AI Overviews, Google’s generative answer block in search results, now appear in 43 percent of Google searches, up from 15 percent a year earlier. The figure comes from Similarweb’s 2026 Generative AI Landscape report, published Monday and covered by TechCrunch reporter Sarah Perez. A rise from 15 to 43 percent in twelve months is close to tripling. The odds of any single query returning an AI-generated answer are now close to a coin flip.
Read the number for what it is. Similarweb builds these estimates from a panel of monitored devices and browser extensions, then models outward to the broader web population. It is not a census of Google’s index. Google has not published its own trigger-rate data. Similarweb remains a credible, industry-standard source for directional search trends, but it is also a company that sells traffic-estimation and AI-visibility products built on this same panel data, a fact worth holding alongside the 43 percent rather than instead of it.
The 43 percent figure describes prevalence: how often an AI Overview shows up on the results page. It says nothing about clicks. It says nothing about referral traffic, and nothing about what any individual site receives when an AI Overview does appear. A query can trigger an AI Overview nine times out of ten and still send a publisher meaningful traffic, or none at all. Conflating a prevalence rate with a traffic rate is the most common misreading of this kind of data.
Publishers feel a related but separate effect in their own referral logs. Similarweb has previously reported a decline in click-through traffic tied to AI citations that answer a query without sending a visitor anywhere. Cloudflare has since rolled out tools letting publishers block AI crawlers outright unless a company pays for access through its content marketplace. That response addresses lost traffic directly. It does nothing to resolve what the 43 percent prevalence figure actually represents, and the two problems call for different fixes.
Similarweb’s report also tracked AI Mode, Google’s conversational search experience: it counted 126 million visits in June 2025 and 279 million by May 2026, better than doubling inside a year. Average Google query length increased over the same window, which Similarweb reads as users replacing short keyword phrases with longer, conversational ones written for a system that answers rather than an algorithm that ranks.
ChatGPT’s pattern runs the opposite direction. Just 6.8 percent of U.S. ChatGPT desktop queries included a citation as of May 2026, according to the same report, though citation frequency in ChatGPT responses has risen more than fivefold over the past year. Travel, retail, and sports queries generate cited responses more often than other categories, Similarweb found. That is a low base climbing fast, not a maturing prevalence curve like Google’s, and the two trends should not be read as one story.
A May 7 update to ChatGPT’s search product changed the referral picture on its own terms. The share of U.S. desktop ChatGPT visits landing on an actual webpage rose from 25 percent in March 2026 to nearly 60 percent by May 30, 2026. That jump is a product-specific referral change, separate from both Google’s prevalence curve and ChatGPT’s slow-building citation rate, and the three figures should stay three figures rather than merge into a single AI-search narrative.
For a search team, the useful break point sits near that 40 percent prevalence line. Below it, an AI Overview is rare enough that a standard rank-tracking report, position by keyword, still tells most of the story. Above it, position three on a query that fires an AI Overview is a different asset than position three on one that does not. Different visibility. Different click behavior. Different value to the business. The move now is to segment the tracked query set by AI Overview trigger rate before the next reporting cycle, and to log citation inclusion as its own KPI, separate from ranking position. A page can hold position one and never appear inside the AI Overview answering that exact query, and a report that does not distinguish the two is measuring less than it claims to.
Sarah Perez reported this data for TechCrunch on July 27, 2026, citing Similarweb’s 2026 Generative AI Landscape report.
