AI search statistics.
The numbers, in context.
AI use is huge, and the traffic it sends is still small. These figures show both, each with its source, its date, how it was measured and what it does not tell you.
Looking for the takeaway? Read what these AI search trends mean for your site.
The short version
- ChatGPT reports more than 1 billion weekly users. OpenAI · Aug 2026
- 68% of US Google searches ended without a click in early 2026. SparkToro · 2026
- AI Overviews showed on 73% of US branded keywords in September. Ahrefs · Sep 2026
- AI chatbots sent 0.28% of website visits. Google sent 28%. Ahrefs · Mar 2026
- Shoppers sent by AI converted 60% better than other US retail visitors. Adobe · Jul 2026
How many people use AI search?
Four AI products now report a billion or more users each. Those totals count people, not searches, and the audiences overlap.
- Nov 2023: 100M weekly users Nov 2023100M weekly usersSam Altman at OpenAI DevDay
- Aug 2024: 200M weekly users Aug 2024200M weekly usersOpenAI, reported by Engadget
- Feb 2025: 400M weekly users Feb 2025400M weekly usersOpenAI COO Brad Lightcap
- Mar 2025: 500M weekly users Mar 2025500M weekly usersOpenAI, reported by TechCrunch
- Oct 2025: 800M weekly users Oct 2025800M weekly usersSam Altman at OpenAI DevDay
- Feb 2026: 900M weekly users Feb 2026900M weekly usersOpenAI, reported by TechCrunch
- Aug 2026: 1B+ weekly users Aug 20261B+ weekly usersOpenAI
Each dot is a date OpenAI announced a figure and links to the report. The line between dots is drawn, not measured.
- ChatGPT 76% to 53%
- Gemini <9% to 27–28%
- Claude ~2% to ~9%
Website visits only. Similarweb gives Gemini and Claude as approximate values, so their dots sit at the middle of the stated range.
-
1B+
ChatGPT weekly active users, worldwide.
- How it was measured
- OpenAI’s own count of people who used ChatGPT in a week. It includes writing, coding and every other use alongside search.
- What it does not tell you
- A count of people, not searches. It says nothing about how often anyone asks about your category.
- Source
- OpenAI: A milestone in expanding access to AI, August 31, 2026. Company-reported weekly active users, worldwide. Counts every use of ChatGPT, not only search.
- Evidence
- Company claim. A platform reporting its own numbers. Not independently audited.
-
1B+
Gemini app monthly active users, worldwide.
- How it was measured
- Google’s count of people who used the Gemini app in a month. The app passed 950 million in July and 750 million in February.
- What it does not tell you
- Monthly, not weekly, so it cannot be set against ChatGPT’s weekly figure.
- Source
- Google: The Gemini app passes one billion monthly users, August 11, 2026. Company-reported monthly users of the Gemini app, worldwide.
- Evidence
- Company claim. A platform reporting its own numbers. Not independently audited.
-
1B+
AI Mode monthly active users, worldwide.
- How it was measured
- Google’s figure for its conversational search tab, which expanded worldwide in October 2025.
- What it does not tell you
- AI Mode and AI Overviews are different features. This is not the AI Overviews audience.
- Source
- Alphabet: Q2 2026 earnings call, CEO remarks, July 22, 2026. Company-reported monthly active users of AI Mode, worldwide.
- Evidence
- Company claim. A platform reporting its own numbers. Not independently audited.
-
2B+
AI Overviews monthly users.
- How it was measured
- Google’s figure for monthly users of AI Overviews, across more than 200 countries and territories and 40 languages.
- What it does not tell you
- AI Overviews appear inside ordinary searches, so a user here may never have chosen to use AI. It is also the latest figure Google has given, from July 2025.
- Source
- Alphabet: Q2 2025 earnings call, CEO remarks, July 23, 2025. Company-reported monthly users of AI Overviews across more than 200 countries and territories.
- Evidence
- Company claim. A platform reporting its own numbers. Not independently audited.
-
10%from 7%
of people used an AI chatbot for news in the past week, up from 7% a year earlier.
- How it was measured
- Reuters Institute online survey, about 2,000 people per market. Among under-35s the figure was 16%.
- What it does not tell you
- News only, not all information seeking. The US, UK, France and Germany saw no increase over the year.
