Seen but Not Clicked: Reading Search Console When Google Answers First

Open the Performance report, tick Clicks and Impressions, and look at the last six months. On a lot of websites the two lines have stopped holding hands. Impressions drift up, clicks slide down, and the average position barely moves. Your page was shown to more people than ever and fewer of them came in. Congratulations, you’re a billboard.

The tempting conclusions are “we’ve been penalised” and “AI stole our traffic”. Sometimes the second one is partly true. Often it isn’t the whole story, and occasionally the chart itself is wrong. This post is about telling those cases apart: what Search Console can and can’t show you in 2026, four patterns behind a page that loses clicks, and what is still worth doing when the results page, not your page, is what changed.

In short
When impressions hold and clicks fall while the position stays put, the results page around your link probably changed: an AI Overview, AI Mode or other features above it. Search Console now has a generative AI report, but it shows impressions only, with no clicks or queries. Before blaming AI, rule out a technical change, a ranking loss, falling demand and plain seasonality, and remember that Google over-reported impressions from 13 May 2025 to 27 April 2026. Some clicks lost to answers on the results page won’t come back through edits on your page. Measure what you can still win, and treat the rest as visibility.

Your page didn’t move. The page around it did.

For most of search history, a stable position meant a stable share of clicks, give or take a season. That link is weaker now. AI Overviews sit above the links for many informational searches, AI Mode answers in a conversation, and both cite sources inside the answer. Your link can keep its place among the results and still start below the first screen of a phone.

How big the effect is depends on who is counting:

• Ahrefs compared 300,000 keywords in December 2023 and December 2025. Where an AI Overview appeared, the top result’s average desktop click-through rate fell from 7.3% to 1.6%. After adjusting for the drop that keywords without an overview also saw, Ahrefs puts the effect at about 58% fewer clicks for position 1, and calls it a correlation.
• SISTRIX looked at over 100 million keywords in Germany. Under an AI Overview, the click-through rate at position 1 drops from 27% to 11%, but the average loss across all keywords is 6.6%, because only about a fifth of the keywords show an overview.
• Chartbeat data reported by Search Engine Journal shows Google Search referrals to its publishers down 40.2% year over year in July 2026. That data leans towards news and media sites, so it says more about publishers than about shops or services.
• Google says its AI features in Search send “billions of clicks to websites every week”, as Nick Fox put it in July. It hasn’t published the data behind that figure.

So the honest summary is: it varies a lot by query type, and for your own site, you’re the only one outside Google who can measure it.

What Search Console shows you, and what it doesn’t

Google added a generative AI performance report to Search Console in June 2026 and rolled it out to all websites on 31 August. It’s useful, but read the label: it counts impressions in AI Overviews and AI Mode by page, country, device and date. It has no clicks and no queries. In October Google told Search Engine Roundtable it will “introduce additional metrics over time”. Until then, it’s a visibility report.

The clicks are in the ordinary Performance report, mixed in with everything else. A few rules from Google’s help page on impressions, position and clicks change how you read that report:

The last row deserves a second look. Google’s data anomalies page says the error “prevented Search Console from accurately reporting impressions” for almost a year, and that after the fix “you may notice a decrease in impressions”. If your impressions line climbed from mid-May 2025 and dropped back in spring 2026, part of that climb may have been bookkeeping. Comparing this autumn’s impressions or CTR with last autumn’s compares real data with a known overcount. Clicks are fine. Use clicks for year-over-year comparisons, and compare impressions only between periods that lie outside that window.

Four reasons a page loses clicks

A page that lost a third of its clicks didn’t necessarily get worse. Look at what happened to its impressions and its position over the same period, and check what the page itself answers today. Four patterns cover most cases:

Helen Pollitt describes the same check in Search Engine Journal’s Ask an SEO column: compare the keywords that now trigger an AI Overview with similar ones that don’t, and look for the case where “rankings and impressions are stable, but click-through rate has declined.” The order matters. Check the technical cause first, because it’s the one you can fix in an afternoon, and the one that would make the other three readings meaningless.

Rule out the boring explanations first

Most panics about AI traffic deserve ten minutes of dull checking before anyone rewrites a page.

The season. Compare the same weeks a year earlier, using clicks. A barbecue guide that lost half its clicks between July and October has had a perfectly normal autumn. With the impressions error above, the year-over-year comparison is one of the few places where clicks are the only number you can trust.

The whole site. If nearly every page dropped at once, single-page explanations are the wrong level. Check a Google ranking update, a site move, a broken analytics or Search Console property, a robots.txt change or a template that lost its titles. One cause, one fix.

Pages that are meant to fade. News, event pages and dated posts lose clicks by design. Judging them by last quarter’s traffic is like judging a newspaper by yesterday’s sales.

Branded and non-branded searches. Searches for your own name behave differently from everything else. Split them before you draw conclusions, or a good month for your brand can hide a bad month for your content.

If the drop survives all four, our post on content decay has the slower routine for deciding whether to update, merge or leave a page.

