AI search statistics 2026: the numbers that matter

800 million people used ChatGPT weekly as of September 2025, a figure Sam Altman confirmed publicly, and OpenAI reported roughly 900 million weekly active users by February 2026. For scale, that September figure represented close to 10 percent of the world's adult population, and it had doubled from 400 million in February 2025. Whatever else is contested in this category, the raw audience is enormous and still compounding.

50 percent of consumers already use AI-powered search intentionally, according to McKinsey's AI Discovery Survey fielded in August 2025 and published that October. The same research found between 40 and 55 percent of consumers in top sectors, including consumer electronics, travel, beauty and financial services, using AI-based search to make purchasing decisions, with adoption spanning every age group including a majority of baby boomers. Bain's February 2025 survey adds the intensity dimension: about 80 percent of search users rely on AI-written summaries in at least 40 percent of their searches.

One widely quoted number deserves a flag. Gartner predicted in February 2024 that traditional search engine volume would drop 25 percent by 2026. That was a forecast rather than a measurement, and as of mid 2026 it has not materialized at that magnitude: Google still handles the overwhelming majority of search volume. Quote it as a dated prediction, never as an observed fact.

94 percent of business buyers used AI during their most recent purchase, according to Forrester's 2026 Buyers' Journey Survey of 18,000 global business buyers, published January 2026. In the same survey, 55 percent used AI tools to compare vendors against each other, 54 percent to research product information, and 47 percent to build internal business cases before contacting any vendor. AI tools outranked vendor websites, product experts and direct sales contact as meaningful sources of purchase information. The same survey counted 13 internal stakeholders and nine external influencers in a typical buying decision, which means one AI answer can now brief a whole committee at once.

51 percent of B2B software buyers now start research with an AI chatbot more often than with Google, up from 36 percent seven months earlier, per G2's 2026 AI Search Insight Report, a March 2026 survey of 1,076 buyers. The consequences are blunt: 69 percent chose a different vendor than they originally expected because of what an AI chatbot told them, 33 percent bought from a brand they had never heard of before the AI named it, and 83 percent reported feeling more confident in their final choice. The fuller picture of this shift is in how B2B buyers research with AI.

$750 billion in US consumer spending will flow through AI-powered search by 2028, per McKinsey's October 2025 projection, which also warns unprepared brands to expect traditional search traffic declines of 20 to 50 percent. Against that backdrop, McKinsey found just 16 percent of brands systematically track their AI search performance today. That gap between where decisions form and what companies measure is the single most actionable statistic on this page.

8 percent versus 15 percent: that is the click rate on traditional search results with and without an AI summary present, from Pew Research Center's July 2025 analysis of 68,879 real Google searches by 900 US adults. Only 1 percent of visits with an AI Overview produced a click on a link inside the summary itself, and 26 percent of AI-summary encounters ended the browsing session entirely, versus 16 percent without one.

34.5 percent is the drop in position 1 click-through rate when an AI Overview is present, from Ahrefs' April 2025 study of 300,000 keywords. By December 2025, Ahrefs' updated data put the reduction at 58 percent for the top-ranking result, so the effect intensified through the year. We break down the full set of CTR studies, and where they disagree, in our AI Overviews click impact analysis.

68 percent of US Google searches ended without any click in the first four months of 2026, per SparkToro's analysis of Similarweb clickstream data, up from 60.45 percent in 2024. Bain's independent February 2025 survey landed nearby, with 60 percent of searches terminating without a click-through and organic web traffic down an estimated 15 to 25 percent. For news specifically, Similarweb measured no-click news searches growing from 56 percent to nearly 69 percent as of May 2025.

1.13 billion referrals flowed from AI platforms to the top 1,000 websites in June 2025, up 357 percent year over year, per Similarweb, with ChatGPT accounting for more than 80 percent of them. ChatGPT referrals to news sites grew 25x, from under 1 million in early 2024 to more than 25 million over the same months of 2025. The volumes remain small next to what search removes, which is the honest framing, but the trajectory is steep.

4.4x is the value multiple of an AI search visitor over a traditional organic visitor, measured by conversion, from Semrush's June 2025 study of more than 500 topics. Adobe Analytics data points the same direction: visitors arriving from generative AI sources spent 32 percent longer on US retail sites, bounced 27 percent less, and during the 2025 holiday season converted 31 percent more than other traffic sources, with AI-driven revenue per visit up 254 percent year over year.

4,700 percent is Adobe's measured year-over-year growth in generative AI traffic to US retail sites in July 2025, and the firm had already logged a 1,950 percent year-over-year spike on Cyber Monday 2024. Holiday 2025 kept the pattern: retail AI referrals up 693 percent, travel up 539 percent, financial services up 266 percent. Growth percentages this large partly reflect a small base, so pair them with the absolute referral counts above when you cite them.

388 percent: the year-over-year growth in referral traffic from Google's Gemini between September and November 2025, per Similarweb, against 52 percent growth for ChatGPT referrals over the same window. ChatGPT still dominates absolute volume, but the second player is closing at a very different slope, which is why single-engine tracking undercounts the surface.