- Source
- Reuters Institute: Digital News Report 2026, June 16, 2026. Online survey of about 2,000 people in each of 48 markets.
- Evidence
- Survey. What people say they do. Strong on intent, weaker on behavior.
These audiences overlap. Adding them would count some people twice and mix weekly with monthly activity. None of them tells you how often buyers ask about your category.
How often does a search send a click?
Most Google searches already end without a click. When an AI summary appears, people click even less.
Two studies, two methods. SparkToro counts any click from 100 searches. Pew counts clicks on ordinary results per 100 visits to a results page. Compare within each side, not across them.
-
68.01%
of US Google searches ended without a click from January to April 2026.
- How it was measured
- SparkToro’s analysis of Similarweb’s desktop and mobile web panel. SparkToro’s 2024 study put the rate at 60.45%, using a different panel from Datos.
- What it does not tell you
- Counts every search with no click, whatever the reason. It is not a count of clicks taken by AI, and the 2024 and 2026 panels are not a matched pair.
- Source
- SparkToro: In 2026, less than one third of Google searches still send a click, June 9, 2026. Similarweb US desktop and mobile web panel, January to April 2026. Excludes the Google mobile app.
- Evidence
- Traffic data. Visits or clicks measured by a panel, analytics or network logs.
-
8%vs 15%
of visits led to a click on a traditional result when an AI summary appeared. Without one, 15% did.
- How it was measured
- Pew tracked real browsing and reconstructed each results page: 68,879 unique searches, 12,593 of them with an AI summary.
- What it does not tell you
- Observational. Searches that show a summary differ from those that do not, so the gap is a relationship, not proof of cause.
- Source
- Pew Research Center: Google users are less likely to click on links when an AI summary appears, July 22, 2025. Browsing by 900 US adults in March 2025: 68,879 unique searches, 12,593 with an AI summary. Results collected April 7 to 17.
- Evidence
- Research. Nonprofit or academic research with a published method.
-
1%
of visits with an AI summary led to a click on one of the summary’s own sources.
- How it was measured
- Same Pew sample. A click on a link inside the summary, counted separately from clicks on ordinary results.
- What it does not tell you
- One month of US browsing in early 2025. Summaries and their link layouts have changed since.
- Source
- Pew Research Center: Google users are less likely to click on links when an AI summary appears, July 22, 2025. Browsing by 900 US adults in March 2025: 68,879 unique searches, 12,593 with an AI summary. Results collected April 7 to 17.
- Evidence
- Research. Nonprofit or academic research with a published method.
-
26%vs 16%
of visits to a results page with an AI summary ended the browsing session. Without a summary, 16% did.
- How it was measured
- Same Pew sample. The share of results pages after which the person stopped browsing.
- What it does not tell you
- Ending a session can mean the answer was enough or that the person gave up. The data cannot tell which.
- Source
- Pew Research Center: Google users are less likely to click on links when an AI summary appears, July 22, 2025. Browsing by 900 US adults in March 2025: 68,879 unique searches, 12,593 with an AI summary. Results collected April 7 to 17.
- Evidence
- Research. Nonprofit or academic research with a published method.
-
58%
lower clickthrough rate for the top-ranking page when an AI Overview appears.
- How it was measured
- Ahrefs compared 300,000 keywords, half with an AI Overview, using aggregated Search Console data, and modeled what the click rate would have been without one.
- What it does not tell you
- A modeled comparison for position one, not a measured 58% traffic loss for every site.
- Source
- Ahrefs: AI Overview clickthrough study, 2026 update, February 4, 2026. 300,000 keywords, aggregated desktop Search Console data from December 2023 and December 2025.
- Evidence
- Traffic data. Visits or clicks measured by a panel, analytics or network logs.
-
31.53%from 33.75%
zero-click rate on the same keywords after an AI Overview appeared, a small fall from 33.75% before.
- How it was measured
- Semrush tracked keywords with more than 100 monthly searches through 2025 and compared each one before and after it gained an AI Overview.
- What it does not tell you
- One tool’s keyword sample. It cuts against the larger drops above, which is why no single figure should stand for every site.
- Source
- Semrush: Semrush AI Overviews study, December 2025. 10M+ keywords for prevalence, January to November 2025. A separate 200,000+ keyword sample for zero-click change.
