AI Search Statistics for 2026: Usage, Clicks and Citations
Usage, clicks and citations, each figure with its source, its date and its limit. Where the studies disagree, both numbers are here.
AI search statistics for 2026 show use rising and clicks falling. Google says AI Overviews have 2.5 billion monthly users, ChatGPT reports 900 million weekly users, and 68% of US Google searches in early 2026 ended without a click.

Open any roundup of AI search statistics and count the dates. Then count the sources.
It won’t take long. A figure with no date, no sample and no seller behind it is a rumor with a decimal point, and the genre runs on them. One page sets a user count from last spring beside a click rate from last year and never mentions the gap.
This page holds fewer numbers. Each one has a name, a date, a link and a note on what limits it. If I couldn’t read the original, the figure stays out. Where two sources disagree, you’ll see both, because the disagreement is the finding.
C4 Solutions sells SEO and AEO services, so you have a reason to doubt me. Click the links.
The Short Version
Use is climbing, clicks are falling and citations are the unsettled part. OpenAI says ChatGPT reached 900 million weekly users in February 2026, Google says AI Overviews has more than 2.5 billion monthly users, and SparkToro put US Google searches that ended without a click at 68.01% for January to April 2026. The share of AI Overview citations that also rank in the top 10 swings furthest, from about 17% in one study to about 38% in another, so quote it with its source or leave it out. The platforms’ own pages publish access rules and no rates.
These numbers sit beside our work on SEO, AEO and GEO for tech companies. The behavior behind the click figures is in our guide to zero-click search, and the MSP-specific measurements are in our AI search visibility benchmarks for MSPs. To see how a share-of-voice percentage is calculated, read how to measure AI share of voice.
AI Search Statistics on Usage: Who Is Asking
Start with the big numbers. They get quoted first and checked last.
Usage Figures, Who Stated Them, and the Limit
| What was counted | The figure | Who said it, when, and the limit |
|---|---|---|
| ChatGPT weekly users | 900 million, plus 50 million paying subscribers | OpenAI, announced February 27, 2026, reported by TechCrunch. A weekly count of one app. |
| ChatGPT, newer report | “On the verge of” 1 billion weekly users | The Information, reported by The Next Web on July 29, 2026. Based on figures OpenAI shared internally, not a published count. |
| Google AI Overviews | More than 2.5 billion monthly users | Google, in the I/O 2026 keynote post, May 2026. Google’s own count, with no method attached on that page. |
| Google AI Mode | More than 1 billion monthly users in about a year | Google, same post. Same limit. |
| AI Mode visits | 126 million in June 2025, 279 million in May 2026 | Similarweb, reported by TechCrunch on July 27, 2026. The unit is visits, not people. |
| Heavy AI tool use | 20% or more of Americans use an AI tool ten or more times a month | SparkToro, June 8, 2026. Stated in passing, with no source given for it. |
Read the first and third rows together. You can’t set 900 million weekly ChatGPT users against 2.5 billion monthly AI Overviews users and call it a race. Weekly isn’t monthly. OpenAI counts people who open its own app, while Google counts people who used a feature inside a search page they were already on.
Different units. Different stories.
The billion mark for ChatGPT is a report, not a count. The Next Web said on July 29, 2026 that ChatGPT was on the verge of 1 billion weekly users, citing figures that OpenAI had shared internally, as The Information first reported. I found no published OpenAI figure above 900 million, so a roundup that states a billion as settled is ahead of its source.
How Often an AI Overview Shows Up
Here the sources stop agreeing. Three named measurements, three answers.
Share of Google Searches With an AI Overview, by Source
| Source | Period | The figure | The limit |
|---|---|---|---|
| Pew Research Center | March 2025 | 18% of Google searches produced an AI summary. 58% of respondents saw at least one. | 900 US adults, one month, Google only, observational. |
| SparkToro | January to April 2026 | “20%+” of all searches | Stated in passing, with no method given for that figure. |
| Similarweb, via TechCrunch | A year to mid-2026 | From 15% of searches to 43% | Similarweb’s panel. The write-up doesn’t say which searches were counted. |
The last two describe overlapping stretches of 2026 and don’t come close. Neither source says, in the pages I read, how it defined a search or whose searches it counted. A panel of US users, a sample of results pages and a list of tracked keywords will each produce a different share, and I can’t tell you which one is right.
I can tell you not to average them.
Look at your own pages instead. Google says sites that appear in AI Overviews and AI Mode are counted in overall search traffic, in the Performance report under the Web search type, so you need no new tool to start looking at your own pages. Pull twelve months of impressions and clicks for the pages that sell, and read them against the table above.
AI Search Statistics on Clicks: What Happens After the Answer
This is the part that touches revenue. The measurements point the same way, though none of them agrees on size.
