SolCrys Logo

AEO Fundamentals

AI search in 2026: a dated timeline and the statistics that hold up

A dated, sourced reference for AI search in 2026: 28 statistics on usage, B2B buyers, clicks, accuracy, and variance; 24 events from September 2025 to September 2026; and 18 widely repeated numbers to stop quoting, with what to cite instead. Every figure carries its number, date, source, and sample, and each was checked against its source on Sept. 26, 2026. The pattern: use of AI answers kept growing, clicks to websites kept falling, and the numbers that travel furthest are often the least scoped.

Updated

Questions this guide answers

  • What are the most reliable AI search statistics for 2026?
  • How many people use ChatGPT, Google AI Mode, and AI Overviews in 2026?
  • How much do Google AI Overviews reduce clicks to websites?
  • What share of B2B buyers use AI to research vendors?
  • Does AI search traffic convert better than organic search?
  • Which AI search statistics are wrong or outdated?
  • What changed in AI search between September 2025 and September 2026?

Direct answer

As of Sept. 26, 2026, the AI search numbers that hold up are dated and scoped: ChatGPT has more than 1 billion weekly active users (OpenAI, Aug. 31, 2026), Google reports more than 2.5 billion monthly users for AI Overviews and more than 1 billion monthly active users for AI Mode (Google, May 19, 2026), and 49% of U.S. adults use AI chatbots (Pew, June 17, 2026).

What fell was the click, not the question: Alphabet said in April that Search queries were at an all-time high, while Chartbeat data show Google Search referrals to publishers down 40.2% in the year to July 2026. B2B buyers moved too, but less completely than the headlines suggest: 94% use AI somewhere in buying (Forrester), 45% used generative AI mainly to research vendors (Gartner), and review sites still narrowly beat chatbots as the top shortlist influence (G2).

Below are 28 statistics, each with its number, date, source, and sample; 24 dated events from September 2025 to September 2026; and 18 widely repeated numbers to stop quoting, with what to cite instead.

How this page chooses its numbers

Every figure here passed four tests. It comes from the organization that produced it, or, where the original is gated, from a named outlet that reported it (marked "via"). It carries the date of the data or the statement. Its sample and scope are stated. And we checked it against its source on Sept. 26, 2026. Usage counts from OpenAI and Google are labeled company-reported, survey results are what respondents said, and arXiv papers not yet peer reviewed are marked as preprints.

We left out statistics from AEO and SEO tool vendors' own studies, including SolCrys's; our own citation study appears only as a qualitative, engine-by-engine cross-check. Vendors publish useful research, but they sell the thing their research measures, and several of the most repeated vendor figures don't survive a look at the method (see the stop-quoting table). Three figures below touch a vendor's data, and each is labeled: Adobe's retail figures come from Adobe Analytics, and Adobe also sells AEO products; SparkToro's zero-click analysis uses Similarweb's panel; and the Reddit citation figures Search Engine Journal reported came from a GEO tracking tool. We keep them because each rests on a large measured sample or was published and scoped by a neutral party.

This page is the sourced companion to Eason Wang's essay A Year in View: The Models Kept Changing. The Facts Had to Hold., which tells the same year as a story.

Usage and scale: how many people use AI search

The largest AI answer surfaces now report users in the billions. Each company counts a user its own way, so compare a figure with its own history, not with another company's.

