Updated Jack F, Co-founder
Part of the Answyn AEO Handbook.
Now readingWhy traffic falls when rankings hold
AI answers can take the click while rankings hold. If clicks fall while impressions and position stay steady, something on the results page is taking it, and an AI Overview is one likely candidate: Pew Research found Google users clicked a traditional result on 8% of visits when an AI summary appeared, against 15% without (US, March 2025). Confirm it query by query in Search Console, and rule out seasonality, tracking changes and lost snippets before blaming AI.
The question usually arrives like this: organic sessions are down on last year, the rank tracker shows the same positions, and someone asks whether ChatGPT or Google's AI Overviews are answering before people click.
Sometimes they are, and sometimes the cause is more ordinary. The two leave different marks in your data. This page sets out the checks that tell them apart, what each one can and cannot prove, and what the published studies measured.
Why does traffic fall when rankings hold?
Because a ranking measures where your link sits, not whether people click it. Clicks are impressions multiplied by click-through rate. When impressions and position hold and click-through falls, something on the results page is taking the click, or the search is ending without one. An AI Overview that answers the question is one candidate. A featured snippet, a video block or a change in what searchers wanted are others.
A search that ends on the results page, with no click to any website, is a zero-click search. Pew Research found that Google users ended their browsing session on 26% of results pages with an AI summary, against 16% of pages with only traditional results (900 US adults, March 2025, Pew Research).
Search Console's counting rules add a wrinkle. It reports the topmost position a link to your site holds, and an AI Overview occupies a single position, with every link inside it assigned that same position (Search Console Help). If an AI Overview at the top of the page cites you, your reported position can hold or even improve while the click on your ordinary result goes. Steady rankings are not proof that the page searchers see stayed the same.
Six checks, in order
Work through these in order. The first two need only Search Console and a browser; the later ones need analytics, server logs or tracking. The last check is last on the list, not last in importance: if Search Console clicks held and only analytics sessions fell, go straight to it.
| Check | What it can show | What it cannot show |
|---|---|---|
| 1. Search Console, per query and page | Which queries and pages lost clicks while impressions and position held, and which pages appear in AI features | Which queries showed an AI Overview, or how many clicks it took |
| 2. AI Overview presence | Whether an AI Overview appears for the affected queries, and which pages it cites | That it caused the drop |
| 3. AI referrals in analytics | The visits AI answers still send you | The visits they did not send |
| 4. AI crawler logs | Which assistants fetch which pages, and whether a fetch is live or for training | Whether you were cited or named, or whether anyone read the answer |
| 5. Named-mention tracking | Whether you or a rival is named and cited for those questions, by engine | A one-for-one link between a mention and a lost click |
| 6. Ordinary causes | Whether seasonality, tracking, lost features or site changes explain the drop | Anything about AI, which is why it sits beside the other five |
First, Search Console, query by query. Compare the same weeks year on year, so seasonality cancels out, and sort queries and pages by lost clicks. The pattern that points at the results page is impressions and position steady or up, with click-through down. Clicks from AI Overviews and AI Mode are counted inside the Web search type of the Performance report, mixed in with ordinary results (Google Search Central).
Search Console also has a separate generative AI performance report, announced on 3 June 2026 (Google Search Central blog) and rolled out to all sites by 31 August 2026. It shows how often links to your site appeared in AI Overviews and AI Mode, by page, country, device and date. Google's help page describes one metric, impressions: clicks, click-through rate, position and queries are not in it (Search Console Help).
Second, find out which of the affected queries show an AI Overview. Search the queries that lost the most clicks, signed out, from the UK, and note whether an AI Overview appears and which pages it cites. Do it on more than one day, because presence varies: across 750 daily checks of 25 questions in Answyn's own tracking, Google showed no AI Overview 30 times (6 October to 4 November 2026). Daily tracking of the same queries replaces the sampling.
Third, look at AI referrals in your analytics. Segment the sessions your analytics attributes to AI assistants and set that trend beside the organic decline. A rise there alongside a fall in organic search is consistent with buyers moving part of their research into assistants. It still cannot show the visits you lost, because an answer that satisfied the reader leaves no trace in your analytics. Answyn's LLM Analytics records referral clicks from cited links in AI answers, alongside crawler fetches.
