Google Search Console added a dedicated filter for AI Overviews and AI Mode inside the Performance report. Open Search results, click the Search type filter, and two new options now sit next to Web, Image, and Video: AI Overviews and AI Mode. Select either one and the same clicks, impressions, CTR, and position columns you already check every week populate with data scoped to just that surface. For the first time, you can see which of your pages Google is actually pulling into its AI generated answers, by query and by page, without guessing from a traffic drop.
Before this filter existed, AI Overview impressions and clicks got folded into your Web totals with no way to pull them back out. A page could hold its ranking position in classic results and still lose clicks because an AI Overview above it answered the question first. Search Console had no column for that. You saw a dip in your overall numbers and had to guess at the cause. The new filter replaces that guess with an actual report.
This still is not a full answer to the question most marketing teams are really asking, which is whether AI is citing their brand at all. Search Console only sees Google’s own AI surfaces. It has nothing to say about ChatGPT, Perplexity, or Copilot. What it gives you is a real, first party view into the part of that picture Google controls directly, and that part is worth understanding in detail before deciding what it does and does not tell you.
The people most likely to open this report first are the ones who already live inside Search Console every week, technical SEOs, content leads, and anyone who reports on organic traffic to a leadership team. That is exactly why the numbers need a different reading than the Web data sitting one tab over. A metric built for ten blue links does not automatically make sense applied to a generated paragraph, and the fastest way to misuse this report is to read it with Web assumptions still attached.
Why this took Google so long to ship
The Performance report has organized data by search type for years. Web, Image, Video, and News each carry their own impression and click totals, switched between with the same filter control at the top of the report. AI Overviews sat outside that structure at first, because the surface blends organic ranking signals with a generated summary, and Google needed a way to attribute an impression to a page that might never get clicked even when it did the work of answering the question.
AI Mode complicated the picture further. It is a fuller, conversational search experience rather than a summary box bolted onto classic results, and a single AI Mode session can touch dozens of source pages across several follow up turns instead of just the one query a person typed. Reporting on that behavior meant building a structure that could handle multi turn sessions, not single queries, which is a different engineering problem than adding one more row to an existing table.
The result is a reporting model that finally treats both AI surfaces as first class parts of Search rather than numbers quietly absorbed into an existing total. It took time to ship because the underlying attribution problem was genuinely harder than adding a new filter to an old report.
Search Console has always lagged the live search experience by design. Google changes the results page constantly, and each change needs its own measurement plan before it can show up as a clean row inside a report property owners already trust. AI Overviews and AI Mode needed that discipline applied to a stranger problem than usual: a page can shape an answer without ever appearing as a link a person could see, click, or ignore. The old click and impression model assumed every impression came attached to a visible result. AI surfaces broke that assumption first and got measured properly second.
What actually changed in the Performance report
The mechanics you already know how to use, filtering by query, by page, by country, by device, all apply to the new data the same way they apply to Web. AI Overviews and AI Mode simply became two more values inside the Search type dropdown that already held Web, Image, Video, and News.
The two surfaces get separate filters because they are different products with different retrieval behavior. An AI Overview is a summary box inserted above or within the classic results for a subset of queries, generated from a handful of source pages and shown alongside the ten blue links a searcher would have seen anyway. AI Mode is a distinct search experience, a full conversational interface where one session can fan out into several related sub-queries and pull from a wider set of sources across multiple turns. Folding both into a single filter would have hidden the fact that a page can perform well in one surface and be invisible in the other.
Filtering to AI Overviews shows you queries where Google generated a summary and your page contributed to it. Filtering to AI Mode shows you the same thing for the conversational surface, including queries that never would have triggered a traditional AI Overview at all, since AI Mode activates for a broader, more exploratory set of searches. Compare a page across both filters and you will frequently find it performs differently in each, the same way a page can rank differently for a query on mobile versus desktop.