StatisticNumberDateSourceSample and scope
ChatGPT weekly active usersMore than 1 billionAug. 31, 2026OpenAICompany-reported, ChatGPT only. Up from 800 million (Oct. 6, 2025) and 900 million (Feb. 27, 2026).
Google AI Overviews monthly usersMore than 2.5 billionMay 19, 2026Google I/O 2026Company-reported. Up from 2 billion monthly users (Nov. 18, 2025). Google's announcement doesn't define a user.
Google AI Mode monthly active usersMore than 1 billionMay 19, 2026; repeated July 22, 2026Google I/O 2026; Alphabet Q2 2026Company-reported, monthly. The 75 million reported on Oct. 29, 2025 was daily users, a different unit.
Gemini app monthly active users950 millionJuly 22, 2026Alphabet Q2 2026Company-reported. Up from 650 million (Oct. 29, 2025) and more than 900 million at I/O (May 19, 2026).
ChatGPT Ads annualized revenue run rate$1 billion, in under 200 daysAug. 31, 2026OpenAICompany-reported. Ads are labeled and shown separately from answers; OpenAI says they are available in more than 40 countries.
U.S. adults who use AI chatbots49%June 17, 2026Pew Research Center5,119 U.S. adults surveyed Feb. 17–23, 2026. Up from 33% in 2024, though Pew notes the 2024 question was worded differently and asked only of those who had heard of chatbots; about 4 in 10 U.S. adults use chatbots for information searching.
People who use AI chatbots for news each week10% (16% of under-35s)June 16, 2026Reuters Institute Digital News Report 2026About 2,000 respondents in each of 48 markets. Up from 7% in 2025; 42% of those who use AI chatbots for news say they always or often click through to original news sources (self-reported, in the 27 markets where the question was asked); measured across all respondents in those markets, the figure is 4%, against 19% for search and 17% for social media.
Trending queries that showed an AI Overview13.7% (64.7% of question-form queries)May 13, 2026Xu, Iqbal, and Montgomery (arXiv)55,393 trending queries across 19 categories, March 13–April 21, 2026. Preprint.

Reading the usage numbers

Units decide the story. ChatGPT reports weekly users, Google reports monthly users, and AI Mode's October 2025 figure was daily. A monthly count is always at least as large as a weekly or daily count of the same audience, so the rise from 75 million to 1 billion for AI Mode is partly a change of unit.

AI Overviews appear inside ordinary Google results, so their user count is best read as exposure to AI answers, not a choice to use an AI tool. The independent surveys put the consumer figures in proportion: about half of U.S. adults use chatbots (Pew), and 10% of people across 48 markets use them for news each week (Reuters Institute).

B2B buyers: how much of the buying process runs through AI

Nearly every B2B buyer now uses AI somewhere in the process; far fewer rely on it alone. The broad and the narrow figures measure different things, and both are true.

StatisticNumberDateSourceSample and scope
Business buyers who use AI during their buying process94%Jan. 21, 2026Forrester, State of Business Buying 2026Nearly 18,000 global business buyers in Forrester's Buyers' Journey Survey. Counts any AI use during the buying process.
B2B buyers who used generative AI, primarily to gather vendor and product information45%May 20, 2026Gartner645 B2B buyers surveyed August–September 2025.
B2B buyers who prefer to validate AI-generated insights with sales reps69%May 20, 2026GartnerSame survey: 645 B2B buyers, August–September 2025.
B2B software buyers who start research in an AI chatbot more often than in Google51% (29% in April 2025)April 15, 2026G2, via PR Newswire1,076 B2B software buyers in North America, EMEA, and APAC, March 2026. G2 runs a software review marketplace.
Buyers who chose a different vendor than planned because of chatbot guidance69%April 15, 2026G2, via PR NewswireSame survey: 1,076 B2B software buyers, March 2026.
Top source shaping which vendors make the shortlistReview sites 38%, AI chatbots 37%July 22, 2026G2 2026 Buyer Behavior ReportMore than 1,000 B2B software buyers and decision-makers, plus interviews with more than 50 sales and marketing leaders. G2 runs a review marketplace.
Buyers comfortable letting an AI agent execute purchases within approved guardrails9% (61% use or plan to use agents in buying)July 22, 2026G2 2026 Buyer Behavior ReportSame survey: more than 1,000 B2B software buyers and decision-makers.

Reading the buyer numbers

Forrester's 94% counts any AI use during buying. Gartner's 45% asks whether buyers used generative AI mainly to gather vendor and product information, in a smaller sample fielded in August and September 2025. When a claim is about vendor research specifically, the narrower figure is the one that fits.

G2 surveys software buyers and runs a software review marketplace. Its April and July reports point in different directions on the same question, which is a reason to cite them together, not to drop either. The verification step hasn't gone away: 69% of buyers prefer to check AI output with a sales rep, and only 9% would let an agent complete a purchase. For what this means for a software company's prompt set and sources, see AEO for B2B SaaS companies.