Fourth, check your server or edge logs for AI crawlers. OpenAI's crawler documentation describes OAI-SearchBot, which surfaces websites in ChatGPT's search features; ChatGPT-User, used for certain actions a user starts in ChatGPT, to which robots.txt rules may not apply; and GPTBot, which crawls content that may be used to train OpenAI's foundation models (OpenAI). A fetch proves an engine can reach the page and read it. It does not prove the page was cited, that you were named, or that anyone read the answer.
AI crawlers largely do not run JavaScript, so a browser analytics tag misses most of their fetches. You need server, CDN or edge logs, which is how LLM Analytics records crawler fetches and separates live fetches from training crawls. If the logs show no AI crawlers at all, run the AI crawler access check to see whether robots.txt or a firewall is turning them away.
Fifth, find out whether you are named in the answers to those questions. Turn the queries that lost clicks into the questions a buyer would ask, put them to the engines your buyers use every day, and record whether you are named, who is named instead and which pages are cited. One manual check is a single draw from an answer that changes from day to day; daily records over weeks are a measurement. If the answer names you and cites your page, you are still in front of the buyer, without the visit. If it names a rival and cites a roundup, the fix is on someone else's site.
This is also where reading a crawl as visibility goes wrong. In Answyn's own tracking, ChatGPT retrieved an Answyn page in 49 runs across 5 questions and left Answyn out of the answer every time (6 October to 4 November 2026).
Sixth, rule out the ordinary causes. Any of these can produce the same pattern on its own:
- Compare with the same weeks last year, not last month, so seasonal demand cancels out.
- Check for consent or analytics changes. A new consent banner, a tag change or a migration moves sessions without moving searches; if Search Console clicks held and analytics sessions fell, the change is in measurement.
- Look for lost featured snippets and other results features. Losing a snippet can cut clicks while the ranking barely moves.
- Review site changes: migrations, redirects, noindex or canonical changes, and template changes that removed content searchers came for.
What the published studies show
Four reports from three publishers, each with a stated sample and each read on 4 November 2026, point the same way: lower click-through where an AI Overview appears. None is a controlled experiment, and none says it is a UK sample. Pew's panel is US adults; Ahrefs and Seer do not name a market.
| Study | Published | Sample | Finding |
|---|---|---|---|
| Pew Research | 22 July 2025 | Browsing data from 900 US adults; 68,879 Google searches in March 2025 | Users clicked a traditional result on 8% of visits to pages with an AI summary and 15% without. They clicked a link inside the summary on 1% |
| Ahrefs | 4 February 2026 | 300,000 keywords; Search Console desktop click-through, December 2023 against December 2025 | An AI Overview correlated with 58% lower click-through for the top-ranking page, against a forecast without one. Its April 2025 study found 34.5% |
| Seer Interactive (2025 update) | 4 November 2025 | 3,119 informational queries across 42 client organisations, June 2024 to September 2025 | Organic click-through on queries with an AI Overview fell from 1.76% to 0.61%, down 61%. Queries without one fell 41%, to 1.62% |
| Seer Interactive (2026 update) | 24 April 2026 | 53 accounts, 5.47 million queries, January 2025 to February 2026 | Organic click-through with an AI Overview fell to 1.31% in December 2025, then rose to 2.36% in February 2026, against 3.82% without one |
Read them as direction, not as a forecast for your site. Ahrefs states its headline finding as a correlation, and Seer's 2026 update says it cannot claim causation. The baseline moved too: click-through on queries without an AI Overview fell 41% in Seer's 2025 data and rose in its 2026 data. Clicks change for reasons that have nothing to do with AI, which is why the sixth check is on the list.
Google's position is that clicks from results pages with AI Overviews are "higher quality", meaning people are more likely to spend more time on the site. The page that says so gives no figures for it (Google Search Central).
What to do about it
If the checks point at AI answers, you cannot make the results page show fewer of them. What you can work on is what the answer says: whose page it cites and whose name it gives.
