The GSC Diagnostic: How to Confirm AI Overviews Caused Your Traffic Drop

Why Search Console is the right tool for this

Search Console is Google's own record of how your site performs in search, which makes it the closest thing to ground truth you have. As the free tool Google provides for monitoring search presence, it reports the exact numbers you need without the guesswork of third-party estimates.

The reason it can confirm an AI Overview squeeze is that it separates the metrics that matter. It shows how often you appeared, how often you were clicked, and where you ranked, as three distinct measures.

When those three move in a particular pattern, they point to a single cause, and no other tool gives you that cleanly.

So before touching anything else, this is where the diagnosis lives. The rest of this guide is a sequence of reads inside Search Console, each one narrowing the possibilities until you can say with confidence whether an overview took your clicks. That certainty is what turns a panic into a clear, ordered plan, and it feeds directly into your wider answer engine optimisation work.

The signature you are looking for

Start by learning the pattern, because the whole diagnosis is really pattern matching. The AI Overview squeeze has a distinctive fingerprint in Search Console, and once you know it, the reports almost read themselves.

The three core metrics tell the story. Google defines impressions, position and clicks as, roughly, how often you were shown, where you ranked on average, and how often people clicked through. In a classic squeeze, impressions hold steady or even rise, average position stays stable, and clicks fall. You are still being shown, still ranking, but the click is being taken before it reaches you.

Contrast that with the other causes and the value becomes obvious. A ranking loss drags impressions and position down together. A technical or indexing problem sends impressions towards zero.

Seasonality moves impressions and clicks together and then recovers. Only the squeeze produces the tell-tale split of stable impressions and position with falling clicks, so that split is what you are hunting for.

Step one: confirm the drop and time it

Begin by establishing that the drop is real and pinning down when it started. Open the Performance report on search results and set the date range wide enough to see the trend clearly, ideally comparing the last few months against the same span before.

Look at the clicks line first and find the point where it turned down. Note the date, because timing is a clue in itself.

A drop that began the week an overview rolled out for your topics points one way, while a drop that lines up with a known algorithm update points another. Google's guidance on why site traffic drops is a useful reference for the shapes these declines can take and what each tends to mean.

Timing also protects you from chasing ghosts. If the decline predates the arrival of overviews on your queries, they are not the cause, and you save yourself a wrong turn. Fixing the date first keeps the whole diagnosis honest.

Watch the shape of the decline as well as its start date. A squeeze usually shows as a step down that begins when an overview rolls out and then holds at the lower level, rather than a slow slide or a sudden cliff.

A gradual drift more often points to rising competition or decaying content, and a vertical cliff to a technical fault or a penalty. The curve itself is a clue, so read it before you read anything into it.

Step two: read clicks against impressions and position

Now put the three metrics side by side for the same period. This is the heart of the diagnosis, and it is where the squeeze reveals itself or rules itself out.

Enable clicks, impressions, average position and click-through rate on the Performance chart together. If impressions and position are roughly flat while clicks and click-through fall, you are looking at the squeeze.

Your content is being served as often as before, at the same rank, but earning fewer clicks, which is exactly what happens when an answer sits above your listing.

If instead impressions and position slid alongside clicks, pause the AI Overview theory and treat it as a ranking or visibility problem.

The numbers are telling you the loss happened higher up the funnel, before the click was ever on offer. Reading these lines together, rather than one at a time, is what separates a real diagnosis from a hunch, and it pairs naturally with your AEO measurement.

If the chart feels overwhelming, work through it methodically. Search Engine Journal's walkthrough on uncovering traffic declines in Search Console is a useful companion for reading the same report, breaking the decline down without jumping to conclusions.

The goal at this stage is simply to decide which of the metrics moved, because that single fact narrows the cause more than any other check.

Step three: filter to the affected pages and queries

The site-wide view hides the detail, so narrow down. Use the Performance report filters to isolate the specific pages and queries that lost clicks, because the cause almost always lives in a subset, not the whole site.

Filter by page to find which URLs bled clicks, then filter by query to see which searches drove the loss. Google's deep dive on filtering performance data explains how to slice the report and where its limits lie, so you read the numbers correctly rather than over-interpreting a sample. Sort by lost clicks and you get a ranked list of exactly where to look.