Clicks, traffic, and conversion

Queries to Google kept growing while clicks to websites fell, and AI referrals stayed a small share of visits. Most measured click data comes from news publishers, so scope matters more in this table than anywhere else on the page.

StatisticNumberDateSourceSample and scope
Google users who clicked a result, with vs. without an AI summary8% vs. 15% of visits; 1% clicked a link inside the summaryJuly 22, 2025 (baseline)Pew Research Center900 U.S. adults, 68,879 Google searches, March 2025. 18% of the searches showed an AI summary.
U.S. Google searches that ended without a click68.01%June 9, 2026SparkToroSimilarweb desktop and mobile web panel, U.S., January–April 2026. Excludes the Google app. The 60.45% 2024 comparison came from a different panel.
Google organic referrals to news sitesDown 33% globally, 38% in the U.S.Jan. 12, 2026Reuters Institute, Trends and Predictions 2026Chartbeat data on 2,576 news sites, November 2024 to November 2025.
Google Search referrals to publishers, year over yearDown 40.2%Sept. 24, 2026Chartbeat, as reported by Search Engine JournalChartbeat's publisher network, July 2025 to July 2026, after a 21.9% decline the year before. Google's share of pageviews fell from about 9% (July 2024) to 5%. The report gives no site count.
AI referrals' share of publisher traffic0.01% (up 8x over 2 years)Sept. 24, 2026Chartbeat, as reported by Search Engine JournalSame Chartbeat publisher network and report. SEJ notes the 0.01% is hard to reconcile with other estimates that put chatbot referrals under 1% of traffic, and the underlying Chartbeat report is gated.
AI referrals' share of total visitsUnder 1% (AI referrals up 155% over 8 months)Nov. 20, 2025Microsoft Bing, on Microsoft Clarity dataMore than 1,200 publisher and news sites; the conversion comparison used 1 month of data. Clarity also found AI referrals converted at up to 3 times the rate of search and social on those sites.
AI-referred traffic to U.S. retail sitesUp 393% year over year in Q1 2026; converted 42% better than non-AI traffic in March 2026, 38% worse in March 2025April 16, 2026Adobe Analytics, via TechCrunchMore than 1 trillion visits to U.S. retail sites. Retail only. Adobe also sells AEO products.

Reading the traffic numbers

Clicks fell, not queries. Alphabet reported Search queries "at an all-time high" on April 29, 2026, and the decline shows up in referrals instead: Pew's 2025 baseline found a result was clicked on 8% of visits with an AI summary against 15% without one, and publisher data for 2025 and 2026 show double-digit drops.

Publisher data is not your data. The Chartbeat and Clarity samples are news and publisher sites. A B2B software site with branded demand can follow a different curve, so check your own Search Console before borrowing a publisher number; this diagnostic shows how.

Conversion claims flip. In Adobe's retail data, AI-referred shoppers converted 38% worse than other traffic in March 2025 and 42% better in March 2026. A conversion multiplier for AI traffic without a vertical and a date is not a finding; the dependable number is your own, and closed-won interviews catch the influence analytics can't see.

Accuracy and variance

AI answers are neither reliably accurate nor stable from run to run, and the research on what moves them is thinner than the advice built on it.

