Are AI Overview impressions the same as AI citations?

An impression, a citation and a mention are different events, and each one is counted by a different party.

• Search Console impressions record that Google showed your page in an AI Overview or AI Mode. Google does the counting.
• Bing citations record that Copilot or a Bing AI summary referenced your page. Microsoft does the counting.
• Tracker mentions record that your brand or URL appeared in an answer to a prompt the vendor sent. The vendor does the counting, on its own sample.

The denominators don't match either. Google counts across all real searches on its surfaces. Microsoft counts across real Copilot and Bing activity. A tracker counts across a few dozen or a few thousand prompts. So a rising tracker score and a flat Search Console line can both be true at once. The fix is not to reconcile them. It's to know what question each one answers.

As of October 2026, Search Console's generative AI report shows impressions for AI Overviews and AI Mode, broken down by page, country, date and device. It has no clicks, no CTR and no position column, which is why you can't see AI Overview clicks in Search Console. It also has no query dimension.

What it is good for: knowing which of your pages Google surfaces in its AI features, and whether that count rises or falls. It is the only first-party view of Google's AI surfaces.

Two cautions. Google says AI Overviews and AI Mode "may use different models and techniques", so treat them as two surfaces even when one report shows both. And a Google-side bug depressed impressions from around 12 August 2026, so check whether a dip in that window is yours before acting on it. Our guide to using Search Console's generative AI report covers the setup.

Bing Webmaster Tools' AI Performance report counts citations: total citations, average cited pages, page-level citation activity and citation share. It covers Microsoft Copilot, AI summaries in Bing and select partner integrations. It has no API as of October 2026.

Its unique value is grounding queries, the phrases the AI actually ran when it retrieved content it cited. Microsoft says that grounding query data is a sample. If you want to know which queries Copilot used when it cited you, this is the only place to look.

Read citation share carefully. Microsoft calls it an observational metric, not a ranking or a competitive scoreboard. It can fall while your absolute citations rise, because whoever already wins the query dominates the denominator.

A prompt tracker sends prompts it chose to engines it supports and records whether your brand or URL appears. That gives you something neither Google nor Microsoft offers: the actual answer text, across engines, including whether the facts about you are right.

It is still a sample, and three things limit it:

• The input is a guess. OpenAI says ChatGPT search rewrites a prompt into one or more targeted queries before searching. The tracker never sees those queries.
• Retrieval is discretionary. Engines decide per query whether to search the web, so a tracker averages grounded and ungrounded answers.
• Citation units differ. Anthropic's API cites text spans, while Perplexity cites whole sources by index. OpenAI returns more retrieved sources than cited ones. A "citation" is not one comparable thing.

A mention without a link is also not a citation. Decide which one your tracker counts before you compare it to anything.

OtterlyAI's 2026 experiment with two AI-generated blogs shows the gap in practice. Otterly's prompt tracker logged 3,934 citations for one site across seven engines. Microsoft's own AI Performance report credited the same site with 42,817. The time windows may not match, so this isn't an exact ratio. It does show that an engine's own count sees far more activity than any prompt sample.

The same test shows why engines must be read separately. Google removed both sites from its index, and Google's AI surfaces stopped citing them. Bing kept them, and Copilot produced 2,718 of the 3,934 tracked citations. A single blended score would have hidden that split entirely. The Bing and Google index split explains which answers draw on which index, and why Bing's counts differ from tracker counts goes deeper on the mismatch.

1. Keep each metric in its own row. Report Search Console impressions, Bing citations and tracker results side by side, never summed or averaged.
2. Compare each to its own past. Week-on-week change within one source is meaningful. Ratios across sources are not.
3. Check access first when a line drops to zero. A blocked crawler or a deindexed page removes you from that engine regardless of content. Look at Google and Bing index status separately.
4. Use trackers for answer quality. They are the only one of the three that shows what an engine says about you and whether it's wrong.
5. Don't buy a rank. Google exposes impressions only, so no position inside an AI answer can be checked.

If you have no tracker yet, the free checker on this site runs up to three buyer questions and shows whether engines name you. Treat it as a snapshot of answers, not a substitute for the first-party reports.

Can I turn Search Console AI impressions into traffic estimates? Not reliably. As of October 2026 the report has no clicks or CTR, and Google AI Overviews send no referral tag you can filter on in analytics. Any traffic number derived from AI impressions applies a click rate nobody can verify for your pages. Track the impression trend itself, and watch overall organic landing-page traffic for the pages that gain or lose AI impressions.

Is Bing's citation share the same as market share? No. Microsoft describes citation share as observational, not a ranking or competitive scoreboard. It does not expose competitor domains, represent traffic share or score content quality. Because the denominator includes everyone cited for those queries, your share can drop even while your own citation count grows. Read total citations alongside it before drawing conclusions.

Why does my tracker show far fewer citations than Bing? The tracker counts answers to its own prompt list, while Microsoft counts real Copilot and Bing activity. In Otterly's test the gap was 3,934 tracked versus 42,817 reported by Microsoft, with possibly different time windows. Engines also rewrite prompts into their own queries, so a tracker's prompts only approximate what users and models actually search.

Should I track brand mentions or only linked citations? Track both, but label them separately. A linked citation means the engine attributed your page as a source. An unlinked mention means the model named you, possibly from training data with no live retrieval. Mentions are where wrong facts about your pricing or products tend to appear, so they matter for accuracy even though they send no traffic.