AI Search & GEO Statistics for B2B SaaS 2026: Primary Research and Benchmarks

B2B buyers are using AI tools to find software, Google is adding AI answers to more search results, and AI assistants are starting to send measurable referral traffic. A brand can appear by name in an AI answer, earn a source link or receive a visitor who clicks through. Those are three different outcomes, and this article measures them separately.

These 42 AI search and GEO statistics come from original surveys and data analyses with 2026 fieldwork or measurement. The research covers B2B software buyers, search results, AI platform visits, citations and technical tests. Each finding links to the organization that gathered the underlying data.

2026 AI search and GEO statistics at a glance

• 84% of surveyed SaaS CMOs used AI during vendor discovery.
• 60% of surveyed US adults had read search AI summaries.
• 71% growth in AI Overviews on commercial queries appeared in Semrush’s panel.
• 9.69% position-one CTR was measured on US desktop AI Overview SERPs.
• 62% of analyzed AI citations lacked an explicit brand mention.
• 97% of valid llms.txt files in Ahrefs’ sample received no requests.

Original research and measurement periods

How B2B SaaS buyers use AI to discover software

1. 84% of B2B SaaS CMOs surveyed by Wynter used AI for software vendor discovery

Wynter asked 101 CMOs at SaaS companies with at least $50 million in revenue about their buying process in January 2026. Eighty-four percent used AI tools or language models to identify vendors, up from 24% in the prior survey. Wynter CMO vendor-discovery survey.

A buyer asking for field-service software with offline time tracking needs to see which vendor supports offline recording, on which plan, and how the data syncs. That’s the level of detail I’d put on product pages.

2. 68% of surveyed SaaS CMOs started vendor research with AI tools

In the same Wynter study, 68% said they began discovering software through AI before traditional search. Only 9% started by searching Google for a software category. Wynter CMO vendor-discovery survey.

After an AI-generated shortlist, buyers still need feature documentation, plan restrictions and comparisons to verify which option actually fits their workflow.

3. 72% of surveyed SaaS CMOs used Google during vendor research

Google appeared in 72% of Wynter’s January 2026 CMO responses, often as a way to investigate or verify recommendations from AI and other channels. Wynter CMO vendor-discovery survey.

If AI recommends a product, the buyer may search for pricing, security documentation or alternatives before contacting sales. Those pages need to answer specific evaluation questions.

4. 65% of surveyed SaaS CMOs used peer communities when finding software vendors

Sixty-five percent of Wynter’s 101 CMOs mentioned private peer communities when describing how they searched for vendors. Wynter CMO vendor-discovery survey.

A buyer may ask peers about deployment problems and compare those answers with the vendor’s setup guide. Publishing detailed implementation material gives people facts to investigate.

5. 42% of surveyed SaaS CMOs ranked word of mouth as their top vendor-consideration factor

Asked to rank the influences that get a vendor considered, 42% of Wynter’s respondents put peers and word of mouth first. AI recommendations ranked first for 10%, ahead of Google Search at 8%. Wynter CMO vendor-discovery survey.

A useful customer story records the old workflow, what changed during setup and what the customer measured afterward. Peers can then judge whether the circumstances resemble their own.

6. 34% of surveyed SaaS CMOs invested in GEO or AEO in 2026

Wynter reported that 34% of its CMO respondents were investing in generative or answer engine optimization in 2026. Wynter CMO vendor-discovery survey.

Begin with ten questions prospects ask on demo calls. Check whether the website answers them, whether the answers are indexed and whether AI tools link to those pages.

AI summaries, search habits and AI platform usage

7. 60% of US adults surveyed by Pew said they had read AI summaries in search results

Pew Research Center’s February 17–23, 2026 survey found that 60% of US adults had encountered and read AI summaries at the top of search results. Pew Research Center AI survey.

A publisher’s audience may read an AI summary and leave the results page. Search Console impressions and clicks show how those visits change on the publisher’s own pages.

8. 44% of US adults surveyed by Pew said they used ChatGPT

In Pew’s February 2026 survey, 44% of US adults reported using ChatGPT at least once, up from 34% in its 2025 wave. Pew Research Center AI survey.

For audience research, ask buyers which tools they used to build their shortlist. Their actual behavior is more useful for selecting platforms than a general-population usage figure.

9. 24% of US adults surveyed by Pew said they used Gemini

Pew’s February 2026 respondents reported Gemini usage at 24%, making it the second-most-used named chatbot in that survey. Pew Research Center AI survey.

If customer interviews repeatedly mention Gemini, add it to the prompt-tracking sample. Monitor the interfaces buyers actually use.

10. 17% of US adults surveyed by Pew said they used Microsoft Copilot

In the same February 2026 Pew research, 17% of US adults said they used Microsoft Copilot. Pew Research Center AI survey.

For a SaaS vendor selling into Microsoft-heavy workplaces, ask buyers whether Copilot played a role in research and check the same evaluation prompts there.