Click Figures, Who Measured Them, and the Limit
| What was measured | The figure | Who measured it, and the limit |
|---|---|---|
| Clicks on a traditional result, with and without an AI summary | 8% of visits with one, 15% without. 1% of visits clicked a link inside the summary. 26% of pages with a summary ended the session, against 16% without. | Pew, March 2025, 900 US adults. Observational, Google only. |
| US Google searches ending with no click | 68.01% in January to April 2026, against 60.45% in 2024 | SparkToro. Similarweb’s panel for 2026, Datos’ for 2024, so the comparison is rough, and SparkToro says so. |
| Top-ranking page’s click rate where an AI Overview shows | 58% lower | Ahrefs, 300,000 keywords, December 2025 data, update dated February 4, 2026. Desktop only and a correlation. The April 2025 study found 34.5%. |
| Google’s share of traffic to 75,000 opted-in sites | Down 8 points, about 22%, from June 2025 to May 2026 | Ahrefs’ tracker, as reported by SparkToro, which calls the panel far from representative. |
| Searches that reach AI Mode, and traffic AI tools send out | 0.34% of searches. Under 1% of all traffic. | SparkToro, January to April 2026. The second figure is stated in passing. |
| ChatGPT visits landing on webpages, US desktop | From 25% in March 2026 to nearly 60% by May 30, 2026 | Similarweb, via TechCrunch, after a May 7 search update. The write-up doesn’t define the unit further. |
Start with Pew’s row, because nobody was guessing there. It logged what 900 US adults did in March 2025. With an AI summary on the page they clicked a traditional result in 8% of visits, and without one, 15%. Only 1% of visits included a click on a link inside the summary. That’s tiny.
Then the two vendor studies. SparkToro’s 68.01% and Ahrefs’ 58% aren’t the same measurement. One counts searches that end with no click at all, while the other compares the click rate of the top-ranking page where an AI Overview shows with the rate where none does. A site can see both at once. Ahrefs flags its own limits, desktop data and a correlation, and found a bigger effect than in its April 2025 study.
Same direction. Different rulers.
None of the studies above counts how many searches happen. They count what happens after the search. That gap is where the doom headlines come from.
The last row is the useful one if your buyers ask an assistant first. In Similarweb’s US desktop data, the share of ChatGPT visits landing on webpages rose from 25% in March 2026 to nearly 60% by the end of May, after a May 7 search update. A link that sends traffic out is a product decision, and a product decision can change in a month. Our guide to why searchers stop clicking covers the behavior behind these figures.
AI Search Statistics on Citations: Who Gets Named
Clicks are the cost side. Citations are the other side of the ledger, and they’re the least settled numbers on this page.
Citation Figures, Who Measured Them, and the Limit
| What was measured | The figure | Who measured it, and the limit |
|---|---|---|
| AI Overview citations that also appear in the first 10 blocks of the results page | 37.9% in the first 10 blocks, 31.2% at positions 11 to 100, 31.0% beyond 100. Blue links only: 37.1%, 26.2%, 36.7%. | Ahrefs, March 2, 2026. 863,000 results pages and 4 million AI Overview URLs, from Ahrefs’ own AI visibility tool. |
| The same overlap, a different study | About 17%, against 54% in October 2025 | BrightEdge, February 12, 2026, as reported by Search Engine Journal. A different method and dataset. I haven’t read BrightEdge’s own paper. |
| Sources per AI summary | 88% cited three or more. 1% cited one. Median length 67 words. | Pew, March 2025. Google only. |
| ChatGPT queries that carried citations, US desktop | 6.8% as of May 2026. Cited responses up more than fivefold in a year. | Similarweb, via TechCrunch. |

Two studies, about 17% and about 38%, on the same question in the same winter. BrightEdge used a different method and dataset, as Search Engine Journal notes. Ahrefs says its own figure can’t be laid over its July 2025 study either, where about 76% of cited URLs also ranked in the top 10, and since then it has improved its parsing to see more citations while Google moved AI Overviews to Gemini 3 in January 2026.
Ahrefs reads its result this way: AI Overviews lean less on the pages that rank for the original query and more on pages found through fan-out searches, the related queries Google says it runs behind the answer. That’s the company’s interpretation, not a measured cause.
So which number do you quote? Both, with the names attached. Then be exact.
Those percentages describe cited pages that also rank, not ranking pages that get cited. Ranking in the top 10 is one route into an AI Overview. Ahrefs found that about 31% of cited URLs didn’t rank in the top 100 at all.
One more figure for the skeptics. Similarweb found that 6.8% of US ChatGPT desktop queries carried citations as of May 2026, even though cited responses rose more than fivefold in a year, so the great majority of those queries carried no citation at all. Which domains get cited, as opposed to how often, is the subject of our post on what ChatGPT cites.
What the Platforms Say in Their Own Documents
The platforms publish rules. They don’t publish rates. I read the four pages below and found no click rate, no citation rate and no share of answers in any of them.