When the results page changed

This is the uncomfortable case. Nothing on your page broke, your ranking held, and an answer above you now satisfies part of the people who used to click. Some of that loss won’t come back, whatever you edit. What you can still do:

1. Look at the results yourself. Search the page’s main queries on a phone, signed out. Note what sits above your link and what it already answers. Clicks rarely come back to a page that only repeats the summary.
2. Make the title promise what the summary can’t give. A table, a calculator, current prices, stock, photos, a worked example or a local detail. “How long to boil an egg” is answered in one line; “Egg timings by size, altitude and fridge temperature” still has a reason to be opened.
3. Measure the queries you can still win. Track clicks on queries that don’t show an overview, and on searches with clear intent to buy, book or compare. Those are where a better page still pays.
4. Count impressions as visibility, not failure. Being cited in an answer is exposure, even without a click. It just isn’t traffic, so don’t put it in the traffic column.
5. Keep the other channels in perspective. Marie Haynes looked at 12 months of GA4 data across 100 websites: sessions from ChatGPT came to 0.62% of the sessions from Google organic, and 0.73% with all other AI tools included. Chasing AI referrals by gutting your Google pages would be an expensive way to swap a big number for a small one.

What not to do: change the date without changing the page, add FAQ blocks for rich results (Google stopped showing FAQ rich results on 7 May 2026), or rewrite every page that lost clicks in the same week. Some of those pages lost clicks for reasons no rewrite will touch, and you won’t know which edit did anything.

How DeepSpider.ai reads it, and what it doesn’t

When you connect Google Search Console, DeepSpider syncs the last 90 days, the 90 days before and the same 90 days a year earlier, and compares them with the pages it found in the audit. The Pages losing Google clicks report lists pages that had at least 50 clicks in the previous 90 days and lost at least 30% of them. It leaves out pages back to at least 80% of their clicks a year earlier (seasonal), pages that only follow a drop of the whole site, and news, archive and event paths. When 70% or more of the pages dropped, it shows one site-wide row instead of a list, because the cause is probably site-wide.

Each remaining page gets one of the four labels from the table above, in that order. Technical cause in the audit comes first: the audit found a redirect, an error, noindex, a canonical to another URL or a non-HTML response. Then impressions down by 15% or more with a position at least one place worse means Ranking loss, impressions down without a worse position means Less demand, and everything else is Search results changed. These are DeepSpider’s review thresholds, not Google rules, and a label is where to start looking, not a diagnosis.

Four related reports help with the same question:

• Pages with impressions but no clicks: indexable pages with at least 100 impressions, no clicks and an average position of 10 or better.
• Click-through rate below the site’s own curve: rows whose click-through rate is under half of your own site’s median for the same position range, so you compare a page with your site rather than with an industry average.
• Branded and non-branded clicks: clicks split by your website name and up to 20 brand terms you add.
• Google ranking updates on the clicks chart: your daily clicks with the dates of Google’s announced ranking updates. A date on the chart doesn’t prove the update affected you.

What DeepSpider doesn’t see: the results page itself, whether an AI Overview appeared for a query, or Google’s generative AI report. An uploaded Search Console export covers a single period, so the click comparison needs the Google connection. And if a comparison period includes days between 13 May 2025 and 27 April 2026, its impressions were overstated by Google’s logging error, so treat the impression-based labels with care. The year-earlier check uses clicks only, which the error didn’t touch.

A 15-minute routine for your next check

1. Open the Performance report for the last 16 months, with Clicks and Impressions on. Mark 13 May 2025 and 27 April 2026, and stop trusting impressions between them.
2. Compare the last three months with the same three months a year earlier, by clicks. Is the drop seasonal?
3. Check whether the whole site dropped or a handful of pages did.
4. For the pages that lost most, read impressions and position next to clicks, and sort each into one of the four patterns.
5. Fix the technical causes the same day.
6. Search the main queries of the “results changed” pages on a phone, and decide per page whether a better title and something an answer can’t give is worth the effort.
7. Write down what you changed and when, so the next comparison means something.

Frequently asked questions

Has my site been penalised?

Probably not, if impressions held and only clicks fell. A manual action shows up in Search Console’s Manual actions report, and algorithmic drops usually take impressions and positions down with them. A stable position with fewer clicks points at the results page, not at a penalty.

Should I switch off AI features in Search Console?

Since 31 August 2026, every website has a Search Console control that keeps its content out of Google’s generative AI features, such as AI Overviews and AI Mode, including the links in them. Google says the control “isn’t used as a ranking or inclusion signal affecting other parts of Search” and “doesn’t affect AI training”. It also says plainly that you won’t receive any traffic or impressions from those features. If your pages earn clicks from AI answers, you’d be giving those up to protect clicks that may not come back anyway. Look at your own numbers before deciding, and treat it as a business decision rather than an SEO fix.

Why does Search Console show queries like “yes go on”?

Google counts a follow-up question in AI Mode as a new query. When someone types “yes go on” in a conversation where your page was cited, that follow-up can appear as its own query in your report. They’re not typos and not bots.

Will structured data get my pages into AI Overviews?

We don’t claim that, and we haven’t seen evidence that it does. Structured data helps search engines understand a page and can make it eligible for some rich results. Being cited in an AI answer is a separate decision Google makes, and no markup guarantees it.

How long should I wait before acting on a drop?

Long enough to see a pattern. A week of fewer clicks is noise; a quarter compared with the same quarter a year earlier is a pattern. Technical causes are the exception: fix those the day you find them.

The short version

When clicks fall and impressions don’t, the results page probably changed around your link. Search Console’s new AI report counts impressions only, AI clicks are mixed into the normal totals, links in an AI Overview share one position, and impressions before 27 April 2026 were overcounted. Rule out technical changes, ranking losses, demand and seasons first, using clicks for year-over-year comparisons. For pages that really lost clicks to answers, give people something the answer can’t, measure the queries you can still win, and stop expecting every impression to turn into a visit. Being seen still counts. It just doesn’t pay the bills on its own.

Sources