- Evidence
- Results study. Search results or AI answers collected at scale, usually by an SEO tool vendor.
To check your own clickthrough rates, try the free Zero-Click Report. Its comparison highlights pages to investigate. It cannot prove an AI Overview caused a particular gap.
How common are AI Overviews?
About one in five search results pages showed an AI Overview in September 2025. On brand-name keywords, Google has since gone much further.
The outlined September 2025 column comes from an earlier Ahrefs crawl with its own sample. The daily rate reached 82.91% on September 29, 2026.
Longer and question-shaped queries were far more likely to get a summary. Shares of results pages, desktop.
-
73.3%from 61.3%
of US branded keywords showed an AI Overview in September 2026, up from 61.3% in July.
- How it was measured
- 92.5 million crawls of branded keywords. The daily rate reached 82.91% on September 29. An earlier Ahrefs crawl, collected in September 2025, measured 13.1%.
- What it does not tell you
- US desktop only, and a share of keywords, not of searches weighted by how often people make them. A branded keyword is any query containing a brand name, not only someone looking for a homepage.
- Source
- Ahrefs: AI Overviews currently appear for 83% of branded searches, October 2, 2026. 92.5 million branded-keyword crawls of US desktop Google results, July to September 2026.
- Evidence
- Results study. Search results or AI answers collected at scale, usually by an SEO tool vendor.
-
+71%
growth in AI Overviews on commercial-intent results pages from November 2025 to April 2026.
- How it was measured
- Semrush tracked 600,000+ keywords in 10 industries. Finance grew most, up 231%. Results pages with both ads and an AI Overview became about twice as common as a year earlier.
- What it does not tell you
- A growth rate in one tool’s sample, not the share of commercial-intent keywords that show an AI Overview.
- Source
- Semrush: AI Overviews are expanding across commercial intent search, July 2, 2026. 600,000+ US keywords in 10 industries, November 2025 to April 2026.
- Evidence
- Results study. Search results or AI answers collected at scale, usually by an SEO tool vendor.
-
15.69%peak 24.61%
of keywords showed an AI Overview in November 2025, after a July peak of 24.61% and 6.49% in January.
- How it was measured
- Semrush checked the same 10 million-plus keyword set each month through 2025.
- What it does not tell you
- Coverage swings month to month, and this sample differs from Ahrefs’. Compare figures within one study, never across them.
- Source
- Semrush: Semrush AI Overviews study, December 2025. 10M+ keywords for prevalence, January to November 2025. A separate 200,000+ keyword sample for zero-click change.
- Evidence
- Results study. Search results or AI answers collected at scale, usually by an SEO tool vendor.
Where do AI citations come from?
Ranking helps, but it is not the whole story. In Ahrefs’ latest study, about four in ten cited pages ranked in the top ten for the same search.
- 37.1%Top 10
- 26.2%Positions 11 to 100
- 36.7%Outside the top 100
Organic positions only. Counting every results block, 37.9% were in the top ten.
-
37.1%
of URLs cited in AI Overviews ranked in Google’s organic top ten for the same query.
- How it was measured
- 863,000 results pages. 26.2% of cited URLs ranked from 11 to 100, and 36.7% ranked outside the top 100. Counting every results block, not only organic links, gives 37.9%.
- What it does not tell you
- Google runs related searches to build an answer, so a page outside the top ten here may rank for another query. Ahrefs’ 2025 study used a different parser, so the two are not a before-and-after.
- Source
- Ahrefs: Where AI Overview citations rank, March 2, 2026. 863,000 search results pages and 4 million cited URLs, January 2026.
- Evidence
- Results study. Search results or AI answers collected at scale, usually by an SEO tool vendor.
-
54.5%from 32.3%
of AI Overview citations also ranked organically in BrightEdge’s sample by September 2025, up from 32.3% in May 2024.
- How it was measured
- BrightEdge tracked citations against organic rankings for the same queries. Healthcare reached 75.3%; restaurants 19.2%.
- What it does not tell you
- A different sample and definition from Ahrefs. The two figures are not points on the same line.
- Source
- BrightEdge: Rank overlap after 16 months of AI Overviews, Data through September 2025. Nine industries monitored from May 2024 to September 2025.