StatisticNumberDateSourceSample and scope
AI Overview claims not supported by the pages they cite11.0% of 98,020 claimsMay 13, 2026Xu, Iqbal, and Montgomery (arXiv)Google AI Overviews only; 55,393 trending queries, March 13–April 21, 2026. Omission was the most common failure. Preprint.
B2B software buyers who often or very often get inaccurate AI chatbot recommendations64%April 15, 2026G2, via PR Newswire1,076 B2B software buyers, March 2026. Self-reported.
Chance that 2 runs of the same prompt return the same brand listLess than 1 in 100 (same order: fewer than 1 in 1,000)Jan. 28, 2026SparkToro600 volunteers, 2,961 runs of 12 prompts on ChatGPT, Claude, and Google AI Overviews (AI Mode when no Overview showed), November–December 2025.
Overlap of retrieved sources across Google Search, Gemini, and AI OverviewsAverage Jaccard similarity below 0.2April 30, 2026Grossman et al., SIGIR 2026 (arXiv)11,500 real user queries, released by the authors as a public benchmark. Peer reviewed (SIGIR 2026).
GEO studies showing a stable, longitudinal, cross-platform causal effectNone of 45July 15, 2026Martinez, critical survey of GEO (arXiv)Review of 45 generative engine optimization studies, 2023–2026. Preprint.
Reddit's share of ChatGPT search citations3.83% (July 18–Aug. 7) to 0.52% (Aug. 14–17, 2026)Aug. 19, 2026Search Engine JournalData from a GEO tracking tool, as reported by SEJ. ChatGPT only, over a short window; the tracker called the size of the drop provisional, and OpenAI did not respond to a request for comment (per Gizmodo, cited by SEJ).

Reading the accuracy numbers

Every accuracy figure is engine- and topic-specific. The 11.0% describes Google AI Overviews on trending queries, not ChatGPT or Perplexity, and not questions about one brand. The 64% is buyers' perception. Neither is an error rate for your company; that takes your own prompts, run repeatedly and graded against facts you can document (see whether AI is telling the truth about your brand).

Variance is a design constraint, not a reason to give up. Identical lists are rare, but rates across many runs are stable enough to track, and how many runs a trustworthy number takes depends on the size of the difference you need to detect.

Sources move for reasons outside your content. ChatGPT's Reddit citations fell sharply in September 2025 and again in August 2026, on that one engine, and OpenAI did not comment on the August drop. A source's AI visibility can change with retrieval or licensing decisions you never see. And no technique has yet shown a stable causal effect across platforms, so treat any "X raises AI visibility by N%" claim as a hypothesis to test on your own prompts.

Timeline: September 2025 to September 2026

These 24 dated events changed how AI answers are delivered, paid for, measured, or regulated. Model releases are left out unless they changed a search surface.

