How to read Google's and Bing's AI reports, and what to do with them

This is a documented guide, not a test. It is built from Google's Search Console help pages and Microsoft's Bing Webmaster Tools documentation and announcements, checked in October 2026. It contains no data or screenshots from a live account, because the numbers belong to your own properties. For the wider list of what to fix first, see the AI search checklist.

The two reports, side by side

From Google's Search Console help article on the report and Microsoft's AI Performance help page and February and June 2026 announcements, October 2026. Both vendors say these reports are summaries, not complete logs.

Google: the Generative AI performance report

Where it is. In Search Console, open Performance and choose the Generative AI performance report (Search). It sits next to the usual Search results report.

What an impression means. Google defines it as how many times links to your site were shown to a user in a generative AI feature. For the chart, two results from the same site in one search count as one impression. The report includes AI Overviews and AI Mode, and Google says the list may expand. Search Labs experiments are excluded.

What you can filter. Pages (the canonical URL after redirects), countries, devices and dates in Pacific Time, with a search type split between Web text and Web multimodal, such as Google Lens or Circle to Search. The usual Performance limits apply, including 1,000 rows, and the newest data can be preliminary for a few hours.

What is missing. No clicks, no click-through rate and no average position. Google says the report's data is part of the Web search type in the main Performance report, which is where clicks for the same search type live. Google's AI features page suggests combining Search Console with Google Analytics to track engagement and conversions.

Bing: AI Performance

What it measures. Visible citation activity: how often your content was shown as a source in AI-generated answers. Microsoft's help page defines total citations, cited pages per day, average cited pages, grounding queries and page-level citations, and a timeline for 7 days, 30 days, 3 months or a custom range.

Grounding queries are not prompts. They are short phrases the AI used when retrieving content that was then cited, grouped across answers. A page can appear under several of them, and a query can map to several pages. The page-to-query mapping lets you click from one to the other. Microsoft says the data does not show individual answers, exact prompts or why a page was referenced, and that it is a representative summary, not a complete log.

June 2026 additions, still in preview. Intents classify grounding queries into broad types such as informational or commercial. Topics group related queries into themes. Citation Share shows your citations as a share of all citations for the same grounding query, which Microsoft calls an observational metric, not a ranking system or a scoreboard. Compare overlays a previous period.

Read it carefully. Microsoft states that trend changes cannot be attributed to a specific cause such as one content update or model change, and that the report does not measure rankings, authority or importance.

What to do with the data

1. List your most-cited and most-shown pages. Ask what they have in common: a first-hand angle, a clear table, a direct answer. Improve the pages next to them.
2. Find pages with impressions but no clicks. Compare the Generative AI report with the main Performance report for the same page. Pages that are shown but not clicked may need a stronger title or answer.
3. Read Bing's grounding queries as topic hints. If a phrase keeps appearing, check that a page answers it fully and accurately.
4. Check for pages that should be cited but are not. The first thing to rule out is indexing: Google says a page must be indexed and eligible to appear with a snippet.
5. Watch country and device splits for markets you did not expect, and for multimodal appearances.
6. Do not chase one number. Both vendors warn that these are aggregated views and that causes cannot be pinned down.

What neither report can tell you

• Why a page was chosen. Neither explains it.
• How you compare with competitors by name. Bing's Citation Share does not reveal competitor domains; Google's report is about your own site.
• What users asked. Bing shows grouped phrases, not prompts; Google does not show queries here.
• Revenue impact. Pair it with analytics and conversion data.
• Claims from third parties. Google's AI guide warns that no third-party tool has access to its internal ranking or AI systems.

The control you should know about, and leave alone

Search Console also has a setting for Search generative AI under Settings. It lets you exclude a property from AI Overviews, AI Mode and Discover's generative AI features, include it again, or inherit from a parent property. Google says exclusion stops your links and content appearing in these features and sends no impressions or traffic from them, takes effect usually within one to two days, and is not a ranking signal. It does not affect AI training, which is controlled with Google-Extended, and it does not remove pages from Search; noindex does that. Unless you have a clear reason, leave it on its default of included.

A monthly routine for 2027

Both reports are young and Microsoft's is explicitly in preview, so do not build a process that depends on today's layout. This routine is our suggestion, not a vendor recommendation.

Saving your own exports matters because the windows are limited: Bing shows history only for the time ranges in its dashboard, and Google's data in this report is subject to the standard Performance limits.

Where Rank Math fits

Rank Math connects to your Google Search Console account, and its MCP tools include a read-only top-keywords tool that returns the Search Console data, so an assistant can summarize it next to your SEO scores. It does not replace Google's Generative AI report, which lives in Search Console itself. The free plugin covers the Search Console connection and the MCP tools; the setup is in Rank Math MCP with Claude. Get Rank Math (affiliate link).

Turning the data into pages

The reports tell you which pages are visible; they do not write the next one. GrowGanic is an autonomous SEO engine: add a domain and it researches the keywords, writes and scores the articles, publishes them to your site or a hosted blog, and tracks the rankings. With a Google Analytics property connected, it reads how organic visitors engage with each article and uses that to order future topics, which is the same engagement data this routine recommends checking. It has no backlink index, and, by GrowGanic's own account, the hosted copy of the free article stays out of Google until the plan is paid (connecting your site or a subdomain first avoids that). Read each draft and add something only you know. Try the free tier (affiliate link).

Who can skip this for now

• New sites with few impressions. Google says the report may not appear without enough generative AI impressions. Fix indexing and publish first.
• Sites that only care about clicks. The Generative AI report has none; the main Performance report is the one to watch.
• Anyone expecting to see rankings in AI answers. Neither report offers them.

The verdict, by situation

Frequently asked questions

Sources

Google's help articles on the Generative AI performance report and the Search generative AI control, and its AI features page; Microsoft's AI Performance help page and its February and June 2026 announcements; plus Rank Math's MCP tools reference. Report layouts and availability change often, so check the help pages against your own account.

Related

Last reviewed 20 October 2026. Both vendors are still changing these reports, so this page is re-checked monthly. Spotted one out of date? Tell me.