- Evidence
- Results study. Search results or AI answers collected at scale, usually by an SEO tool vendor.
-
13.15%
of ChatGPT’s US citations went to Wikipedia in January and February 2026. Reddit took 11.97%.
- How it was measured
- 897,359 ChatGPT citation events. In Google AI Mode the top domain was fandom.com, at 7.16% of 591,000 citations.
- What it does not tell you
- Two months in one country. Citation shares move quickly as the engines change how they search.
- Source
- Similarweb: The most cited domains by LLMs, April 14, 2026. 897,359 ChatGPT and 591,000 Google AI Mode citation events, US, January to February 2026.
- Evidence
- Results study. Search results or AI answers collected at scale, usually by an SEO tool vendor.
-
15%vs 17%
of links in Google’s AI summaries pointed to Wikipedia, YouTube or Reddit. In standard results, 17% did.
- How it was measured
- Same Pew sample. The three sites were the most cited sources in both AI summaries and standard results.
- What it does not tell you
- Google only, and from early 2025.
- Source
- Pew Research Center: Google users are less likely to click on links when an AI summary appears, July 22, 2025. Browsing by 900 US adults in March 2025: 68,879 unique searches, 12,593 with an AI summary. Results collected April 7 to 17.
- Evidence
- Research. Nonprofit or academic research with a published method.
Google says its AI features can run related searches to assemble an answer, so a page can be useful beyond the original query. Read how AI assistants choose citations for the mechanism.
How much traffic does AI send?
Very little, so far. In Ahrefs’ March 2026 sample, AI chatbots sent 0.28% of visits. Google sent a hundred times more.
AI chatbot visits by platform
AI visits that arrive without a referrer are counted as direct, so the AI share is likely understated.
-
12.1%from 0.5% of visits
of Ahrefs’ own signups came from AI search visitors, who made up 0.5% of its traffic.
- How it was measured
- Ahrefs reported its own signup data in the same article.
- What it does not tell you
- One company’s site. Treat it as a hint about lead quality, not a benchmark for yours.
- Source
- Ahrefs: AI chatbot traffic: what it is, and how to get more, May 15, 2026. Referral traffic to 74,752 websites in Ahrefs’ tracker, March 2026.
- Evidence
- Traffic data. Visits or clicks measured by a panel, analytics or network logs.
What happens after an AI referral?
The visits AI does send tend to be engaged. In US retail they convert better than other traffic. In travel they have caught up.
- Retail · Jul 2026 +60%
- Travel · Jul 2025 −47%
- Travel · Jul 2026 −1%
The difference in conversion rate between AI-referred visits and every other visit in the same month.
-
+62%
growth in AI referrals to US retail sites, July 2026 against July 2025.
- How it was measured
- Adobe Analytics data on more than 1 trillion visits to US retail sites. AI referrals are up 1,219% since October 2024.
- What it does not tell you
- Growth in visits from AI, not AI’s share of all retail traffic.
- Source
- Adobe Digital Insights: US consumers are embracing LLMs to make travel plans, August 19, 2026. Adobe Analytics, more than 1 trillion visits to US retail sites. July 2026 against July 2025.
- Evidence
- Traffic data. Visits or clicks measured by a panel, analytics or network logs.
-
+60%
higher conversion rate for AI-referred visits to US retail sites than for other visits, July 2026.
- How it was measured
- The 11th month in a row AI traffic converted better. AI visitors also spent 59% more time on site, were 33% less likely to bounce and added to cart at a 28% higher rate.
- What it does not tell you
- A relative difference in conversion rate, not a 60% purchase rate.
- Source
- Adobe Digital Insights: US consumers are embracing LLMs to make travel plans, August 19, 2026. Adobe Analytics, more than 1 trillion visits to US retail sites. July 2026 against July 2025.
- Evidence
- Traffic data. Visits or clicks measured by a panel, analytics or network logs.
-
−1%from −47%
conversion gap for AI-referred visits to US travel sites in July 2026. A year earlier they converted 47% worse.
- How it was measured
- Same Adobe dataset. AI traffic to US travel sites grew 119% year over year, and AI visits ran 67% longer than other visits.
- What it does not tell you
- A sector benchmark. Retail and travel already behave differently, so neither is a forecast for a software business or a publisher.