AI Overviews and AI Mode are two different surfaces with two different filters. A page that shows strong impressions under one can show almost nothing under the other. Check both before deciding your AI visibility is fine or broken.
How to find the AI Overviews and AI Mode report
The path is short. Most of the work is knowing what to look at once you get there.
- Open Search Console and select your property. Use the domain property if you have it verified, since it rolls up data across subdomains and protocols instead of splitting it apart by each one.
- Go to Performance, then Search results. This is the same report you already use for keyword and page level data.
- Click the Search type filter near the top of the report. Web is selected by default. AI Overviews and AI Mode now appear in the same dropdown.
- Select AI Overviews or AI Mode. Run them one at a time. The underlying queries for each surface frequently do not overlap, so comparing both inside one chart gets confusing fast.
- Switch to the Queries and Pages tabs. Sort by impressions first, since clicks will be low across most rows by design for both AI surfaces.
- Set the date range to at least 28 days. AI surfaces trigger on a smaller share of total queries than classic Web results, so a short window can make a real pattern look like noise.
Once both filters are open in separate tabs, the comparison that matters most is not AI Overviews against AI Mode. It is either one against your Web numbers for the same queries. That comparison tells you whether an AI surface is pulling attention away from your normal ranking or adding a second, separate way to appear.
Segmenting the data by query and page
The Queries tab is where the report earns its keep. Sort by impressions and look for two things: queries where you already rank well under Web but show zero AI impressions, and queries with climbing AI impressions that you have never specifically targeted. The first group points to a technical or structural gap keeping you out of a summary you should be part of. The second group shows you what your buyers are actually asking, sometimes phrased differently than the keywords sitting in your content plan.
The Pages tab tells a different story. A handful of pages usually account for most of your AI impressions, often the same pages that already carry your strongest schema markup, clearest answer first structure, and most direct heading hierarchy. If your AI impressions concentrate on two or three pages out of a much larger site, that concentration is itself useful information. It shows you which content patterns Google’s AI systems already trust, and it gives you a template to test against the rest of your site instead of guessing at what to change first.
Filter by country and device inside either AI search type the same way you would for Web. AI Overviews and AI Mode both roll out unevenly by market and by surface, so a page showing strong impressions in one country can show nothing in another even when the underlying content is identical.
What the numbers actually mean
Impressions
An impression under the AI Overviews or AI Mode filter means Google used your page as one of the sources behind a generated answer, whether or not the searcher ever saw your URL or clicked it. This is the biggest difference from how impressions worked under Web only reporting. A Web impression meant your listing appeared somewhere on the results page. An AI Overview impression can mean your content got absorbed into a paragraph of synthesized text with no visible link at all. The impression count tells you that you were used as a source. It does not tell you that anyone saw your brand name.
This matters most when a page shows a large impression count that never turns into meaningful clicks over several months. That pattern usually means the page is doing real work behind the scenes, feeding facts or figures into a summary, without earning any of the credit a visible citation would carry. A rising impression count on its own is not proof of growing visibility. It is proof of growing use, and use without credit is a problem worth naming precisely rather than celebrating as a win.
Clicks
Clicks behave the way you would expect, counted whenever someone follows a link out from the AI generated response back to your page. Expect this number to be small relative to impressions. That is not a sign your content failed. It is how AI Overviews and AI Mode are built to work: the summary answers the question directly, so a large share of readers get what they need without clicking anything.
Watch the ratio between clicks and impressions over time rather than the raw click count in any single month. A page that holds a steady ratio while its impression count grows is scaling in a healthy way. A page whose ratio quietly erodes while impressions climb is being cited less and summarized more, even if the total click number looks unchanged on the surface.
