Zero-click searches: how to tell if AI Overviews are taking your clicks

What is a zero-click search?

A zero-click search is a search where the person sees the results page and then does not click any result. They may have their answer, they may search again, or they may leave.

Zero-click searches are not new. Weather boxes, featured snippets and knowledge panels have answered questions on the results page for years. A 2024 SparkToro and Datos study estimated that 58.5% of US Google searches ended with zero clicks, and that every 1,000 US searches sent only 360 clicks to the open web (SparkToro, 2024).

What changed is the size of the answer. AI Overviews write a full summary above the results. AI Mode is a chat-style tab where people ask follow-up questions without leaving Google.

Do AI Overviews really reduce clicks? What the studies say

Yes, on the queries where they appear. The studies differ on size, but agree on direction. Our AI Overviews CTR study roundup compares the methods of each one.

Google's view is different. In August 2025 Google said total organic click volume was "relatively stable" year over year and that clicks from AI results are "higher quality" (Google, Aug 6, 2025). Both can be true: total clicks across the web can hold while some sites lose a lot. Your own data is what matters.

How AI Overviews show up in Search Console

Four facts decide how you read your data:

1. AI Overviews and AI Mode are counted in your normal Performance report, under the Web search type. There is no AI filter there (Google).
2. An AI Overview takes one position. Every link inside it gets that same position. A cited page can log a position near 1 even if it ranks lower in the blue links (Search Console Help).
3. The Generative AI performance report (all sites since Aug 31, 2026) shows impressions in AI Overviews and AI Mode by page, country, device and date. It shows no clicks, queries, CTR or position (Search Console Help). Read our guide to the Generative AI performance report.
4. An AI Mode follow-up counts as a new query, with its own impressions and clicks.

Watch two data traps. Around September 10 to 14, 2025, Google stopped supporting the &num=100 parameter used by rank trackers. Many sites saw impressions fall and average position improve overnight, with clicks unchanged (Locomotive). Then, in April 2026, Google fixed a logging error that had over-counted impressions from May 13, 2025 to April 27, 2026; clicks were not affected (Google, Data anomalies). So compare impressions and CTR only between windows that both fall after April 27, 2026. More in why Search Console impressions dropped.

How to tell if AI Overviews are taking your clicks: the 4-sign test

Check one group of queries at a time. If all four signs are yes, AI answers are the likely cause.

If impressions dropped more than 20%, look at demand, indexing or tracking first. If position fell more than one spot, you have a ranking problem, not an AI problem.

The steps in Search Console

1. Open Performance > Search results. Click Date > Compare. Use the last 28 days vs the 28 days before. Avoid year-over-year impressions for now: impressions before April 27, 2026 were over-counted.
2. Turn on all four metrics: clicks, impressions, CTR, position.
3. Click + Add filter > Query > Custom (regex) and paste a group filter (below).
4. Open the Queries tab and sort by Clicks difference. The top 20 losers are your shortlist.
5. Search each shortlisted query in a private window. Note: AI Overview yes or no, and are you cited.
6. Open the Generative AI performance report. Pages with rising AI impressions and falling clicks are your first fixes.

Sample from the guide: two of the 9 regex filters

Questions:

^(who|what|when|where|why|how|which|is|are|can|does|do|should|will)

Long queries (6+ words):

^(\S+\s+){5,}\S+$

All 9 filters, with what each one tells you, are on our Search Console regex page.

Which searches lose the most clicks?

The pattern: the more a search is a question someone can answer in three sentences, the more clicks it loses. Buying searches still send clicks.

A worked example (fictional)

northwind-outdoors.com is a fictional outdoor gear shop with a blog. Blog clicks fell 38%, shop clicks were flat. On the "Questions" regex group: impressions +8%, position 3.4 to 3.2, CTR 5.9% to 3.4%. All four signs said yes. One query, "how to waterproof hiking boots", held position 2 but CTR fell from 6.8% to 2.9%, and the AI Overview listed the steps. They changed the title to "How to Waterproof Hiking Boots (We Tested 5 Methods for 30 Days)", put the full short answer in the first two sentences, and added their own soak test with photos and a product pick per boot type. They judged success by clicks from the blog to product pages, not by last year's blog traffic.

How to win clicks back from AI Overviews

Google says AI features need no special tricks: pages must be indexed and eligible for a snippet, and normal SEO applies (Google). So aim every change at one goal: a reason to click after reading the summary. To be the source the summary links to, see how to get cited in AI Overviews.

• Answer first. Give the short answer in the first two sentences. Hiding it makes you less likely to be cited.
• Put the reason to click in the title. A test, a tool, a price, a local fact, a template.
• Add what a summary cannot replace: your own data, real photos, calculators, live prices and hours, and a clear recommendation.
• Use structured data for rich results, not for AI. Google says no special schema is needed for AI features (Google, July 2026). FAQ rich results stopped showing on May 7, 2026 (Google).
• Grow your brand. People who know your name click you even under an AI Overview. Fix wrong facts about you at the source.
• Do not add llms.txt for Google, chop pages into chunks, or mass-produce AI articles. Google says the first two do nothing for Search; the third can break spam policies.

Who this is for

Site owners, marketers and small agencies who see clicks falling in Search Console and want to know if AI Overviews are the cause, and what to do about it. No paid tools needed.

Use SEOmatic to check whether AI Overviews and AI Mode are taking clicks from my site, mysite.com, following the "Zero-click searches / AI Overviews clicks guide" (Updated October 2026).

Step 0. If I am not signed in, run check_site on mysite.com first, then connect_seomatic so I can sign in and connect Google Search Console.

Step 1. Compare periods.
- gsc_performance action compare_periods with dimension query.
- gsc_performance action compare_periods with dimension page. Note: this compares the latest period with the preceding period of the same length. If either window falls before April 27, 2026, warn me that impressions and CTR are skewed: Google over-counted impressions from May 13, 2025 to April 27, 2026, and the num=100 change cut impressions in mid-September 2025. Clicks are not affected.

Step 2. Group my queries. Using the query rows, sort each query into: questions (starts with who/what/when/where/why/how/which/is/are/can/does/do/should/will), long (6+ words), comparisons (vs, versus, compare, difference between, alternative, best), buying and local (buy, price, cost, near me, book, quote), and brand (my brand terms: ___). Also run gsc_insights action brand_split with my brand terms.

Step 3. Run the 4-sign test per group and show a table: impressions, average position, CTR and clicks for both periods. Flag a group as "likely AI Overviews" only if impressions are within -10% or higher, position moved less than 1, CTR fell 25% or more, and the loss is bigger in questions/comparisons than in brand/buying. If impressions fell more than 20% or position fell more than 1, say it is probably NOT AI Overviews and say what to check instead.

Step 4. Run gsc_insights action ctr_outliers to list pages with CTR far below normal for their position. Cross them with the top losing pages from Step 1.

Step 5. Read ai_citations action observed_citations to see where my site has been seen cited in AI answers. If I ask for a fresh scan, run run_ai_visibility_scan (it costs AI credits, max 1 per hour) and read the result with ai_citations action scan_results.

Step 6. For the 5 pages that lost the most clicks while keeping position, use site_pages action crawl to read the current title, meta description and first heading, then draft: a title that gives a reason to click (a test, data, price, tool or local fact), a 2-sentence opening answer, and a meta description that promises the next step.

Step 7. Stage those drafts with task_manage action create, type ctr_fix, including the affected queries, and show them with show_approval_cards. Do NOT publish anything: I will approve or dismiss each one.

Finish with: the groups table, the 5 pages, and one line on what to re-check in 28 days.