- Source
- Adobe Digital Insights: US consumers are embracing LLMs to make travel plans, August 19, 2026. Adobe Analytics, more than 1 trillion visits to US retail sites. July 2026 against July 2025.
- Evidence
- Traffic data. Visits or clicks measured by a panel, analytics or network logs.
These are sector benchmarks. Use your own referral, trial and sales data before forecasting revenue from AI search.
How do buyers use AI to choose?
Nearly every business buyer now uses AI along the way. They still check what it says against reviews and people.
-
94%
of business buyers say they used AI during their buying process.
- How it was measured
- Forrester’s Buyers’ Journey Survey. Forrester reports that buyers validate AI output against trusted outside sources.
- What it does not tell you
- Self-reported, and “used AI” covers anything from one question to a whole shortlist.
- Source
- Forrester: The State of Business Buying, 2026, January 21, 2026. Buyers’ Journey Survey of nearly 18,000 business buyers worldwide.
- Evidence
- Survey. What people say they do. Strong on intent, weaker on behavior.
-
82%
of B2B software buyers sourced recommendations from an AI chatbot in the last 24 months.
- How it was measured
- G2 surveyed more than 1,000 B2B software buyers and interviewed 50+ sales and marketing leaders.
- What it does not tell you
- A survey of what buyers say, not observed behavior. G2 runs a software review marketplace.
- Source
- G2: 2026 Buyer Behavior Report, July 22, 2026. More than 1,000 B2B software buyers, plus interviews with 50+ sales and marketing leaders.
- Evidence
- Survey. What people say they do. Strong on intent, weaker on behavior.
-
38%vs 37%
of buyers named review sites as the top influence on their shortlist, just ahead of AI chatbots at 37%.
- How it was measured
- Same G2 survey. It was the first time review sites passed AI chatbots on this question.
- What it does not tell you
- A one-point gap is too small to call a winner.
- Source
- G2: 2026 Buyer Behavior Report, July 22, 2026. More than 1,000 B2B software buyers, plus interviews with 50+ sales and marketing leaders.
- Evidence
- Survey. What people say they do. Strong on intent, weaker on behavior.
How much do AI bots take from websites?
AI companies crawl far more pages than they send visitors back. Training is now the biggest single reason bots crawl.
Log scale. Each gridline is ten times the last. Ratios move month to month, and Cloudflare publishes current ones on Radar.
-
52%from 22%
of crawler requests were for AI training in June 2026, up from 22% in spring 2025.
- How it was measured
- Cloudflare sorts crawler requests on its network by purpose, into training, search and user action.
- What it does not tell you
- A share of crawler requests on Cloudflare’s network, not of all web traffic.
- Source
- Cloudflare: Content Independence Day, one year on, July 1, 2026. Crawler requests across Cloudflare’s network.
- Evidence
- Traffic data. Visits or clicks measured by a panel, analytics or network logs.
-
38,066:1
pages crawled by Anthropic for each visit it referred in July 2025. Google’s ratio was 5:1.
- How it was measured
- Cloudflare divides each company’s crawler requests by the referrals its products send. OpenAI was at 1,091:1, Perplexity 195:1 and Microsoft 41:1.
- What it does not tell you
- Ratios swing month to month. Cloudflare’s 2025 review put Anthropic between about 25,000:1 and 100,000:1 after May. Live figures are on Cloudflare Radar.
- Source
- Cloudflare: The crawl-to-click gap, August 29, 2025. Crawler requests compared with referrals from each company’s products, January to July 2025.
- Evidence
- Traffic data. Visits or clicks measured by a panel, analytics or network logs.
-
15x
growth during 2025 in crawling triggered by people asking a chatbot to read a page.
- How it was measured
- Cloudflare’s “user action” crawls are visits an AI tool makes because someone asked it about a page, as opposed to training or search indexing.
- What it does not tell you
- Growth from a small base. At peak, training crawls still ran 32 times higher.
- Source
- Cloudflare: 2025 Cloudflare Radar Year in Review, December 15, 2025. AI crawler requests across Cloudflare’s network in 2025, split by purpose.
- Evidence
- Traffic data. Visits or clicks measured by a panel, analytics or network logs.