What Each Platform Publishes
| Platform | What its page says | Note |
|---|---|---|
| A page must be indexed and eligible to show with a snippet. There are no additional requirements, no new machine readable files and no special schema.org markup. Traffic from these features is reported in the Web search type of the Search Console Performance report. | AI features and your website. No date read on the page. | |
| OpenAI | OAI-SearchBot surfaces sites in ChatGPT search. GPTBot is the training crawler. Each setting is independent, and changes take about 24 hours. A site opted out of OAI-SearchBot isn’t shown in ChatGPT search answers, though it can appear as a navigational link. | Overview of OpenAI Crawlers. Undated. |
| Anthropic | Claude-SearchBot improves the quality of search results. ClaudeBot relates to training. Claude-User retrieves a page when someone asks Claude a question. | Anthropic help page. Undated. |
| Perplexity | PerplexityBot surfaces and links sites in Perplexity results and isn’t used to crawl content for foundation models. Perplexity-User supports actions users take inside Perplexity. | Perplexity crawler documentation. Undated. |
That’s still useful. Access is the one thing on this page you control, and it isn’t a statistic. Check your robots.txt and your CDN’s bot rules against the table, bot by bot. The longer argument is in our post Is SEO Dead? What AI Search Changed and What It Didn’t.
Statistics I Wouldn’t Quote
Five habits from the roundups. I’d cut each one from a deck. Cut them all.
- A user count with no unit. A billion users of what, counted over which period?
- The drop from about 76% to about 38%, presented as a collapse in how often AI Overviews cite the top 10. Ahrefs says its method changed, so it isn’t one line on one chart.
- Any rate with no sample size, no country and no month. If the page can’t say who was counted, it isn’t a number yet.
- A tools vendor’s figure about its own panel of sites, offered as the whole market. SparkToro itself calls the Ahrefs tracker’s 75,000 opted-in domains far from representative.
- A forecast dressed up as a count. Nobody has counted next year.
What to Do With AI Search Statistics
Quote fewer. Measure more of your own, in this order.
- Pick three numbers for the board, one on use, one on clicks and one on citations. Put the source and the date beside each on the same slide.
- Pull twelve months of Search Console impressions, clicks and average position for the pages that sell. Flat impressions with falling clicks point at answers on the results page. Our organic traffic drop checklist tells that pattern from a lost ranking.
- Count leads by source from the CRM. Sessions from analytics don’t count here, and our guide to SEO KPIs that tie to pipeline lists what to track.
- Once a month, ask ChatGPT, Perplexity and Gemini the ten questions a buyer would ask, and log which companies come back. No study can supply that log, and our guides to AI share of voice and Perplexity SEO cover how to read it.
- Check the table again before you reuse a figure. User counts and the AI Overview share go stale fastest, and citation overlap is already contested.
If you’d like a second read on your own numbers, book a growth strategy call.
Before You Quote These Numbers
Is the 68% zero-click figure about every search?
Google searches in the US only, from January to April 2026. SparkToro measured it on Similarweb’s panel, and the 2024 comparison of 60.45% came from a different panel, so SparkToro calls the two-year comparison rough.
It says nothing about ChatGPT or any other engine.
How many people really open ChatGPT each week?
900 million, per OpenAI’s announcement of February 27, 2026, which also reported 50 million paying subscribers.
Reports of a billion trace to figures OpenAI shared internally, as The Information reported. They aren’t a published count. Say so.
Does an AI Overview really cut clicks by more than half?
Ahrefs found a 58% lower click rate for the top-ranking page. Pew found 8% of visits clicked a result with a summary and 15% without.
The first is a ratio on one ranking position, from tracked keywords. The second is a share of visits, from 900 people. Both point down. Neither can be laid over the other. Quote them separately.
Why can’t two studies agree on how often AI Overviews appear?
Each one counts something different: a panel of people, a sample of results pages or a list of tracked keywords.
Pew found 18% in March 2025, SparkToro said 20% or more in 2026, and Similarweb said 43% by mid-2026. Pick the source closest to your buyers, name it, and don’t blend it with the others.
If a page ranks in the top 10, will an AI Overview cite it?
Not reliably. No study here says otherwise. Ahrefs found that 37.9% of cited URLs also appeared in the first 10 blocks of the results page, and BrightEdge reported about 17% overlap with the top 10.
Those figures describe cited pages that also rank, not ranking pages that get cited.
Do I need a special file or markup to appear in Google’s AI features?
Google says no. Its page lists no additional requirements, no new machine readable files and no special schema.org markup, only that a page be indexed and eligible to show with a snippet.
Other engines publish their own crawler rules, which the platform table above sets out.
Put your own numbers next to these
Bring us your Search Console export and your CRM lead sources, and we will set them beside these figures, check which crawlers can reach your pages, and say what the gap means for your plan.
Book a Growth Strategy Call