What the Semrush study found
Semrush checked the same 400 U.S. desktop keywords each month from September 2025 through August 2026. It asked one question: did an AI Overview appear?
The panel covered banking, ecommerce, and software. Coverage rose from 45.75% to 88%—a gain of 42.25 percentage points.
AI Overview coverage in the 400-keyword panel
U.S. desktop · same fixed keywords
Source: Semrush, October 21, 2026. Share of tracked keywords with an AI Overview; no clicks measured.
Approximate labels reflect Semrush's reported values and thresholds; they do not describe every query in each category.
Across industries
Ecommerce had the largest rise: 32.62% to 85.11% of its selected keywords. Banking and software also ended above 85%.
Within ecommerce
Price and cost queries rose from 4.76% to 71.43%. Twelve of the 15 industry-and-modifier groups ended at or above 80%.
The numbers come from Semrush's original study. Search Engine Land, which is owned by Semrush, also summarized it. The study tracks Overview appearance, so it cannot tell us whether an individual site lost traffic.
How to read the 88% and 97% figures
Both percentages describe the same 400-keyword panel. They answer different questions:
88%
Keywords with an Overview
Each tracked keyword counts once, regardless of its estimated search volume.
97.08%
Estimated volume represented
Higher-volume keywords contribute more. This is estimated demand within the same panel.
Neither is a click-loss rate. The study did not measure answer quality, citation clicks, leads, or revenue.
The boundary of the study
The selected modifiers cover three industries on U.S. desktop. They are not a random sample of every buying query, country, or device. A monthly SERP observation also cannot represent what every searcher saw.
What changes on buying searches
A commercial query asks for help choosing. An AI Overview can give the first framing before someone opens a vendor page, review, or video.
Buyer asks
A comparison, price, or review query
AI frames
The Overview gives an initial answer
Sources shape
Official and third-party pages supply facts
Visitor decides
Clicks and conversions reveal the outcome
Pricing
Check current prices, plans, and qualification details.
Reviews
Check product facts and legitimate independent feedback.
Comparisons
Explain the use case and tradeoff for each option.
Accuracy matters even when your official page ranks. In Semrush's software-comparison example, the Overview drew on a vendor guide, a third-party comparison, and a YouTube video. A single organic position misses that source mix. Editing one source does not guarantee Google will cite it.
How other evidence fits
Three sources answer different parts of the question. Their percentages should stay attached to their own samples.
8% vs. 15%
Clicks in Pew's 2025 browsing study
Pew Research Center observed traditional-result clicks on 8% of visits to pages with an AI summary, versus 15% without one, among 900 U.S. adults. The study covered many query types and cannot estimate conversions for Semrush's commercial panel.
82.91%
Branded results in Ahrefs' crawl
Ryan Law's Ahrefs analysis found this rate on U.S. desktop branded-keyword crawls on September 29, 2026. Branded queries and the method differ from Semrush's commercial panel.
Google guidance
Eligibility and useful content
Google Search Central says pages need to be indexed and snippet-eligible. It recommends original, useful content and says no special AI markup or llms.txt file is required. None of this guarantees a citation.
A commercial-query audit you can run this week
Start with 20 to 30 of your valuable queries. Use the same fields each time so the audit becomes a baseline.
Choose queries by business value
Pull comparison, alternatives, best-for, review, and pricing terms from Search Console and sales conversations. Include brand, competitor, and category terms. Use the Keyword Search Volume Checker to compare demand after identifying buyer language.
Capture the result
Record date, country, device, Overview presence, major claims, visible citations, organic position, and nearby ads or shopping modules. Repeat the check. The AI Overview Analyzer can inspect a sample; check whether each result uses live SERP data or an AI estimate.
Mark factual gaps
Save a screenshot and source URL for an old price, wrong integration, missing limitation, or outdated review. Distinguish an error from Google choosing another valid source.
Find the source of truth
Identify the official pricing, product, comparison, help, or merchant-feed record. Name its owner and last update date. Note which reputable third-party pages carry facts that may need correction.
Check eligibility and outcomes
Confirm the key page is indexable, canonical, available, and snippet-eligible. A Technical SEO Audit can flag common issues. Compare Search Console clicks and visibility with actual leads or sales.
Keep one audit row per query
Query · device and country · Overview · answer accuracy · cited domains · official page · owner · next action · review date
What to improve after the audit
1. Correct the facts
Update stale prices, availability, features, and eligibility rules on the official page. Give the decisive fact enough context to avoid a misleading excerpt.
For ecommerce, keep Merchant Center data current too. Product feeds can help surface product information; they do not force a citation.
2. Improve the comparison
Add firsthand testing, clear use cases, and honest limitations. A Helpful Content Checker can prompt a review, while a product owner verifies the facts.
3. Make the click useful
Put plan details or a comparison summary near the top of the landing page. Use SERP Preview for the organic snippet, keeping in mind that Google may rewrite it.
4. Keep pages distinct
Google says special AI markup and page lengths are unnecessary. Maintain a smaller set of useful pages instead of near-duplicates for every query variation.
See our branded AI Overview audit for brand queries and zero-click search analysis for the wider click picture.
How to measure progress without inventing a ranking
One AI answer is a snapshot. Track these three observations separately across a fixed query set:
Overview present?
Did Google show an AI answer for the query?
Brand mentioned?
Did the answer name your brand accurately?
Page cited?
Did it link to an official page you own?
A mention does not require a citation, and a citation does not guarantee a recommendation. Keep date, country, device, and query list consistent between checks.
Connect visibility to outcomes
- Google's generative AI report shows impressions, pages, countries, devices for Search, and dates. It rolled out worldwide by August 31, 2026.
- The regular Performance report gives query and page trends. Analytics and CRM data show meaningful leads or sales.
- Review after a justified correction and recrawl. Check that the source is accurate and eligible, then compare the observed answer and landing-page outcomes.
The dedicated AI report is aggregate visibility data, not a transcript of every answer. An increase in impressions with fewer qualified leads needs investigation, and a wrong price needs correction even when clicks stay steady.
What to watch next
Semrush's measurement ended in August, although its report appeared in October. The next useful studies would:
- Repeat the same panel in later months.
- Add mobile results, more countries, and independent keyword panels.
- Connect Overview appearance with observed clicks and conversions.
For your site, keep watching the top buying queries. Note when the answer changes, whether the official source appears after a correction, and whether qualified traffic follows. One screenshot is not a permanent ranking.
Frequently asked questions
Sources
- Semrush: commercial modifier study and methodology
- Search Engine Land: coverage of the Semrush study
- Google Search Central: guide to generative AI features
- Google Search Central: generative AI reports in Search Console
- Pew Research Center: clicks on results with AI summaries
- Ahrefs: branded-query AI Overview study
- Google: Search Console Performance report
About the Author
Co-Founder & SEO Execution
Co-founder of PikaSEO. 11 years in corporate tech, then bootstrapped entrepreneur. Leads SEO execution and content-led growth for SaaS companies.