Google AI Overviews Now Cover 88% of Commercial Keywords

What happened

Google's AI-written summaries now sit above nearly every high-value shopping search. Over the same period, the AI chatbots that were supposed to become a new source of web traffic still send only a small fraction of what organic search delivers. Semrush published a 12-month study on October 21, 2026, that followed a fixed panel of 400 US desktop commercial keywords in banking, software and ecommerce. AI Overviews appeared on 45.75% of those keywords in September 2025 and on 88% by August 2026.13 When Semrush weighted each keyword by its estimated monthly search volume, coverage reached 97.08%, and the 12% of keywords that never triggered an overview made up less than 3% of the panel's estimated volume.1214

The increase showed up across the board. Searches containing "reviews" went from 30.14% coverage to 89.04%. "Best X for Y" queries reached 96.12%. All five modifier groups Semrush tracked finished above 80%.13 Ecommerce rose the most of the three industries, from 32.62% to 85.11%. In September 2025 the three verticals were more than 35 points apart. By August 2026 the gap was under six points.1317

Trade coverage linked the finding to two other developments from the same week. Fresh traffic studies put ChatGPT visits at roughly 0.62% of Google organic traffic. Separately, Ahrefs reported that Google is sending altered results to SEO data tools.11 Together these suggest search has become harder to win and also harder to measure.

Why the 88% figure needs context

The outlets reporting the study all repeat the same caveats, and they matter. The panel covers only US desktop results, it contains only decision-stage queries, and it records whether an overview appeared, not whether anyone clicked.13 Coverage of the study notes that a 400-keyword panel built from purchase-intent searches says little about the share of all Google queries that get an overview.13 One tracker of AI Overview research makes the same point: a fixed list of hand-picked keywords cannot estimate how common overviews are across search as a whole.16

Broader measurements show how much the panel shapes the result. Semrush's own panel of more than 10 million keywords put AI Overviews at 15.69% of queries in November 2025. Pew counted roughly 18% of real user searches in March 2025. A Similarweb clickstream figure reported in July 2026 put US exposure at 43%.16 An agency study of 218 client sites found overviews on only 22.7% of ecommerce and retail queries, compared with more than 79% for media and education.6 That conflicts with Semrush's 85% ecommerce figure, and the likely reason is the type of query each study sampled.

The readings fit together once you separate scale from importance. Overviews are not on almost every search. They are on almost every search where a buyer is choosing between products, and those searches carry the most commercial value. Earlier research pointed the same way. Semrush reported in July that the share of commercial-intent results with an overview grew 71% between November 2025 and April 2026, across more than 600,000 keywords.13 DemandSphere tracked overviews on branded queries rising from about 26% to above 80% in late September.20

The click problem

The Semrush study does not measure clicks, but other research does, and it consistently finds losses. Ahrefs' updated analysis of December 2025 data links an AI Overview to a 58% lower click-through rate for the top-ranking page, up from the 34.5% gap it reported in April 2025.1 Pew's browsing panel found users clicked a traditional result on 8% of visits that showed an AI summary, compared with 15% without one. Only about 1% clicked a link inside the summary.8 Two randomized experiments now support those correlations. One found outbound clicks per search fell 39.8% when an overview appeared. Another found click-through rose 8.8 percentage points when overviews were hidden.16

Not everything points the same way. Seer Interactive found organic click-through on overview queries fell to 1.31% in December 2025 and recovered to 2.36% by February 2026, though it stayed below the 3.82% baseline for queries without overviews.4 Being cited inside the summary also helps. Seer measured about 20,743 organic clicks per million impressions for cited brands versus about 9,445 for brands that were not cited. Cited brands still trailed pages with no overview by 38%.4 Google's own position remains that total organic clicks are "relatively stable" and that click quality has improved.16 One agency's data partly supports the quality argument: visitors who do click through stay longer and convert to leads more often.6

Our reading is that the quality argument is real but small. Better visitors do not make up for having fewer of them, especially for publishers. Chartbeat data shows search referrals over two years falling 60% for small publishers, 47% for mid-sized ones and 22% for large ones.16 The agency study recorded a 47.3% drop in organic sessions for media and publishing, the steepest of any industry it tracked.6

AI assistants have not replaced the lost traffic

One hope in the industry has been that chatbots would make up for what Google keeps. The numbers do not support that yet. Ahrefs data for September 2026 found AI assistants sent 0.43% of visits across 115,254 sites, while search sent 26.82%.28 Cloudflare Radar data put chatgpt.com at 0.53% of crawler-observed referrals in August, compared with 88.55% for Google.30 Depending on the panel, AI platforms account for between 0.2% and 1.08% of website visits.23

There is growth. ChatGPT's worldwide referral share jumped from 0.23% to 0.32% between April and May 2026, its highest level so far.24 BrightEdge says ChatGPT referrals doubled between January and August, and that it now sends 95.1% of all AI referral traffic.29 The concentration statistics disagree with each other, though. Six panels put ChatGPT's share of AI referrals anywhere from 62.6% to 92.4%, and they do not even agree on whether that share is rising or falling.23 After a May update, about 62–63% of ChatGPT's referrals now go to homepages rather than specific pages.21 That helps brands more than publishers whose business depends on article pages.

Measurement is getting harder

The third development weakens confidence in the first two. On October 22, Ahrefs told customers that Google has been "poisoning" results served to suspected SERP data providers, meaning it returns different results to them than to real users. Ahrefs said 5% to 20% of results are affected, depending on the provider.31 Vendors give very different numbers. Nozzle measured 8.4% of SERPs affected since mid-September, with a peak of 28.6% on September 29. One later report says Nozzle's rate jumped to 35.7% on October 21.32 AccuRanker reported rates below 1% after mid-October validation, and Semrush described the impact on its users as "limited."3337

This matters because a swapped page can make a site appear to drop out of the top 10. Nozzle found tracked sites missing from poisoned pages 60.2% of the time, compared with a 2.7% baseline. The problem also coincided with Google's September 24–October 8 spam update, so real ranking drops and fake ones are mixed together in the same weeks of data.31 Google has not explained the behavior. Some observers think the real targets are AI companies and scraping services.32

The effect on the 88% figure is unclear. The Semrush panel ends in August, before the large wave of altered results that vendors date to late September.1331 Still, any study built on scraped results now carries a reliability question it did not have a year ago.

What it means for SEO and answer engine optimization

The practical conclusion is fairly clear. On commercial queries, ranking first no longer guarantees the first impression, because a summary is now on the page almost every time and it often draws on third-party sources such as Zapier, YouTube and review sites rather than the brand's own page.13 Semrush argues that marketers now need to target and track the AI answer as well as rankings. It also concedes, without traffic data to back it, that a number-one position probably brings less traffic than it did a year ago.13

That makes answer engine optimization, the work of getting cited inside the summary, a necessity. Citation does not bring traffic back to where it was, but it is the clearest lever the data identifies.4 Publishers are in a worse position. Their informational content is the easiest for an overview to answer fully, and the legal and regulatory response is growing, including Penske Media's antitrust suit and a complaint to the European Commission.5 Google's May additions, including "Further Exploration" links and subscription labels, are its response.5 So far, the traffic data has not shown a recovery.