Google AI Overviews appear on 43% of US searches as of May 2026. According to Similarweb's 2026 Generative AI Landscape report, that share climbed from roughly 15% in January 2025 over about 18 months. Over the same period Google AI Mode monthly visits grew from roughly 120 million in June 2025 to near 320 million in March 2026. An AI answer presented above the results is no longer an exception but the default surface. ASAP summarizes this from Similarweb's data and primary reporting.

The path to 43% was not a straight line

The AI Overviews share started near 15% in January 2025 and reached a high of 43% in May 2026. The trajectory is not a smooth adoption curve. Similarweb's tracking shows it rising to about 40% in August 2025, falling to roughly 27% in October 2025, then rebounding to its peak in May 2026. A 13-point drop and a 16-point recovery both occurred inside eight months. That amplitude indicates AI Overviews is a dial Google adjusts rather than an organic uptake curve. The exposure share is tuned between search-quality experiments and revenue structure, so 43% is a current setting rather than a fixed state. When building content strategy on this number, trust the direction but avoid plans that depend on one specific value holding.

Exposure is rising while referrals are not

In a separate Similarweb analysis, Google's standard search referral rate holds at 17 to 19% while AI Mode sits at 1.6 to 2.5%. The same analysis reports that 68% of Google searches now end without a click. Overlaying the two numbers makes the structural shift clear. The share of screens where AI summarizes the answer grew from 15% to 43%, but the rate at which those screens send people to websites is about one tenth of standard search. For publishers, this means exposure and traffic have decoupled. Ranking used to be traffic; now being cited inside an AI answer and that citation producing a visit are two separate events. Any organization treating traffic as its only success metric will record decline even while its actual influence grows.

What a 6.8% ChatGPT citation rate says in both directions

ChatGPT's citation rate rose from 1.3% in June 2025 to 6.8% in May 2026, more than a fivefold increase. The number reads two ways at once. One is growth: the share of answers carrying sources multiplied five times in a year, and after the May 7, 2026 update referrals landing on homepages rose from roughly 25% to nearly 60%. The path from citation to actual visit is strengthening. The other is the ceiling: 6.8% also means the remaining 93.2% of answers carry no source attribution at all. Citation distribution reinforces this, with reviews and user-generated content at 28.9% and news and publishers at 26.0%. More than half of all citations concentrate in community content and journalism, leaving ordinary company sites with far lower odds.

The pie grew while the share split apart

Generative AI platforms averaged 9.5 billion monthly web visits and 655 million monthly unique visitors between June 2025 and May 2026, with visits up 70% year over year. While the total expanded, internal share moved the other way. ChatGPT's share of chatbot web visits fell from above 80% in late 2024 to just over 50% in May 2026, with Gemini, Claude, Perplexity, and DeepSeek absorbing the decline. Read together, the two trends describe the market's character: the pie is growing while the entry points are fragmenting. Optimizing for a single AI service has a shorter shelf life, and structuring content so that any model can cite it stays valid longer.

95% overlap means coexistence, not replacement

Similarweb data puts the audience overlap between ChatGPT users and Google users at 95%. That figure fits poorly with the narrative that AI search is replacing traditional search. What actually happens is overlap rather than substitution. The same person asks an AI during exploration and returns to search for verification and purchase. Similarweb's consumer survey found AI tools ranked above search engines at every stage of the purchase journey except the final transaction. The practical implication is attribution distortion. When a user first encounters a brand through AI but arrives via Google, analytics records that visit as search traffic. The AI contribution disappears from the dashboard and is credited to search. Left uncorrected, this distortion pushes budget away from the channel that is actually working.

Three things to audit right now

Teams should audit three things in the second half of 2026: exposure metrics, citation format, and attribution design. First, is there a system measuring how often the brand is mentioned and cited inside AI answers, separate from rankings and sessions? Second, is the content shaped to be citable? Passing through a 6.8% citation gate requires a first sentence that answers the question directly, verifiable numbers, and clear source attribution. Third, has a survey or referral question been added to correct AI-assisted visits being miscounted as search performance? One caveat matters: this data covers the US market and is largely desktop-based. Korean-language search and mobile-first usage produce different values, so apply the direction while re-measuring the absolute numbers locally.

The caveats that come with this data

Similarweb figures are panel-based estimates rather than measured logs, and they are not official disclosures from Google or OpenAI. Some items vary by reporting date, such as AI Mode visits, which peaked near 320 million in March 2026 before easing to 279 million in May 2026. Individual numbers are therefore more accurate as trends and relative comparisons than as confirmed facts. The direction the data shows is nonetheless consistent. Screens where AI answers first keep expanding, and the rate at which those screens send people to the web stays far lower. Content strategies built on the assumption of web traffic need revision in proportion to how far those two curves diverge.

Source: Similarweb 2026 Generative AI Landscape report (AI Overviews from about 15% in January 2025 to 43% in May 2026, near 40% in August 2025 and 27% in October 2025; AI Mode monthly visits from about 120 million in June 2025 to near 320 million in March 2026 and 279 million in May 2026; ChatGPT citation rate from 1.3% in June 2025 to 6.8% in May 2026; homepage referrals from about 25% to nearly 60% after the May 7 update; citation distribution of 28.9% reviews and UGC and 26.0% news and publishers; 9.5 billion average monthly visits, 655 million monthly unique visitors, up 70% year over year; ChatGPT chatbot web visit share from above 80% in late 2024 to just over 50% in May 2026; 95% audience overlap with Google), Similarweb AI search trends analysis (June 24, 2026; Google referral 17 to 19% versus AI Mode 1.6 to 2.5%, 68% zero-click), and TechCrunch reporting (July 27, 2026). Compiled by ASAP.