You pull up Search Console for a page you care about and the numbers look strange. Impressions are up. Clicks are flat, or they slid a little. Your first thought is probably AI Overviews, because everyone is talking about them right now. That thought might be right, but Search Console alone will not confirm it just because the shape looks familiar. This piece walks you through what the impressions up clicks down Search Console pattern actually tells you, and which other causes you should rule out first, using only the data already sitting in your own account.
By the end you will be able to look at one page's clicks, impressions, CTR, and average position, plus the newer Generative AI performance report, and read the AI Overview impact on Search Console data with some confidence, rather than guessing from the shape of one chart. What you will not get is proof. Google does not publish a report that assigns a lost click to an AI Overview, and this piece will not pretend otherwise.
What the numbers actually mean before you diagnose anything
It helps to be precise about what each metric is counting, because the diagnosis depends on reading them correctly, and because clicks and impressions decoupling is the thing you are actually trying to read. A click is a click from Google to your site. Clicking the same result twice after coming back to the results page usually counts as one click, and a different result counts separately. An impression means a link to your site appeared, or could have appeared, in front of a user. It is an opportunity to be clicked, not proof that anyone noticed it. CTR is just clicks divided by impressions, and it is the clearest way to describe clicks and impressions decoupling: when impressions grow faster than clicks, or clicks fall while impressions hold, CTR drops.
Average position is the average rank of your site's topmost result for a query, averaged across the property in the chart and by the row's grouping in a table. A move from position 2 to 4 is a small change by Google's own example. A move from 4 to 29 is a large one. Position is not fixed. It shifts by device, location, time, and search history, so "steady" here does not mean identical, it means no meaningful deterioration across comparable periods, with the query-level rows agreeing with that story.
Google now also documents a Generative AI performance report inside Search Console, covering AI Overviews and AI Mode together. It counts generative-AI impressions: the number of times a link to your site showed up inside a generative-AI feature. This report was rolled out worldwide by the end of August 2026, though a low-traffic property may still show thin data. One thing to hold onto for the rest of this piece: the report is documented as an impressions report. It does not give you clicks, CTR, or position for the AI feature itself. You get that from the standard Performance report, and you have to line the two reports up yourself.
Start with a query-mix or demand change
This is usually the first thing worth ruling out, because an aggregate chart can hide what actually moved underneath it.
What it looks like. Total impressions and clicks shift, but average position for the page does not clearly worsen. When you break the page down by query, you find a different mix than before, or the same queries at different volumes. Device or country rows sometimes explain the whole thing on their own.
How to confirm it. Filter the standard Performance report to the one page, then look at it by query, device, country, and date. Ask whether the apparent gain in impressions is coming from new queries with low click intent, rather than from the page's established terms. Compare equal, completed periods, and do not read a property-level chart as if it described one page.
What to do. Do not touch the page just because the aggregate numbers moved. Separate the mix change from the page's core demand first. If the established queries are holding position and CTR while generative-AI impressions climb elsewhere, AI displacement stays a live possibility, worth checking next. If the query set itself changed, call it a mix change and move on, because there is nothing to fix on the page.
The AI Overview signature itself
This is the pattern the title of this piece is named for, and it is a relationship among several metrics, not a single number you can read off a chart.
What it looks like. For the same page, over comparable periods: standard impressions are stable or rising, clicks are flat or falling, CTR is falling, and average position is broadly stable. On top of that, the Generative AI performance report shows impressions for that page in the same window, ideally rising or shifting close to when the click pattern changed. The decline tends to sit in queries the page already ranked for, rather than being explained away by a new query mix.
How to confirm it. To diagnose AI Overview traffic loss with any confidence, filter the Generative AI report to the affected URL and line its dates up with the standard report for the same page. The strongest signal is a generative-AI impression increase that overlaps in time with a falling CTR, without a matching drop in position. Remember the report's chart defaults to property-level totals; add a URL filter so you are reading the one page you care about, not the whole site blended together.
What to do. Do not assume you lost rankings. The real question here is attribution, not repair. If the pattern holds, keep the distinction between visibility and click yield in mind: the page is still being surfaced, it is just converting fewer of those surfacings into visits. Any change you make to the page should come from thinking about its usefulness and intent match, not from an assumption that a ranking fix will bring the clicks back, because there may be no ranking problem to fix.
What this cannot tell you. Search Console cannot, from these metrics alone, assign a specific lost click to an AI Overview, tell an AI Overview apart from some other search feature sitting above your result, or show you the clicks you would have gotten without it. It also cannot tell you why a page was left out of an AI Overview it might belong in. That is a different diagnosis, outside what this piece covers.
A genuine ranking or visibility drop
What it looks like. Average position gets meaningfully worse, usually alongside a decline in impressions, clicks, or both. It might touch one page, a cluster of related pages, or many pages across the site. A sharp position change tells you more than a small wobble.
How to confirm it. Compare the page's average position across equal periods, then check queries and dates underneath it. Is the loss concentrated in the page's usual queries? Look at other pages in the same topic area to see whether the drop is isolated to this one page or spread wider. An impression decline running alongside the position decline strengthens the read.
What to do. Look at ranking, relevance, indexing, and page-quality explanations before assuming clicks were intercepted by an AI feature. Do not rewrite the page just because traffic fell. Google's own guidance on traffic drops is clear that a decline does not automatically mean something is fundamentally wrong with the content, and that positions move because the web and the results around them keep changing.
A pattern that spans more than one page
What it looks like in Search Console. It resembles a plain ranking decline, often showing up across several pages or query groups around the same time. Position may worsen, and impressions and clicks may fall with it. Nothing in the four standard metrics uniquely identifies this as an algorithm update rather than something else.