Cloudflare publishes the current ratio for each company on Cloudflare Radar.
What does GEO research actually show?
Generative engine optimization means improving how content appears in AI answers. The research gives you ideas to test, not a guaranteed lift.
-
Up to 40%
visibility improvement in the GEO paper’s experimental benchmark.
- How it was measured
- Aggarwal and colleagues tested changes such as adding relevant statistics, quotations and source citations. Results varied by domain and method.
- What it does not tell you
- Visibility inside generated answers in a research benchmark. Not traffic, not sales, and not a promise about today’s ChatGPT or Google.
- Source
- Aggarwal et al.: GEO: Generative Engine Optimization (KDD 2024), KDD 2024. Experimental benchmark of generative-engine answers. Measures visibility inside an answer, not visits or sales.
- Evidence
- Research. Nonprofit or academic research with a published method.
For Google, the practical starting point remains accessible, useful, indexable content. Its official guidance on AI features says no special AI markup or schema is required. Our GEO vs SEO guide explains how the work fits together.
Turn the benchmarks into a check on your site.
Keep three things apart in your reporting. Improving one does not prove the others moved.
Does AI mention you?
Ask the questions your buyers ask and read the answers. ContextBolt SEO checks live answers from ChatGPT, Gemini, Claude and Perplexity, and keeps mention history for ChatGPT and Google AI.
Does it cite your page?
Open the pages it cites. If one belongs to a competitor, read it as your brief, and ask what it answers that your page does not.
Does anyone arrive and buy?
Compare with your Search Console clicks and your own referral, trial and sales data. ContextBolt SEO connects Search Console and keeps every finding on your SEO Dashboard.
7-day trial with a 100-credit cap and card required. Then $35/month for 1,000 credits. Cancel before the trial ends to avoid the charge.
The companion article walks through what to measure, and our AI visibility tools comparison explains the buying options.
Common questions about AI search statistics.
What percentage of Google searches end without a click?
SparkToro reported 68.01% for US Google searches from January to April 2026, using Similarweb desktop and mobile web data. The Google mobile app was excluded. This measures every zero-click search in that panel, not only searches showing an AI Overview, and it is not a worldwide rate.
How many people use ChatGPT, Gemini and Google AI Mode?
OpenAI reported more than 1 billion weekly active ChatGPT users on August 31, 2026. Google reported more than 1 billion monthly Gemini app users on August 11, 2026, and more than 1 billion monthly AI Mode users on July 22, 2026. These are different reporting periods and overlapping audiences, and ChatGPT use includes far more than search, so the counts cannot be added.
Do AI Overviews reduce website clicks?
Several studies find fewer clicks when AI summaries appear. Pew observed traditional-result clicks on 8% of visits with a summary against 15% without one. Ahrefs estimated 58% lower clickthrough for the top-ranking page in its desktop sample. Semrush found only a small fall in zero-click rate on the same keywords. None of these predicts the change on a particular site; query intent and study design matter.
How often do AI Overviews appear?
It depends on the query and the date. Ahrefs found AI Overviews on 20.5% of 146 million results pages in September 2025, and on 57.9% of question queries. On US branded keywords, Ahrefs measured 73.3% in September 2026, up from 61.3% in July.
Does a high Google ranking guarantee an AI citation?
No. Ahrefs found 37.1% of cited URLs in its March 2026 study ranked in the organic top ten for the same query. Its roughly 38% headline includes other search-result blocks. Google can retrieve related queries when building an AI answer, so being absent from the original top ten does not mean SEO is irrelevant.
How much traffic does AI search send to websites?
Not much yet. Ahrefs found AI chatbots sent 0.28% of visits to 74,752 websites in March 2026, while Google sent 28.12%. Conductor measured 1.08% across enterprise sites in 2025. Some AI visits arrive without a referrer, so both figures likely undercount.
Does AI referral traffic convert better?
It depends on the sector and period. Adobe found AI referrals converted 60% better than non-AI traffic on US retail sites in July 2026. On travel sites that month, AI visits converted about 1% worse, up from 47% worse a year earlier. A retail benchmark is not a forecast for a software business, publisher or individual website.
Can GEO guarantee a 40% visibility increase?