11. ChatGPT accounted for roughly 53% of generative AI website visits in Similarweb’s May 2026 data

Similarweb’s global panel placed ChatGPT at about 53% of visits to major generative AI websites in May 2026. The share had been about 76% in June 2025. Similarweb AI search market-share analysis.

Track the same few product-category questions in more than one assistant. Results can differ between interfaces even when the buyer asks the same thing.

12. Gemini accounted for roughly 27–28% of AI website visits in Similarweb’s May 2026 data

Similarweb measured Gemini’s share of worldwide AI platform website traffic at approximately 27–28% in May 2026, up from less than 9% in June 2025. Similarweb AI search market-share analysis.

A vendor serving teams already using Google Workspace might prioritize Gemini in its monitoring panel; another serving Microsoft enterprises might examine Copilot more closely.

13. Claude accounted for nearly 9% of AI website visits in Similarweb’s May 2026 data

Similarweb found Claude had grown to nearly 9% of generative AI website visits by May 2026, from around 2% the previous year. Similarweb AI search market-share analysis.

I’d revisit platform selection quarterly using customer interviews and referral logs. A fixed dashboard built around one assistant can miss where buyers have moved.

AI Overview appearance in Google search results

14. Semrush measured 71% growth in AI Overviews on commercial-intent searches

Semrush analyzed more than 600,000 keywords across ten industries in a six-month study published in July 2026. The prevalence of AI Overviews on commercial-intent results increased by an average of 71%. Semrush AI Overviews commercial-search study.

Track commercial searches such as ‘best payroll software for contractors’ and ‘[product] alternatives’ alongside definition queries. Those searches reflect different stages of evaluation.

15. Semrush observed a 5% decline in AI Overviews on transactional searches

In the same six-month analysis, Semrush recorded a 5% decline in AI Overview appearance on transactional-intent queries. Semrush AI Overviews commercial-search study.

Group queries by informational, comparison and purchase intent. The same percentage change in AI Overviews will have different implications for each group.

16. AI Overviews and Google Ads appeared together about twice as often in Semrush’s 2026 study

Semrush reported that ads and AI Overviews co-occurred on the same result page at roughly twice the earlier rate in its tracked keyword groups. Semrush AI Overviews commercial-search study.

For a priority comparison query, record whether ads, an AI Overview and your organic listing appear together. That gives both teams a shared view of what the buyer sees.

17. 59.73% of SE Ranking’s tracked queries showed an AI Overview in February 2026

SE Ranking followed 100,000 keywords across 20 niches and found AI Overviews on 59.73% of its tracked queries in February 2026. SE Ranking AI Overviews tracking.

Create a set of fifty high-intent queries from Search Console and sales conversations, then check how many display an AI Overview in a consistent location and device setting.

18. 60.85% of SE Ranking’s tracked queries showed an AI Overview before its January 2026 model change

In the January 2026 wave before Google’s move to Gemini 3 for AI Overviews, SE Ranking recorded AI Overviews on 60.85% of the same tracked-query program. SE Ranking AI Overviews tracking.

Keep query selection, device and location stable when comparing AI Overview occurrence from one month to the next. Otherwise changes in the sample can obscure the trend.

19. 13.7% of 55,393 Google searches in a 2026 academic experiment triggered AI Overviews

Researchers Xu, Iqbal and Montgomery issued 55,393 trending queries across 19 topical categories from March 13 to April 21, 2026 and recorded a 13.7% overall AI Overview activation rate. Xu, Iqbal and Montgomery AIO measurement paper.

The rate depends heavily on what a query sample contains. A tracked set of trending searches will differ from a panel selected for informational or commercial questions.

20. 64.7% of question-form searches triggered AI Overviews in the 2026 academic experiment

Within the same March–April 2026 experiment, 64.7% of queries phrased as questions triggered an AI Overview. Xu, Iqbal and Montgomery AIO measurement paper.

For a how-to query, open the AI Overview and read the linked sources. A product page can add an actual walkthrough, screenshots and implementation decisions beyond a short summary.

Organic clicks and AI search referral traffic

21. 9.69% was the first-position organic CTR on AI Overview SERPs in Advanced Web Ranking’s July study

Advanced Web Ranking measured 9.69% organic click-through rate for position one on Google US desktop result pages with AI Overviews in July 2026, compared with 20.02% across its broad SERP sample. Advanced Web Ranking CTR study.

If the target keyword triggers an AI Overview on desktop, use the study’s AI Overview position-one benchmark as context and compare it with your actual Search Console CTR.

22. 74.3% of search sessions with AI Overviews had no organic click in Advanced Web Ranking’s study

Advanced Web Ranking’s July 2026 data showed a 74.3% zero-click rate on AI Overview desktop SERPs, versus 53.6% on SERPs without AI Overviews. Advanced Web Ranking CTR study.

If impressions rise but visits fall for one query group, inspect its ranking changes and results page before concluding that the content itself lost relevance.

23. AI referrals represented 0.32% of site traffic in SE Ranking’s January–April 2026 panel

SE Ranking reported an average 0.32% of website visits from AI search engines during January–April 2026. The same panel measured traditional organic search at 42.75% of visits. SE Ranking AI referral study.