DateWhat happenedSource
Sept. 29, 2025OpenAI launches Instant Checkout: U.S. ChatGPT users can buy from Etsy sellers in the chat, through the Agentic Commerce Protocol built with Stripe.OpenAI
Oct. 21, 2025OpenAI launches the ChatGPT Atlas browser, on macOS only.OpenAI
Oct. 29, 2025Alphabet reports more than 75 million daily active users for AI Mode and more than 650 million monthly users for the Gemini app.Alphabet Q3 2025
Nov. 18, 2025Gemini 3 ships in AI Mode on launch day for Google AI Pro and Ultra subscribers, which Google calls "the first time we are shipping Gemini in Search on day one." Google puts AI Overviews at 2 billion monthly users.Google
Dec. 9, 2025The European Commission opens a formal antitrust investigation into Google's use of publishers' content for AI Overviews and AI Mode.European Commission
Jan. 16, 2026OpenAI says it will start testing ads in ChatGPT in the U.S., on the Free and Go tiers.OpenAI
Feb. 4, 2026Anthropic says "Claude will remain ad-free."Anthropic
Feb. 9, 2026The ChatGPT ads test begins in the U.S. for logged-in adults on Free and Go. Plus, Pro, Business, Enterprise, and Education tiers carry no ads.OpenAI
Feb. 10, 2026Bing Webmaster Tools adds an AI Performance report in public preview: citations and grounding queries across Copilot and Bing's AI summaries. It reports no clicks.Microsoft Bing
March 9, 2026Gartner publishes a Market Guide for Answer Engine Visibility Tools.Gartner
March 24, 2026OpenAI lets merchants use their own checkout and refocuses ChatGPT shopping on product discovery, saying the first Instant Checkout "did not offer the level of flexibility that we aspire to provide."OpenAI
April 29, 2026Alphabet says Google Search "queries are at an all-time high."Alphabet Q1 2026
May 13, 2026Google Analytics 4 adds an AI Assistant default channel for traffic from ChatGPT, Gemini, Copilot, and similar assistants. It excludes Google's AI Overviews and AI Mode.Google Analytics; channel definition
May 15, 2026Google publishes its guide to generative AI features in Search: "optimizing for generative AI search is optimizing for the search experience, and thus still SEO." No special files, schema, or chunking are required (updated July 10). The same day, its spam policies extend to generative AI responses, naming "attempting to manipulate generative AI responses in Google Search."Google Search Central; spam policies
May 19, 2026At I/O, Google reports more than 2.5 billion monthly users for AI Overviews, more than 1 billion monthly active users for AI Mode, and more than 900 million for the Gemini app.Google
May 28, 2026The Regional Court of Munich rules that AI Overviews are Google's own content and that Google is liable for false statements in them. The ruling is not final.The Decoder
June 3, 2026The UK Competition and Markets Authority requires Google to let publishers opt out of their content powering AI features and to attribute it with clear links. Google has 9 months to implement.UK CMA
July 1, 2026Cloudflare gives every customer separate Search, Agent, and Training controls for AI traffic.Cloudflare
July 22, 2026Alphabet says AI Mode has "surpassed 1 billion monthly active users" and the Gemini app has 950 million.Alphabet Q2 2026
Aug. 8–17, 2026ChatGPT all but stops citing Reddit: 3.83% of its citations before the break, 0.52% after, per tracker data. In SolCrys's cross-engine data, Google AI Overviews, Perplexity, and Gemini did not follow.Search Engine Journal; SolCrys citation study
Aug. 9, 2026ChatGPT Atlas stops working. Its agentic browsing moves into the ChatGPT desktop app and a Chrome extension.OpenAI Help Center
Aug. 31, 2026OpenAI says ChatGPT Ads reached a $1 billion annualized revenue run rate in under 200 days, and that ChatGPT has more than 1 billion weekly active users.OpenAI
Aug. 31, 2026Search Console's generative AI performance report reaches all websites worldwide. It shows impressions from AI Overviews and AI Mode, with no clicks, click-through rate, or queries.Google Search Console Help
Sept. 15, 2026Cloudflare adds a Disallow AI Training setting that keeps Googlebot, Bingbot, and Applebot crawling for search, while an explicit Block for Training now stops them too. Existing settings migrate to Disallow AI Training, and new ad-monetized domains are offered Search allowed, Training set to Disallow AI Training, and Agent blocked on pages with ads.Cloudflare

What the timeline shows

Three patterns run through the year. Answers started carrying ads: ChatGPT's ads went from a U.S. test in February to a reported $1 billion run rate by Aug. 31, while Anthropic said Claude will stay ad-free. Buying inside the chat retreated while discovery grew: Instant Checkout was phased out and Atlas shut down, while OpenAI refocused ChatGPT shopping on product discovery. And free platform reports arrived that measure presence, not outcomes: Bing reports citations without clicks, GA4's AI channel leaves out Google's own AI surfaces, and Search Console reports AI impressions only.

AI search statistics to stop quoting

These numbers circulate widely. Some were predictions, some described one website or one content type, and some mix units. Where a number originated with an AEO or SEO tool vendor, we describe it instead of linking it, in line with how the rest of this page is sourced.