The pattern that emerges is usually telling. Informational, question-shaped queries tend to dominate the losses, because those are the searches overviews answer most.

If your biggest drops cluster on how-to and definition queries while your transactional pages hold, that distribution itself is strong circumstantial evidence, which our guide to optimising for AI Overviews then helps you act on.

Keep a note of the pages that did not lose clicks as well as those that did. The survivors are just as informative, because they show which of your content types the squeeze has not yet reached, and they often point to where your safest, most clickable opportunities still lie.

A diagnosis that maps both the losses and the holds gives you a fuller picture than one fixated only on the damage.

Step four: compare periods to isolate the change

Numbers in isolation can mislead, so compare. Use the date comparison feature to set the weeks after the drop against an equivalent span before it, for the same filtered pages and queries.

This does two things. It quantifies the loss precisely, turning a vague sense of decline into a percentage you can report, and it confirms the loss is concentrated where you think it is.

If a filtered set of query shows stable impressions and position but a sharp fall in clicks between the two periods, you have isolated the squeeze to specific searches. That is the level of proof worth acting on.

Keep the comparison fair. Match the length of the periods, avoid straddling a major seasonal event, and compare like for like, so the difference you see is the change you are diagnosing and not an artefact of the calendar.

Step five: check the live results

Data points to the cause, but a live look confirms it beyond doubt. Take the queries that lost the most clicks and run them in a real search, ideally the way a customer would phrase and locate them.

Look at what sits above your listing. If an AI Overview or another answer feature now occupies the space your result used to own, the circle is closed: stable impressions and position, falling clicks, and a visible answer intercepting the click.

That is a confirmed AI Overview squeeze, not a theory. Extending the check across ChatGPT search and Perplexity shows whether the same questions are being answered off Google too.

Do this manually rather than trusting a screenshot from a report, because overviews appear and vanish by query, location and phrasing. Seeing what your customers see is the final, decisive piece of the diagnosis.

It is worth checking from a location and device that match your audience, since an overview that shows for one searcher may not show for another.

If you serve a specific area, search as someone there would, because the results a customer actually meets are the only ones that matter for your traffic. A handful of these live checks, spread across your worst-hit queries, gives you a truer picture than any single screenshot ever could.

Step six: rule out the other causes

A responsible diagnosis eliminates the alternatives, so run the checklist. Each other cause leaves its own fingerprint, and confirming their absence strengthens your conclusion.

Check for a known core or algorithm update around the drop date, since that would point to a ranking cause rather than a squeeze. Confirm the pages are still indexed, because a technical problem, not an overview, would explain impressions collapsing.

Consider seasonality by comparing the same period last year, and verify your analytics and tags did not change, which can fake a drop that never happened. Semrush data on the prevalence of AI Overviews helps you judge how likely an overview is for your kind of query in the first place.

When the other causes are ruled out and the signature holds, you have your answer. A full AI visibility audit can formalise this across your whole site, and it complements a broader website audit for generative AI search when you want the complete picture.

What this looks like in practice

The value of the method shows up when it stops a wrong turn. A common story is a team that saw traffic fall, assumed a ranking problem, and spent weeks on links and on-page tweaks that changed nothing, because the position was never the issue.

The Search Console signature would have told them so in an afternoon.

Once they read impressions against clicks and checked the live results, the overviews sitting above their best queries were obvious, and the plan wrote itself. Our case study shows the same shift, where progress came from diagnosing correctly and then acting, rather than throwing effort at a symptom. The diagnosis is not busywork, it is what makes the fix land on the real problem.

What to do once you have confirmed it

Confirmation changes the work. You now know the position is fine and the click is the problem, so the fix is to be inside the answer and to win the searches that still click, not to chase a ranking you already hold.

Start from your home base and turn the confirmed queries into a priority list. For each, aim to be the cited source in the overview and to strengthen the pages that still earn clicks, recovering the ground the squeeze took.

That recovery is exactly what turning lost traffic into AI citations is built around, and extending it to Google's AI answers keeps you visible as the results page evolves.

Aligning the whole effort with an AI SEO plan turns a one-off diagnosis into an ongoing habit, so the next time your traffic moves you already know how to read it and where to act.