Some activity is measured, but the reports answer different questions
Google includes AI Overviews and AI Mode in Search Console’s Web performance data. That total does not tell you how many clicks came specifically from AI answers. Bing has a separate AI Performance report for citations across supported AI experiences. A citation count is not a traffic count.
This guide reflects documentation checked on 5 October 2026. Google describes the combined reporting in AI features and your website; Bing describes its dedicated report in the AI Performance announcement.
| Question | Useful evidence |
|---|---|
| Did Google Search send more clicks? | Search Console Web performance, including eligible AI search activity. |
| Was a page cited in supported Microsoft AI answers? | Bing’s AI Performance report. |
| Did an identifiable AI referral visit my site? | Your website analytics or logs, when referral information is available. |
| Did those visitors sign up or buy? | Your own conversion measurement, with its attribution limits. |
What Google’s numbers do—and do not—tell you
Clicks from external links in AI Overviews and AI Mode count as clicks. For AI Overviews, links within the overview share its search-result position. In AI Mode, a follow-up question is treated as a new query. Google documents these rules in its definitions of clicks, impressions, and position.
Suppose a page gets 1,000 Web clicks this month. That number does not establish whether 20, 200, or none came from AI Overviews. The overall CTR also cannot provide that split. Do not label the whole total “AI traffic” or try to infer an AI click count from a change in average position.
A query containing a question word is still just a question-shaped query. A regex for “how,” “why,” or “what” cannot identify which result format produced its clicks.
When impressions rise and clicks fall, investigate the affected queries, pages, devices, and search-result context. AI features may be part of that context, but the aggregate pattern alone does not prove that they caused the change. Our clicks-down, impressions-up example shows another way the same pattern can arise.
Bing separates citation activity
Bing’s AI Performance report covers Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations. Its original metrics include total citations, average cited pages, sampled grounding queries, and page-level citation counts. A grounding query is a phrase the system used to retrieve supporting content; it should not be assumed to be a verbatim user prompt. See Bing’s metric definitions.
In June 2026, Bing announced preview capabilities for Intents, Topics, Citation Share, and period comparisons. Citation Share describes your portion of citations for a grounding query across cited sites. It is not your share of visits, a ranking position, or a list of competitors. Topic and intent labels are classifications, and their precision can vary during the preview. Bing’s June update explains these additions.
Use the report to find which of your pages are being referenced and whether that changes over time. It is not an inventory of every AI answer everywhere, and a missing page in this report is not proof that no AI system has ever cited it.
More citations can coexist with fewer visits
Consider this fictional monthly report:
| Metric | Earlier month | Recent month |
|---|---|---|
| Bing-reported AI citations | 800 | 1,200 |
| Identified AI-referral visits in site analytics | 60 | 55 |
| Signups attributed to those visits | 3 | 5 |
You can report that citations rose 50%, identified visits fell from 60 to 55, and attributed signups rose from three to five. You cannot calculate an AI CTR by dividing those 55 visits by 1,200 citations: the systems do not necessarily cover the same surfaces, people, or events.
The small signup count is worth watching, not enough to establish a reliable trend on its own. Keep the three measures separate so a larger citation number does not conceal a decline in visits—or a useful improvement in the visitors who arrive.
A measurement checklist you can repeat
- Choose a small set of pages. Start with important guides or product pages. Record their exact URLs so later comparisons concern the same content.
- Record Google Web performance. Keep the dates, device, country, and other filters with the clicks, impressions, CTR, and position. Label it “Google Web,” not “Google AI.”
- Record Bing citation data separately. Note the report’s date range, cited URLs, and the metrics available in your account. Keep preview reporting changes in your notes.
- Check identifiable referrals and outcomes. If you have site analytics or logs, look for visits with relevant referral information and the actions those visitors took. Missing referrers and consent or tracking gaps make this an incomplete view.
- Annotate meaningful changes. A rewritten guide, a new product, or a site migration gives a later comparison context. A before-and-after change is evidence to investigate, not proof of causation.
A simple tracking sheet can have one row per URL and period, with separate columns for Google clicks, Bing citations, identifiable referral visits, and conversions. Leave unavailable data blank. Do not replace it with an estimated “AI score.”
Where Numic fits
Numic brings Google and Bing search performance reports to iPhone, iPad, and Mac. Its current Bing integration does not include the AI Performance dashboard described here; use Bing Webmaster Tools for those citation reports.
For a closer look at a Google traffic change, the CSV comparison tool can show which exported rows gained or lost clicks. It does not identify AI-only clicks. Keep that distinction when sharing its results.