Claim as usually quotedWhat's wrongWhat to cite instead
"Search volume will drop 25% by 2026 because of AI chatbots" (Gartner)A February 2024 prediction about query volume, not a measurement. Alphabet said on April 29, 2026, that Search queries were at an all-time high. Clicks to websites fell; queries did not.Measured referral declines with their scope: Reuters Institute and Chartbeat (news sites, down 33% globally, November 2024 to November 2025) and Chartbeat via SEJ (publishers, down 40.2%, July 2025 to July 2026).
"AI search visitors are worth 4.4x organic visitors"From a widely shared 2025 SEO-tool study built on digital-marketing and SEO topics. It states the multiplier without the sites, conversions, or period behind it, next to projections the vendor calls extrapolations.Measured conversion with a vertical and a date: Adobe retail data via TechCrunch (42% better in March 2026, 38% worse in March 2025). Adobe also sells AEO products.
"AI visitors convert 23x better than organic"From one SEO-tool company's analytics on its own website over 30 days: 0.5% of its traffic produced 12.1% of its signups. One site is not a sample.Your own data: GA4 AI Assistant channel conversions, remembering that it excludes AI Overviews and AI Mode.
"AI Overviews cut clicks by 34.5%"An April 2025 estimate from an SEO-tool vendor, which the same vendor revised upward in February 2026. The revision is desktop-only and compares against a modeled counterfactual.Pew's measured behavior: a result was clicked on 8% of visits with an AI summary vs. 15% without (900 U.S. adults, March 2025).
"Only 12% of AI citations rank in Google's top 10, so SEO doesn't matter"The 2025 SEO-tool study behind it compared ChatGPT, Gemini, Copilot, and Perplexity citations with Google's top 10 for the original prompt, and the 12% averages engines that differ widely. It did not cover Google AI Overviews, the surface built on Google's own ranking systems.Google: optimizing for its generative AI features is "still SEO." Xu et al.: nearly 30% of AI Overview–cited domains were not in the co-displayed first-page results, so most were.
"Zero-click searches rose from 56% to 69%"The figure covers news-related searches only, May 2024 to May 2025.SparkToro: 68% of all U.S. Google searches were zero-click in January–April 2026, on a web panel that excludes the Google app.
"94% of B2B buyers use AI to choose vendors"Forrester's 94% counts any AI use during the buying process, not AI as the deciding source.Pair it with Gartner: 45% used generative AI primarily for vendor and product information, and 69% prefer to validate AI output with sales reps.
"Half of B2B buyers now start with AI instead of Google"Accurate for G2's March 2026 sample (51%). G2's July report found review sites (38%) narrowly ahead of chatbots (37%) as the top shortlist influence, and G2 runs a review marketplace.Both reports together: April 2026 and July 2026.
"AI is wrong 45% of the time"The EBU and BBC study tested news questions on free or consumer versions of ChatGPT, Copilot, Gemini, and Perplexity from May 24 to June 10, 2025. "Significant issue" includes sourcing problems; about a fifth of answers had major accuracy issues. It is not an error rate for brand facts.For claims in AI Overviews, Xu et al. (11.0% unsupported). For buyer experience, G2 (64% report frequent inaccuracy). Name the engine either way.
"AI visibility can't be measured, because every answer is different"The SparkToro study behind it measured identical lists, and concluded that "visibility % across dozens to hundreds of prompts run multiple times is a reasonable metric."Rates across repeated runs. Treat a rank in a single answer as noise.
"GEO boosts visibility by up to 40%"A 2023 lab result on a benchmark with a fixed set of sources per query.The Martinez survey: none of 45 studies showed a stable, longitudinal, cross-platform causal effect.
"Google says AI clicks are higher quality"Google's Aug. 6, 2025 post said it is sending "slightly more quality clicks" but published no numbers.Pew's measured click behavior, and your own analytics.
"Search Console now shows your AI Overview and AI Mode clicks"The generative AI performance report shows impressions only: no clicks, click-through rate, or queries.Google Search Console Help
"AI Mode grew from 75 million to 1 billion users"The 75 million figure was daily active users (October 2025); the 1 billion figure is monthly active users (May and July 2026).Each figure with its unit: Alphabet Q3 2025 and Alphabet Q2 2026.
"ChatGPT hit 1 billion weekly users on July 31"OpenAI's July 31 statement said its "models now reach more than one billion active users," a count across its products, as reported by BNN Bloomberg.OpenAI's Aug. 31, 2026 statement: more than 1 billion weekly active users of ChatGPT.
"Ads now run inside AI Overviews worldwide"Ads inside AI Overviews run in 12 countries, in English only.Google Ads Help
"ChatGPT's Instant Checkout launched in February 2026"It launched Sept. 29, 2025, and OpenAI scaled it back on March 24, 2026.OpenAI: launch and retreat.
"Reddit is dead in AI search"The August 2026 break was on ChatGPT. In SolCrys's cross-engine data, Google AI Overviews, Perplexity, and Gemini did not follow.Search Engine Journal for ChatGPT, and the SolCrys citation study for the engine-by-engine picture.