No. The KDD 2024 GEO paper reported improvements of up to 40% in an experimental benchmark, with results varying by domain and method. That is a measured research result, not a promise about current ChatGPT or Google responses, website traffic or revenue. Test changes against your own questions and outcomes.
AI search statistics: sources, dates and limits
Reviewed October 7, 2026. This is a curated reference, not original ContextBolt research. We link to the organization reporting each result. Company user totals, panel measurements, surveys and controlled experiments are different kinds of evidence, and each figure's tag says which.
-
Company-reported weekly active users, worldwide. Counts every use of ChatGPT, not only search.
Used for 1B+
-
Company-reported monthly users of the Gemini app, worldwide.
Used for 1B+
-
Company-reported monthly active users of AI Mode, worldwide.
Used for 1B+
-
Company-reported monthly users of AI Overviews across more than 200 countries and territories.
Used for 2B+
- Traffic dataSparkToro: In 2026, less than one third of Google searches still send a clickJune 9, 2026
Similarweb US desktop and mobile web panel, January to April 2026. Excludes the Google mobile app.
-
Online survey of about 2,000 people in each of 48 markets.
Used for 10%
- ResearchPew Research Center: Google users are less likely to click on links when an AI summary appearsJuly 22, 2025
Browsing by 900 US adults in March 2025: 68,879 unique searches, 12,593 with an AI summary. Results collected April 7 to 17.
-
300,000 keywords, aggregated desktop Search Console data from December 2023 and December 2025.
Used for 58%
-
10M+ keywords for prevalence, January to November 2025. A separate 200,000+ keyword sample for zero-click change.
-
92.5 million branded-keyword crawls of US desktop Google results, July to September 2026.
Used for 73.3%
-
146 million desktop search results pages, collected September 2025.
Used for 20.5%
-
600,000+ US keywords in 10 industries, November 2025 to April 2026.
Used for +71%
-
863,000 search results pages and 4 million cited URLs, January 2026.
Used for 37.1%
-
Nine industries monitored from May 2024 to September 2025.
Used for 54.5%
-
897,359 ChatGPT and 591,000 Google AI Mode citation events, US, January to February 2026.
Used for 13.15%
-
Referral traffic to 74,752 websites in Ahrefs’ tracker, March 2026.
-
3.3 billion sessions on enterprise sites in 10 industries, May to September 2025.
Used for 1.08%
- Traffic dataAdobe Digital Insights: US consumers are embracing LLMs to make travel plansAugust 19, 2026
Adobe Analytics, more than 1 trillion visits to US retail sites. July 2026 against July 2025.
-
Buyers’ Journey Survey of nearly 18,000 business buyers worldwide.
Used for 94%
-
More than 1,000 B2B software buyers, plus interviews with 50+ sales and marketing leaders.
-
Crawler requests across Cloudflare’s network.
Used for 52%
-
Crawler requests compared with referrals from each company’s products, January to July 2025.
Used for 38,066:1
-
AI crawler requests across Cloudflare’s network in 2025, split by purpose.
Used for 15x
-
Experimental benchmark of generative-engine answers. Measures visibility inside an answer, not visits or sales.
Used for Up to 40%
-
Official eligibility and content guidance for Google’s AI features. Guidance, not an effectiveness study.
The ChatGPT milestone chart links each dot to its report: TechCrunch for November 2023, March and October 2025 and February 2026, and Engadget for August 2024 and February 2025.
To cite a figure, keep its geography, period and denominator, and link to the original study. Each figure's Cite button copies a citation that does this, and its Link button copies a link to that row. Spotted an error? Email hello@contextbolt.com.
What changed on October 7, 2026
- Added three chapters, on how much traffic AI sends, how buyers use AI and how much AI bots crawl.
- Added Gemini’s and AI Overviews’ user counts, AI Mode’s share of US searches, AI chatbot visit share and Reuters Institute news use.
- Added Ahrefs’ 2026 branded-search data and its September 2025 coverage crawl, and Semrush’s commercial-intent study.
- Updated the travel conversion figure with the year-earlier comparison, and moved Semrush’s zero-click comparison beside the other click studies.
- Corrected SparkToro’s publication date to June 9, 2026. Dropped Semrush’s informational-share figure; newer coverage data replaces it.
- Every figure now carries its method, its limits, a ready-made citation and a direct link.