Are AI Overviews Taking Your Traffic? A GSC Test

Your traffic is down 30% and someone on LinkedIn told you AI Overviews ate it. Maybe. Or you got hit by a core update, or you slipped from position 3 to position 7, or it is January and nobody buys a new furnace in January.

These four causes look identical on a traffic graph and completely different in Google Search Console. This post is the differential diagnosis. It takes about 20 minutes and needs no tools you do not already have.

One thing first: this post assumes you used to rank and the clicks went away. If you never ranked in the first place, you have a different problem and we wrote about it separately in why your service business isn't ranking on Google. Do not diagnose an AI Overview problem on a page that Google has never sent traffic to.

Are AI Overviews actually taking your clicks, or did you lose rankings?

Three numbers separate them: impressions, average position, and clicks. AI Overview click compression is the only cause where impressions stay flat, average position stays flat, and clicks fall anyway. Every other cause moves at least one of the first two numbers.

That is the whole test. Google still counts an impression when your link appears on a SERP with an AI Overview on it, and your organic position is still your organic position. What changed is how many people scroll past the summary and click. So the loss shows up in clicks and CTR, and nowhere else.

Pull it in Search Console: Performance > Search results, 16 months, compare the last 3 months to the same 3 months last year. Not last quarter — the same quarter last year, or seasonality will lie to you. Turn on all four metrics: clicks, impressions, CTR, average position.

What does the Search Console pattern for AI Overview click loss look like?

It looks like a widening gap between two flat lines: impressions holding steady (or rising) while clicks bend downward underneath them, with average position unchanged within roughly half a position. Here is the full lookup table.

The second row is the one that matters most, and it is the row most agencies quietly skip. If your average position got worse, you did not lose clicks to an AI Overview — you lost the ranking. Blaming AI is a convenient way to avoid saying the SEO stopped working. Anyone who shows you a traffic chart, says 'AI Overviews,' and never shows you the position line is managing your emotions, not your search visibility.

If you have never opened these reports before, start with our walkthrough of Google Search Console for service businesses and come back.

How do you separate an AI Overview hit from a core update or seasonality?

By shape and by date. A core update lands on a known date and hits the position line; an AI Overview hit has no date and never touches the position line. Seasonality repeats on the same calendar week every year.

Run these three checks in order, and stop at the first one that explains the drop:

• Date-match the cliff. Note the exact week clicks fell. Cross-reference Google's Search status dashboard for a confirmed core or spam update in that window. A drop that starts on a rollout date, hits the position line, and hits your whole site at once is a core update — not AI.
• Year-over-year the same weeks. Compare weeks 1–12 of this year to weeks 1–12 of last year, not to last quarter. A plumber's search demand in July is not the same animal as in January. If last year's line has the same dip, it is seasonality and there is nothing to fix.
• Segment branded vs non-branded. Filter queries containing your business name, then invert it. If branded clicks are flat and non-branded informational clicks fell, that is a SERP-feature story. If branded clicks fell too, something is wrong with your brand or your business — not with Google.

A drop that survives all three checks — no update date, no seasonal twin, branded traffic intact, position line flat — is click compression. Now you can say 'AI Overviews' and mean it.

Which of your queries even trigger an AI Overview?

Roughly one in five. Ahrefs analyzed 146 million SERPs and found AI Overviews appear on 20.5% of keywords — but that average hides everything useful, because the trigger rate depends almost entirely on what kind of query it is.

Read the bottom of that table again. AI Overviews trigger on 4.3% of commercial-intent and 2.1% of transactional-intent keywords, versus 21.4% of informational keywords (Ahrefs, 146M SERPs, September 2025). Your 'how does a heat pump work' post is in the blast radius. 'Emergency AC repair Tampa' mostly is not.

Since June 3, 2026, you can also check this directly. Google launched a Generative AI performance report in Search Console showing impressions, pages, countries, devices, and dates for AI Overviews and AI Mode. Two catches, both stated by Google: it is impressions only — no clicks, no position, no CTR, and Google is rolling it out to a subset of websites, so your property may not have it yet.

So the new report tells you which of your pages AI features are showing — it does not tell you what that cost you. You still need the impressions-flat / position-flat / clicks-down test above to price the damage.

Are near-me and emergency queries safe from AI Overviews?

Mostly, but the honest answer is that the two best datasets disagree, and you should know why before someone sells you a panic retainer. Ahrefs measured a 7.9% AI Overview trigger rate on local searches; Seer Interactive measured 76.9% on 'near me' queries.