How to cite an AI search statistic responsibly

Four fields make a statistic citable: source, date, sample, and scope. Three more keep it honest: unit, interest, and kind. These are the same rules in our published editorial standards.

  • Source: the originator, not the aggregator. Link the study or statement that produced the number. Statistics roundups drop the scope, which is how a news-only zero-click figure becomes a claim about all searches.
  • Date: when the data was collected, as well as when it was published. Pew's click study was fielded in March 2025 and published in July 2025. A 2024 prediction is not a 2026 fact.
  • Sample: who, and how many. A survey of 645 buyers and one of 18,000 can both be right about different questions. One company's website is not a sample.
  • Scope: engine, surface, geography, device, and vertical. An AI Overviews finding is not a ChatGPT finding, a U.S. retail conversion rate is not a B2B software rate, and a web panel that excludes the Google app misses much of mobile search.
  • Unit: daily, weekly, or monthly; share of answers or share of citations; relative or absolute change. "Down 86%" and "down 3.3 points" can describe the same drop.
  • Interest: who benefits. Platforms report their own usage, review sites survey buyers about review sites, and tool vendors publish studies that sell tools. Disclose it; it doesn't make a number wrong.
  • Kind: measurement, self-report, or prediction. Say which. A survey answer describes what people say they do; a panel describes what a sample did; a forecast describes neither.

The same fact, cited two ways

Weak: "AI Overviews have 2.5 billion users." Citable: "Google said on May 19, 2026, that AI Overviews had more than 2.5 billion monthly users (company-reported)."

Weak: "AI traffic converts 4x better." Citable: "In Adobe Analytics data on U.S. retail sites, AI-referred traffic converted 42% better than other traffic in March 2026, after converting 38% worse a year earlier (TechCrunch, April 16, 2026)."

How SolCrys applies these rules

SolCrys is an AEO platform built around Measure → Diagnose → Execute → Verify, and the discipline on this page shapes both what it measures and what it declines to claim.

  • Engine by engine. SolCrys runs a fixed set of buyer prompts on ChatGPT, Google AI Overviews, Gemini, Perplexity, and Claude, and keeps each engine's results separate. That is how the August 2026 Reddit break showed up as one engine's change in our 36,268-citation study of the AEO category, which states its window, prompt count, engines, and method.
  • Rates, not ranks. In line with the variance research above, SolCrys's core metrics are rates across repeated runs, not a position in a single answer.
  • Observations, not proof. After a team ships a change, Verify re-runs the same frozen prompt set on the same engines and, for a page fixed from a Content Audit, re-audits the page. A before-and-after number is reported as an observation, not as proof of cause, which is the bar the 45-study survey sets.
  • Accuracy with evidence. Answer Accuracy grades AI answers against your organization's Corporate Context and cannot mark a claim wrong without pointing to the evidence.

What SolCrys doesn't measure

SolCrys measures answers, not visits. It does not report clicks, AI referral traffic, or conversions; GA4's AI Assistant channel, Search Console, and Bing Webmaster Tools each cover part of that, with the gaps listed in the timeline and set out report by report in what the free AI reports show. Its Google measurement is AI Overviews: it does not capture AI Mode answers, and it does not measure Microsoft Copilot. And it publishes no usage or market-size estimates of its own, which is why none appear on this page.

To see what ChatGPT says about your brand, Start Free: the free workspace includes 10 ChatGPT prompts and 1 content audit, with email verification and no credit card. Paid plans add more engines; to choose a plan, talk to us.

Last updated and changelog

Last updated Sept. 26, 2026. We will review this page every quarter, re-check each number on its source page, and log every change below with its date. The next review is due in December 2026.

  • Sept. 26, 2026: First edition. 28 statistics, 24 dated events, and 18 numbers to stop quoting, each checked against its source on Sept. 26, 2026.

Sources

FAQ

How many people use ChatGPT, Google AI Mode, and AI Overviews in 2026?