On Google itself, Pew found 58 percent of US adults in its panel ran at least one search that produced an AI summary during March 2025 alone, so exposure to AI answers is already a mainstream Google experience rather than an early adopter one. Engines also differ in what they read: 5W's 2026 index found only 36 percent of Claude's journalism citations came from the past 12 months, versus 56 percent for ChatGPT, so the same question can be answered from different eras of the web depending on the engine.

More than 25 percent of US ChatGPT citations come from just two domains: Wikipedia at 13.15 percent and Reddit at 11.97 percent, per 5W Research's 2026 analysis. The firm's AI Platform Citation Source Index synthesized over 680 million citations across ChatGPT, Google AI Overviews, Perplexity, Gemini and Claude, and found Reddit at or near the top for every engine. The Wall Street Journal, The New York Times and Bloomberg were absent from ChatGPT's top 20. The per-engine breakdown is in our guide to ChatGPT citation sources.

Under 1 percent: the chance that two runs of the identical prompt in ChatGPT return the same list of recommended brands, from the SparkToro and Gumshoe.ai study of late 2025, in which 600 volunteers ran 12 prompts a combined 2,961 times across ChatGPT, Claude and Google's AI surfaces. The same list in the same order appeared less than 0.1 percent of the time. The researchers' conclusion is the methodological core of this whole field: track appearance rate across many runs, because a single check is noise. Why answers move this much is covered in why AI answers change between runs.

Citation shares are volatile at the platform level too. 5W's index documents Reddit's share of ChatGPT citations falling from roughly 60 percent to roughly 10 percent within two weeks in September 2025. A source mix that can swing that hard in a fortnight is another argument for continuous measurement over annual audits.

Up to 40 percent: the visibility lift inside generative engine responses from adding statistics, quotations and cited sources, measured by the Princeton and Georgia Tech team that coined the term GEO in the paper published at KDD 2024. The study tested nine optimization methods across 10,000 queries in its GEO-bench benchmark, spanning nine datasets. For sources sitting in lower ranked positions, citing external sources improved visibility by as much as 115 percent.

The same paper falsified a tactic: keyword stuffing, the classic manipulation from early SEO, performed below the unoptimized baseline in generative engines. That pairing of validated and falsified tactics is what makes the paper worth reading in full, and our plain language walkthrough is in the GEO research paper, explained. It is also why this page formats every claim the way it does. Statistics with named sources are precisely the content feature the research showed engines reward.

3.5x: the growth in OAI-SearchBot crawl events after GPT-5 launched in August 2025, from Botify's 2026 analysis of roughly 7 billion OpenAI bot log events across enterprise sites, spanning November 2024 through March 2026. GPTBot, the training crawler, grew 2.9x over the same period. Growth was uneven by sector: healthcare sites saw about 740 percent more OAI-SearchBot activity and media sites about 702 percent, while travel saw roughly 30 percent. Scale still matters: in Botify's most recent 30-day window, Googlebot logged 18.2 billion events against 887 million for all OpenAI bots combined, so OpenAI crawls at roughly 4 percent of Google's volume.

The exchange rate between crawling and traffic is the tension to watch. Cloudflare Radar data from 2026 put Anthropic's crawl-to-refer ratio around 10,300 pages crawled per referral sent and OpenAI's GPTBot near 900 to 1, against single digits for Googlebot. The ratios move with product changes: after Claude added web search with clickable citations, Anthropic's ratio dropped 86.7 percent in six months. Who these bots are and how to handle them is mapped in our AI crawlers guide.

$1.01 billion: the global GEO services market in 2025 per Intel Market Research, projected at $1.48 billion for 2026 and $17.02 billion by 2034, a 45.5 percent compound annual growth rate. Valuates Reports runs a more conservative model and still lands at $7.3 billion by 2031 on a 34 percent CAGR. The forecasts differ because the market definitions differ, which is worth stating whenever you quote either one, but every published model agrees on the direction and the steepness.

Set the market growth against McKinsey's finding that only 16 percent of brands systematically track AI search performance and you get the shape of the next few years: budgets arriving faster than measurement discipline. The brands that win that period will be the ones whose numbers survive an audit.

Three habits keep a stats deck honest. First, quote the window with the number: Ahrefs' 34.5 percent and 58 percent are both real, eight months apart. Second, separate predictions from measurements, as with Gartner's 25 percent forecast. Third, respect the variance research: since identical prompts agree less than 1 percent of the time, any claim about what an engine says of a brand needs repeated runs behind it.

That third habit is exactly how Reachroller measures. It runs your buying questions across ChatGPT, Claude, Gemini, Perplexity and Grok through official APIs, scores literal mentions over repeated runs, and stores every raw answer so the score is auditable. The full method, including what gets excluded from the headline number, is public on the methodology page. If you want your own row in a table like the one above, the free checker on the Reachroller homepage produces a first read on your brand in about a minute, and it is the direct place to start.