CTR and average position
Click through rate under an AI search type filter will read lower than the same query’s CTR under Web, often by a wide margin. Suppose a page shows 800 AI Overview impressions and 12 clicks in a month. A 1.5 percent CTR looks alarming next to a typical Web CTR of 25 to 30 percent for a page ranking near the top, but the two numbers are not measuring the same behavior. Average position inside an AI surface also means something different: it reflects where your source appeared within the generated citations or reference list, not where a blue link sat on the page. Read both metrics as AI specific numbers, not as a worse version of your Web metrics.
A shift in average position under either AI filter is worth watching even when the impression count barely moves. Google reordering which sources it credits first inside a summary, even without changing which pages it draws from, can shift which brand a searcher notices before they stop reading.
What counts as your page being used, and what does not
Not every page that ranks for a query gets pulled into the AI Overview or AI Mode answer for that same query, and the report only reflects the pages that actually did. A page can hold position two in classic Web results and never register a single AI impression, while a page ranking position six on the same query becomes the primary source behind the generated summary. Rank and source selection are related but separate decisions, made by different parts of the system.
Source selection tends to favor pages that state a claim plainly, attach a number or a specific fact to that claim, and avoid burying the answer under a long introduction. A page that spends three paragraphs building context before stating its point is harder for the system to lift cleanly than a page that opens with the fact itself. This is the same answer first structure that already matters for classic AEO work, and the AI Overviews and AI Mode filters give you a direct, page level way to confirm whether your content actually behaves that way in practice instead of just in theory.
| Search type | What triggers it | Typical CTR pattern | Best signal to track |
|---|---|---|---|
| Web | Standard organic ranking for a query | Highest CTR near position one, declining by position | Position and click trend over time |
| AI Overviews | Google generates a summary box for a subset of queries, pulled from a handful of sources | Low, since the summary often answers the query directly | Impression volume against your Web impressions for the same queries |
| AI Mode | Full conversational search session, often multi turn, pulling from a wider source set | Lowest of the three, spread across more source pages per session | Which queries and pages appear that never show up under Web or AI Overviews |
A worked example
Suppose a mid-size services company checks its AI Overviews filter for the first time and finds three pages carrying most of the data. Its pricing page shows 1,400 impressions and 6 clicks for the month. Its comparison guide shows 300 impressions and 40 clicks. Its homepage shows 90 impressions and 1 click.
The pricing page looks like a citation problem, not a content problem. High impressions with almost no clicks usually means Google is lifting a specific number or claim directly into the summary, satisfying the question before the searcher ever needs to visit. The fix is not more content on the page. It is making the page harder to fully summarize without a visible source link, through a named claim structure, a proprietary figure, or a comparison only that page can make.
The comparison guide shows the opposite pattern: lower impressions, but a far stronger click rate, which suggests the AI Overview is pointing searchers toward the page rather than replacing it entirely. That is the behavior worth protecting and replicating on other pages, since it means the page is earning attribution instead of just supplying raw material.
The homepage barely registers, and that is normal. Homepages rarely answer a specific enough question to get pulled into an AI Overview or AI Mode session, and a low number there is not a signal of a problem the way it would be on a page built to answer one clear query.
Four patterns worth acting on
Most accounts fall into one of four patterns once the AI Overviews and AI Mode filters have a few weeks of data behind them.
- High AI impressions, flat or falling Web clicks. Your content is getting used as a source and satisfying the searcher before they reach your site. This is not a technical problem to fix. It is a visibility problem, and the fix is the same content and entity work covered in a full AI visibility audit, aimed at earning a named citation or a link inside the summary instead of an uncredited assist.
- Zero AI impressions on pages that rank well under Web. Check the basics first: whether AI crawlers can reach the page, whether the content leads with a direct answer, and whether comparable competitor pages use structured formatting yours does not. A page ranking on classic Web signals but never appearing in AI Overviews or AI Mode usually has a structural gap, not a ranking gap.
- Strong AI Mode impressions with almost no AI Overviews impressions, or the reverse. This tells you the two surfaces are drawing on different query patterns for your category. Treat them as two separate targets rather than assuming success in one carries over to the other.