ChatGPT has more than 1 billion weekly active users (OpenAI, Aug. 31, 2026). Google reports more than 2.5 billion monthly users for AI Overviews and more than 1 billion monthly active users for AI Mode (May 19, 2026, repeated for AI Mode on July 22), and 950 million monthly users for the Gemini app (July 22, 2026). All are company-reported, and the units differ: weekly for ChatGPT, monthly for Google.

How much do Google AI Overviews reduce clicks?

The clearest neutral measurement is Pew's: in March 2025, people clicked a traditional result on 8% of visits when an AI summary appeared and 15% when none did, and clicked a link inside the summary on 1%. Publisher data since then show larger falls in Google referrals overall, including a 40.2% year-over-year decline to July 2026 in Chartbeat's network, but those figures cover more than AI Overviews and come from news sites.

What share of B2B buyers use AI to research vendors?

It depends on the question asked. Forrester found 94% of business buyers use AI somewhere in the buying process. Gartner found 45% used generative AI primarily to gather vendor and product information, and 69% prefer to validate AI output with sales reps. G2 found 51% of software buyers start research in a chatbot more often than in Google (April 2026), and that review sites (38%) narrowly edged chatbots (37%) as the top shortlist influence (July 2026).

Does AI search traffic convert better than organic search?

Sometimes, and not consistently. In Adobe's U.S. retail data, AI-referred traffic converted 42% better than other traffic in March 2026 but 38% worse in March 2025. Microsoft Clarity found AI referrals converted at up to 3 times the rate of search and social on more than 1,200 publisher and news sites, in 1 month of data, while staying under 1% of visits. The widely shared 4.4x and 23x multipliers come from a vendor study that doesn't disclose its data and from one company's own website. Measure your own with GA4's AI Assistant channel, which excludes Google's AI Overviews and AI Mode.

Which AI search statistics are wrong or outdated?

The most common: Gartner's 2024 prediction of a 25% drop in search volume by 2026 (queries grew; clicks fell), "AI Mode grew from 75 million to 1 billion users" (daily vs. monthly), "ChatGPT hit 1 billion weekly users on July 31" (that was a cross-product count; the weekly figure dates from Aug. 31), "AI is wrong 45% of the time" (news questions in mid-2025), and "Search Console shows AI clicks" (impressions only). The full table, with what to cite instead, is on this page.

How accurate are AI search answers?

On Google AI Overviews, a 2026 study found that 11.0% of 98,020 claims were not supported by the pages cited for them, with omission the most common failure (preprint). Separately, 64% of B2B software buyers told G2 they often or very often get inaccurate AI chatbot recommendations. Neither is an error rate for a specific brand; that takes repeated runs of your own prompts, graded against facts you can document.

Can AI visibility be measured if answers change every time?

Yes, as a rate. SparkToro found less than a 1 in 100 chance that 2 runs return the same brand list, but concluded that visibility percentage across many prompts run multiple times is a reasonable metric. A rank position in a single answer is not.

How often is this page updated?

Quarterly. Each review re-checks every number on its source page, and each change is logged with its date in the changelog at the end of the page. It was last updated Sept. 26, 2026.

Free ChatGPT visibility check

See where AI answers skip your brand — then fix it, free

Start a free workspace with your domain: 10 buyer-intent prompts through ChatGPT show where you are mentioned, cited, or skipped, and who gets recommended instead. A free content audit in the same workspace hands you the first fix to ship.

Start Free

Free · No credit card · About 5 minutes

Related guides

Strategy & Positioning

AI Search Is a Decision Layer

Treating AI search as 'another channel' alongside SEO and paid undercounts its impact. AI search is a decision layer that influences buyers across all other channels. This essay explains the implication for measurement and investment.

How SolCrys Works

SolCrys Editorial Standards

SolCrys publishes its editorial standards in full: the 5 mandatory DO's, the 7 named anti-patterns we refuse to recommend, and the pre-publication checklist every SolCrys asset runs through. We hold our content to a higher bar than the AEO category average — and we want buyers to be able to check.