How to confirm it. Compare the affected page against other pages and query groups in the same property. A broad, simultaneous move across many pages points toward a site-wide or system-wide event. A change confined to one page or one query cluster keeps a page-specific explanation on the table.
What to do. Describe what you see as a ranking loss that looks consistent with a broader change, rather than naming it a confirmed update, since Search Console alone cannot tell you which system moved. Google says its ranking systems are updated continually, and that a traffic decline after one of those changes does not automatically mean your content is at fault.
A technical, indexing, or canonical issue
What it looks like. Impressions and clicks fall sharply, sometimes close to disappearing, often with a large position change. The page's numbers may look inconsistent with the rest of the site's pattern, or data you expect to see under one URL is missing.
How to confirm it. Check the affected URL against the canonical version shown in the Pages report. Search Console generally assigns data for page variations to the canonical URL Google selected, so make sure you are not looking for data under a URL that Google folded into a different one.
What to do. Settle the data-ownership question before you diagnose anything else. If the numbers moved to a canonical URL, do your analysis there. If the impressions genuinely vanished, you are not looking at the impressions up clicks down Search Console pattern at all, you are looking at a technical problem.
A reporting or aggregation artifact
What it looks like. The chart and the table seem to disagree with each other, or a pattern visible at the property level disappears once you look at a single page or query. The most recent days in the range can also shift as Search Console finishes processing them.
How to confirm it. Check whether you are comparing a property-aggregated chart against a page-aggregated table. Google has said directly that chart and table totals can differ for this reason. Use the Pages dimension when you want a page-level answer, and apply a URL filter inside the Generative AI report too. Treat the newest few days of data as provisional.
What to do. Re-run the comparison with the same page, search type, date range, device, and country scope, and use completed periods only. Do not build a story about AI Overviews, or anything else, out of a mismatch that aggregation caused.
If none of these match
Sometimes the pattern in front of you does not cleanly fit any of the sections above, or two of them seem to apply at once. That is a reasonable place to land. Write down what you actually saw: the two comparison periods, the page-level metrics, the query breakdown, and the Generative AI report view for that page, before you change anything on the page itself. A later comparison means much more when you have the original pattern on record rather than a memory of "traffic went down around then."
If you are still unsure after working through the sections above, treat the diagnosis as open rather than forcing a conclusion the data does not support. Continuing to log the same metrics over the next few weeks, on a fixed cadence, will usually resolve the ambiguity faster than a single snapshot ever could.
Reading the frame together
None of this works as a single threshold you check once. It is a set of questions you ask in sequence, and where you stop tells you the likely explanation.
Did clicks fall or hold flat while impressions rose? If yes, check CTR next. Did CTR fall? If yes, you have confirmed decoupling, though not its cause yet. Did average position materially worsen? If yes, a ranking or indexing explanation moves ahead of AI displacement. Did position stay broadly stable instead? Then an AI Overview or another search-result feature becomes a more plausible explanation. Did the page pick up generative-AI impressions over that same window? If yes, that strengthens the AI displacement read. Did the pattern only show up after a shift in query, device, or country mix? If so, the aggregate is compositional, and you should not pin it on AI without a like-for-like comparison first. Did impressions collapse along with clicks, rather than holding or rising? That points back toward visibility, ranking, indexing, or demand. Is the movement showing up across many pages, not just this one? Then consider a broader event rather than something specific to this page.
Treat this as a way to organize your thinking, not a scoring system with a pass or fail line. Nothing here assigns points.
A practical way to run the comparison
Pick two equal, completed periods to compare, sized to the amount of traffic the page gets. A high-traffic page can show a pattern over a shorter window; a low-traffic page needs more time before you can trust what you are seeing. Start at the page level by filtering the standard report to the affected URL, rather than reading the whole property and assuming it describes one page.
Record clicks, impressions, CTR, and average position for both periods, and check the Pages and Queries dimensions to see whether the page's established queries are behaving the same way as the aggregate. Look at device and country breakdowns too, since a shift toward mobile or a different country mix can move CTR without any ranking change behind it. Then check the Generative AI report for the same page and dates, filtered by URL, and record the impressions figure it shows, since that is the only metric it currently documents.
Compare the timing of any generative-AI impression change against the timing of the CTR decline, not just the totals on their own. A rise that begins right around the point clicks flattened tells you more than a lifetime total ever could. When you are done, classify what you found honestly: consistent with AI Overview displacement, a ranking loss, a query-mix change, or unresolved. Save yourself the trouble of a stronger word than the data supports.
Diagnosing AI Overview traffic loss without overclaiming
The full method above exists to diagnose AI Overview traffic loss carefully, and carefully means accepting its limits along the way. A stable average position does not prove AI Overviews took your clicks. It only rules out a clear position decline as the primary explanation, and it says nothing about which other search feature might be involved. Rising generative-AI impressions on top of a stable position is stronger evidence than a stable position alone, but it is still an attribution signal, not a causal test.
For context on why this distinction matters: to even qualify as a supporting link inside an AI Overview or AI Mode, a page has to already be indexed and eligible to appear in regular Google Search with a snippet. There is no separate technical requirement or special markup needed for these features. Both AI Overviews and AI Mode can pull from multiple related searches across a topic to build their answer, and they can use different underlying models, which is part of why the pages they cite can shift over time even when your own page has not changed.
If your numbers do fit the signature, the honest next step is not necessarily a content rewrite. It is deciding whether the page still does its job for the reader who lands on it, since a ranking fix will not help a click-yield problem that was never about ranking in the first place. Tools that already track how your pages get cited inside AI answers, alongside your ordinary Search Console data, can make this kind of comparison faster to run on an ongoing basis instead of a one-off pull, which matters more the more pages you are watching.