Be the cited and named source. In Seer's 2026 data, informational queries where the brand was cited in the AI Overview averaged 2.07% organic click-through over the year, against 0.94% where it was not cited and 3.35% where no AI Overview appeared. Seer says it cannot claim causation, because higher-authority brands are also more likely to be cited, so treat it as a reason to compete for the citation, not as proof that one brings the clicks back. What earns one is set out in how brands earn citations in AI answers.
Update the pages assistants already read. A page an assistant fetches and does not cite is within reach and not persuading. Answyn's Pages screen puts citations, AI crawls and Search Console clicks for each URL on one row, with a filter for pages crawled and never cited. Those are usually the cheapest pages to improve, and what counts as a worthwhile update is in the freshness handbook.
Measure named mentions as well as clicks. When clicks fall for reasons you cannot reverse, the useful question becomes whether answers still name you for the questions that used to send traffic. Track those questions every day, by engine, and report mentions and citations beside clicks, so a fall in visits and a fall in visibility are never mistaken for each other.
Answyn keeps that daily record of who is named and cited for the questions you track, and every figure opens back to the answer behind it. Access is by invitation, and every plan is quoted for the business using it. Enquire if you want it for your brand, or read how Answyn pricing works.
A note on sources
External figures come from the studies and documentation listed under Sources, each read on 4 November 2026 and dated where it appears. We cite Seer Interactive's own reports rather than press coverage of them.
Answyn's figures come from its own tracking of its category: 25 questions asked in the UK of ChatGPT, Google AI Overviews and Claude, 6 October to 4 November 2026, 1,575 answers. Claude is not on Answyn's standard roster, and a plan adds it only by agreement; we collect it for our own project.
Questions people ask
- Can AI Overviews reduce clicks when our rankings haven't changed?
- Yes. A ranking measures where your link sits, not whether people click it, and when an AI Overview answers the question fewer people need to. Search Console also gives every link inside an AI Overview the Overview's single position, so being cited there can keep your reported position steady while clicks fall.
- Does Search Console show AI Overview traffic separately?
- Partly. Its generative AI performance report, announced on 3 June 2026 and rolled out to all sites by 31 August 2026, shows impressions of your URLs in AI Overviews and AI Mode by page, country, device and date. It has no clicks and no queries; clicks from AI features are counted inside the Web search type of the main Performance report.
- How much do AI Overviews reduce click-through rates?
- The studies agree on direction and differ on size. Pew Research found Google users clicked a traditional result on 8% of visits with an AI summary against 15% without (US, March 2025), and Ahrefs found 58% lower click-through for the top-ranking page (December 2025). Seer Interactive saw organic click-through on AI Overview queries recover from 1.31% to 2.36% between December 2025 and February 2026, still below the 3.82% on queries without one.
- Do AI crawler visits mean we appear in AI answers?
- No. A fetch shows an engine could reach and read the page, not that it cited the page or named you. In Answyn's own tracking, ChatGPT retrieved Answyn's pages in 49 runs and left Answyn out of every one of those answers (6 October to 4 November 2026).
- Should we block AI crawlers to protect our traffic?
- Blocking takes you out of the answers without stopping the answers. OpenAI says sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links. Google's nosnippet, max-snippet and noindex controls limit what Search shows from your pages, AI Overviews included.
- What is zero-click search?
- A search that ends on the results page without a click to any website, for example because an AI Overview answered it. Pew Research found Google users ended their browsing session on 26% of results pages with an AI summary, against 16% of pages with only traditional results (US, March 2025).
Sources
- Pew Research, Google users are less likely to click on links when an AI summary appears in the results (22 July 2025)
- Google Search Central blog, Search Console generative AI performance reports (3 June 2026)
- Search Console Help, Generative AI performance report (Search)
- Search Console Help, What are impressions, position, and clicks?
- Google Search Central, AI features and your website
- OpenAI, Overview of OpenAI crawlers
- Ahrefs, Update: AI Overviews reduce clicks by 58% (4 February 2026)
- Seer Interactive, AIO impact on Google CTR: September 2025 update (4 November 2025)
- Seer Interactive, AIO impact on Google CTR: 2026 update (24 April 2026)