- AI impressions concentrated on a handful of pages sitewide. This is not a red flag on its own. It shows you which existing pages already meet whatever bar Google’s AI systems apply, and it gives you a working template, structure, schema, and answer placement, to test against the pages that show nothing.
None of these patterns require a different measurement tool to diagnose. They require reading the AI Overviews and AI Mode filters against your Web numbers for the same queries, not in isolation, since an AI impression only means something in relation to what was already happening on that query before the AI surface existed.
A page with rising AI impressions and falling clicks is not failing. It is being used and not credited. That is a citation problem, not a traffic problem, and it calls for a different fix than a ranking drop does.
What this report still cannot show you
Search Console only measures what happens inside Google. It has no visibility into ChatGPT, Perplexity, Copilot, or any other engine your buyers might ask instead. A brand can score well across both the AI Overviews and AI Mode filters and still be completely absent from a ChatGPT answer to the exact same question, because the two systems retrieve from different source pools and rank them differently. Our breakdown of how ChatGPT, Perplexity, and Google AI Overviews choose sources covers why that gap exists and why closing it on one engine does nothing for the others.
Search Console also cannot tell you whether you were named, linked, or simply absorbed into a paragraph with no attribution at all, and it cannot show you what a competitor is getting cited for instead of you. Those are the questions a dedicated cross platform view answers. Our guide to AI citation tracking tools covers the manual and automated methods for watching your brand across every engine your buyers actually use, Google, ChatGPT, Perplexity, and Copilot alike.
Treat the AI Overviews and AI Mode filters as the Google specific slice of a bigger picture, not the whole picture. It is real data, first party, free, and more useful than the guessing that came before it. It is also incomplete by design, because Google built it to report on Google, not on the AI search category as a whole.
Common mistakes when reading this report
A handful of habits quietly undermine this data before anyone gets to use it.
- Judging CTR against Web benchmarks. A CTR that would signal a problem under Web, a top ranking page barely getting clicked, often means nothing under AI Overviews. Judge each search type against its own baseline, not against the others.
- Checking a date range shorter than 28 days. AI surfaces trigger on a smaller share of queries than Web, so a seven day window frequently shows a handful of rows that read as noise rather than a pattern worth acting on.
- Treating AI Overviews and AI Mode as one number. Combining the two filters, or reading the AI Overviews number and assuming AI Mode looks the same, hides the fact that they pull from different query sets and different parts of a site.
- Assuming a zero row means a penalty. Most queries never trigger an AI Overview or AI Mode session at all. A page showing nothing under either filter is not being punished. It may simply be answering a question Google has no reason to summarize.
- Stopping the analysis at Google. A strong AI Overviews report can create false confidence that AI visibility overall is solved, when ChatGPT, Perplexity, and Copilot are three separate conversations this report never touches.
Building this into a monitoring routine
Check both filters monthly at minimum, alongside your normal Web performance review. During a content push or a schema rollout, weekly checks catch whether the change moved AI impressions before a full month has passed. Compare the same query set every cycle so the trend means something, the same discipline that makes any AI visibility tracking useful rather than a one time screenshot.
Layer this data into whatever you already use to track Share of AI Voice or your AEO Maturity Model score. Search Console’s numbers make a clean input for the Technical Foundation and Content Optimization pillars specifically, since a page with strong Web rankings and zero AI impressions is telling you exactly where those two pillars diverge.
Pair the Google side of the picture with a cross platform view. AEO Hunt builds AI visibility reporting that pulls Search Console’s AI Overviews and AI Mode data alongside ChatGPT, Perplexity, and Copilot citation tracking into a single monthly report, as part of our analytics and reporting service. You get the Google specific numbers plus the parts of the picture no Google dashboard will ever show you.
The report Google shipped is a genuine improvement over guessing from a traffic graph. Use it for what it is built to do, then build the rest of your AI visibility measurement around the gaps it leaves open.
