AI Overviews can answer a question before someone visits a website. That makes them a plausible contributor to fewer clicks, but the timing of a traffic drop does not establish its cause. Start by identifying which measurement changed and which pages are affected.
This guide investigates Google search traffic. It does not assume access to your analytics or claim that every publisher's decline comes from AI. For the broader planning question, see what remains worth doing in SEO.
What the Evidence Shows, and What It Does Not
Pew Research Center's July 2025 analysis of 900 U.S. adults found traditional-result clicks on 8% of visits with an AI summary versus 15% without. This observational 2025 sample is not a Malaysian or 2026 estimate, or proof of the cause of one site's decline.
Searches with and without an AI answer can differ in purpose and wording. A comparison between those groups does not isolate the effect for your page. Equally, a page can be useful and lose clicks: a decline is not sufficient evidence of poor content.
1. Confirm What Actually Fell
Website analytics sessions and Search Console clicks are different measurements. If only analytics falls, check recent consent, tagging, reporting and deployment changes before editing content. If Search Console clicks also fall, continue with the affected pages and queries.
Record the start date, report, search type and comparison period. Check ordinary seasonal changes and whether the comparison includes the same mix of weekdays. Search Console's latest data can be incomplete. A sudden sitewide loss also warrants checking availability, indexing restrictions, redirects, manual actions and security issues.
The Search Console metrics guide explains what clicks, impressions, click-through rate (CTR) and average position can establish.
2. Separate Pages and Query Groups
Consider an appliance retailer with a washing-machine care guide, comparison pages and actual product pages. This is an illustrative diagnosis, not a report of a client's data. If only definition queries lose clicks, the next check differs from a loss across every product page.
In Search Console's Search results report, compare dates and inspect Pages. Choose an affected page, then inspect Queries, Countries and Devices. Separate brand searches from relevant non-brand tasks where the data allows. Keep the query list's omissions and row limits in mind.
| Observed pattern | Possible explanation | Next check |
|---|---|---|
| Impressions and clicks both fall | Demand, eligibility, relevance or ranking changed | Indexing, affected query groups, comparison dates and current results |
| Impressions broadly hold, clicks fall | Result layout, query mix or title/snippet appeal changed | Comparable query groups and visible search features; do not assume AI alone |
| Analytics falls but Search Console clicks do not | Measurement or landing-page differences may contribute | Tagging, consent, redirects and report definitions |
| Visits hold, useful enquiries fall | Offer, audience mix or enquiry path changed | Real product terms, contact function and sales records |
A stable average position can hide changes within the query mix. Compare relevant groups rather than interpreting one sitewide average as a controlled experiment.
3. Inspect the Relevant Search Experience
For a small sample of affected queries, record the exact phrase, date, language, country or location context, device and page shown. Note whether AI answers, ads, local results, shopping or other features appear. Save what was observed separately from your interpretation.
A result viewed today does not reconstruct what everyone saw before the drop. Repeating relevant samples can reveal a pattern, but cannot recover an unrecorded baseline. If you have no earlier observation, keep “AI layout change” as a hypothesis rather than a confirmed cause.
Also inspect the destination. For the retailer, check whether the guide still answers the care question and whether the product page has current dimensions, delivery and installation information. Missing or outdated details may need correction regardless of the traffic explanation.
4. Use the AI Report for What It Measures
As of October 2026, Search Console has a Generative AI performance report for supported Google Search AI features. It shows link impressions and page, country, date and device breakdowns. Google says the report may not appear without sufficient impressions. It is not a full dataset of brand mentions or a complete separate AI-click attribution report.
Check the property's Search generative AI inclusion choice and inherited parent setting if expected visibility is absent. Exclusion affects supported AI features, not ordinary Search ranking. Do not confuse this control with Google-Extended's training control.
Ordinary Search performance includes AI-feature traffic in Web data. Compare that evidence with actual customer outcomes without claiming the reports expose every hidden query or explain each lost visit. Our AI-search measurement guide covers the reporting boundary.
Choose a Response the Evidence Supports
If the retailer's guide is indexed and accurate but fewer people click, making it longer is not an automatic remedy. Check whether it still serves a useful learning task and offers a relevant next step, such as a genuine comparison or product detail. Do not manufacture personal experience or delete it solely for low Google clicks.
If the page has an access problem, fix that first. If visitors cannot find delivery conditions, supply them. If a contact route fails, repair it. Record the change and compare the same page groups later, alongside demand and enquiry quality. A subsequent improvement is evidence to assess, not proof that the one edit caused it.
Frequently Asked Questions
Does Lower CTR Prove AI Overviews Took the Clicks?
No. Query mix, ads, other result features, titles and snippets can also change click-through rate. Investigate comparable groups and observed results before attributing a cause.
Does an AI Citation Guarantee Traffic?
No. A source link can be shown without being clicked. Measure link visibility, site visits and useful customer actions separately.
Should I Delete Articles That Lost Search Traffic?
Not on that evidence alone. Check accuracy, reader purpose, other traffic sources and existing links before deciding whether to improve, combine, retain or remove a page.