assign quality scores. The reporting covers citations across Microsoft Copilot, Bing, and select partner AI experiences.

Microsoft also names the volatility directly, listing user behavior, evolving models, freshness signals, partner refresh cycles, and broader changes across the web as reasons citation patterns shift. Read that as an instruction about sample size. A single week of Citation Share on a low-volume query is noise.

#Two denominators, two questions

The cleanest way to hold both metrics is to name what each one divides by.

Report	Unit	Denominator	Question it answers
Search Console generative AI	Impression	Your own site's activity	How much of my Google visibility is happening in AI surfaces?
Bing Citation Share	Percentage	All cited sites for that grounding query	How much of the answer's source set is mine?

Google's report is an absolute count with an internal denominator, so it grows when you publish more and shrinks when a surface stops firing. Bing's is a relative share with an external denominator, so it can fall while your citations rise, simply because the candidate pool got bigger. These will disagree, and the disagreement is information rather than an error to reconcile.

#The blind spot neither one covers

Both reports stop at their own platform. ChatGPT, Claude, and Perplexity publish no equivalent for site owners. OpenAI's bot documentation covers the mechanics of being eligible, stating that OAI-SearchBot surfaces websites in ChatGPT's search features and that sites opted out of it will not be shown in ChatGPT search answers, and recommending you allow it in robots.txt. It is access control documentation, not analytics. There is no dashboard behind it.

So the coverage map looks like this: Google tells you about Google, Microsoft tells you about Copilot and Bing, and for the assistants where a large share of high-intent research now happens you are on your own. That gap is why prompt-panel measurement still exists. At Crawlmind we run repeated prompt sets against ChatGPT, Perplexity, and Claude and compute citation share from the answers themselves, because there is no first-party feed to read instead. Sampling an assistant is worse data than a platform report by construction, and it is the only data available for those surfaces.

#How to structure the report

Keep the three streams separate and label them by what they measure.

Track Google's generative AI impressions as a ratio against Web search impressions, monitored for trend rather than reported as a total. Track Bing Citation Share per query group rather than per page, over multi-week windows, and note when a move is your citations changing versus the candidate set changing. Track assistant coverage separately as sampled data, with the number of runs stated next to the number.

The mistake worth avoiding is a single "AI visibility score" that averages these together. Three different denominators averaged into one figure produces a number that cannot go up or down for a reason you can explain. Given that neither platform gives you query-level data, the explanation is the entire value of the report.