Those are not the same measurement. Ahrefs' 7.9% covers all queries its system classifies as local intent, across a 146-million-SERP census. Seer's 76.9% covers informational queries containing the phrase 'near me' inside 53 commercial brand accounts. Different denominators, different populations. Anyone who averages them or quotes one without the other is guessing.

What both are consistent with: an AI Overview is far more likely on a research question than on a job-ready one. 'Best time of year to replace a furnace' is a question query — 57.9% trigger rate. '24 hour plumber near me' is somebody standing in water. Google's own trigger data puts transactional intent at 2.1%.

For a trades business, that is the whole story: your plumbing and HVAC money pages sit on the intent classes AI Overviews largely leave alone, and your blog sits on the ones they do not. Check yours before you assume it — segment your queries into emergency, near-me, service+city, cost, and research, and look at the CTR line on each of the five separately. The five will not move together.

Which pages should you stop investing in, and which should you re-point?

Sort every affected page into one of three buckets, and only one of them gets more money. The test is not 'did it lose clicks' — it is 'did the clicks it lost ever turn into work.'

• Re-point (fund this). The page ranks, held position, lost CTR, and sits near a buying decision — repair-vs-replace, cost, 'is X worth it', 'do I need a permit.' Add a real answer in the first 60 words, add the price range, add a booking path. You want the click that survives the summary, and those clicks are disproportionately the ones with a wallet attached.
• Maintain, do not grow. Definitional and top-of-funnel explainers that hold impressions and feed brand recall. Keep them accurate, keep them cited, stop expanding them. They are now a citation asset, not a traffic asset.
• Stop (cut this). Pages that lost clicks, never produced a call, and answer a question an AI Overview now answers completely in 67 words. There is no version of that page that wins. Retire the budget, not necessarily the URL.

The bucket-3 pages are where most agencies keep billing you. A monthly retainer that keeps 'refreshing' a post about what a P-trap is has stopped being SEO and started being an annuity. That is exactly the kind of line item our 90-day kill switch exists to end: a channel or an asset with no qualified leads in 90 days gets cut.

For the pages you do keep, the goal shifts from 'get the click' to 'be the citation.' Seer Interactive analyzed 5.47 million queries across 53 brands and found that being cited inside an AI Overview delivers 120% more organic clicks per impression than appearing on the same SERP uncited — though still 38% fewer clicks than a SERP with no AI Overview at all. Being in the box is not as good as no box. It is twice as good as being outside it.

How much click loss is normal and survivable for a service business?

On the informational queries that trigger an AI Overview, the measured hit is severe: Ahrefs' Search Console data across 300,000 keywords shows the presence of an AI Overview correlates with a 58% lower clickthrough rate for the #1 organic result. The penalty shrinks as you go down the page — 46% at position 3, 19% at position 10.

Pew Research Center measured the behavior directly rather than modeling it: tracking 900 US adults across 68,879 real Google searches, users clicked a search result on just 8% of visits when an AI summary appeared, versus 15% when it did not, and clicked a link inside the summary itself on only 1% of visits (Pew Research Center, March 2025 data).

Now the part nobody puts in the panic deck. The decline stopped. Seer's April 2026 study found organic CTR on AI-Overview queries rebounded from a floor of 1.3% in December 2025 to 2.4% in February 2026 — an 85% increase — and every informational and transactional segment beat their model's projection in January and February 2026. Seer is careful to say two months is not a trend. But 'CTR is in freefall' is no longer what the most current data says.

So the survivable pattern, for a US service business, is this: informational blog clicks down materially, money-query clicks and branded clicks roughly flat, lead volume roughly flat. If your blog traffic fell 30% and your booked jobs did not move, you did not lose anything you were monetizing. If booked jobs fell, the cause is somewhere other than AI Overviews and you should stop looking here.

And if a proposal lands on your desk that guarantees you a spot in AI Overviews: Google's own documentation states there are no additional technical requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible for a snippet. There is no lever to buy. What we sell is the same thing that has always worked — pages that answer the question better than the summary does, on the queries that actually pay.

Want the diagnosis run on your own property instead of a blog post? We will pull your Search Console, run the impressions/position/clicks test on your top 50 queries, and tell you plainly whether AI Overviews cost you anything or whether your rankings just slipped. Get my free audit — no contract, no lock-in, and we will tell you if the answer is 'nothing is wrong.'