Keep AI referrals and organic search as separate channels. For each, compare landing pages, engaged visits and leads instead of treating raw visit counts as equivalent.

24. SE Ranking recorded organic search at 42.75% of website traffic in early 2026

In its January–April 2026 traffic-panel analysis, SE Ranking reported organic search accounting for 42.75% of website visits. SE Ranking AI referral study.

For a SaaS site, Search Console remains central to understanding organic discovery. AI referral logs add another view of which content earns visits.

25. Similarweb measured 770.7 million monthly AI referral visits through May 2026

Similarweb’s global referral dataset counted an average of 770.7 million monthly visits from AI platforms between June 2025 and May 2026, across desktop and mobile web. Similarweb AI referral traffic study.

Set up an AI referral report with source platform, landing page, engaged sessions and qualified actions. The global figure gives context for the channel’s size.

26. Similarweb found 117.4% annual growth in AI referral visits through May 2026

Similarweb recorded 117.4% year-over-year growth in average monthly AI referral visits during the June 2025–May 2026 window. Similarweb AI referral traffic study.

Show both the baseline and new AI referral visits. Going from 10 to 20 visits is 100% growth but represents ten additional people.

27. Similarweb measured 312.5% annual AI referral growth in the beauty category

Within the same Similarweb industry analysis, beauty websites recorded 312.5% year-over-year growth in AI referral visits. Similarweb AI referral traffic study.

A beauty retailer’s referral growth tells little about demand for compliance software. Use the industry numbers to identify patterns, then benchmark the relevant software niche.

AI citations, brand mentions and original sources

28. 62% of analyzed AI citations did not mention the cited brand in Semrush’s 2026 study

Semrush and Kevin Indig analyzed thousands of AI answer domain appearances and found that 62% of citations linked to a source without naming its brand in the response. Semrush Ghost Citations study.

Add two columns to the prompt audit: ‘brand named’ and ‘site URL linked.’ One answer may satisfy the first and miss the second.

29. Gemini named brands in 83.7% of appearances in Semrush’s AI citation study

In Semrush’s analysis, Gemini included a brand mention in 83.7% of appearances involving a tracked brand. Semrush Ghost Citations study.

Record each occurrence as a mention, citation, both or neither. That lets you see whether the platform is recommending the brand or using it only as a source.

30. Gemini linked to brands in 21.4% of appearances in Semrush’s citation study

The same Semrush study recorded a citation link in 21.4% of Gemini brand appearances. Semrush Ghost Citations study.

A linked product-documentation page may supply a fact without the assistant recommending the product. Review those cases separately in reporting.

31. ChatGPT cited brands in 87% of relevant appearances in Semrush’s AI citation study

Semrush reported that ChatGPT cited brands in 87% of tracked appearances, with explicit brand mentions occurring in 20.7%. Semrush Ghost Citations study.

Compare the answer’s explicit brand recommendation with its linked sources. A product can appear in one without appearing in the other.

32. ChatGPT averaged 15 citations per response in Semrush’s 2026 AI Visibility Index

Semrush analyzed 126 million US prompts from January to April 2026 across four AI search platforms. ChatGPT responses averaged 15 cited sources, while Gemini responses averaged three. Semrush 2026 AI Visibility Index.

A longer citation list creates more possible source placements. For a vendor, the practical audit remains whether the linked page is relevant and what claim it supports.

33. Only 36 brands sustained top-100 visibility across four AI platforms in Semrush’s index

Semrush reported that 36 brands held top-100 visibility across all four tracked platforms in every month from January to April 2026. Semrush 2026 AI Visibility Index.

A company visible in ChatGPT might appear less often in Gemini or Google AI Mode. Track consistent buyer prompts in each interface instead of one combined visibility score.

34. Nearly 30% of cited domains in the 2026 AIO research were absent from the first organic results page

The March–April 2026 academic study found that nearly 30% of domains cited by AI Overviews were absent from the co-displayed first page of standard results. Xu, Iqbal and Montgomery AIO measurement paper.

Review the pages that AI Overviews cite for your key queries, then compare them with the top organic listings. The two sets may include different sites.

35. 11% of atomic claims in sampled Google AI Overviews lacked support in cited pages

Researchers decomposed AI Overview answers into 98,020 individual claims and found 11% unsupported by the cited source pages in their March–April 2026 study. Xu, Iqbal and Montgomery AIO measurement paper.

For consequential claims, show the exact feature, plan, date or measurement on the linked source page so the reader can verify it quickly.

Tests of GEO tactics and technical implementations

36. 28% of websites in Ahrefs’ May 2026 analytics panel published an llms.txt file

Ahrefs examined 137,210 domains receiving traffic through Ahrefs Web Analytics in May 2026. Twenty-eight percent had a valid llms.txt file at the domain root. Ahrefs llms.txt server-log study.

Before creating an llms.txt file, check your robots rules, indexability, documentation and whether the important product pages are current. Those checks affect ordinary discovery too.

37. 97% of the llms.txt files in Ahrefs’ May 2026 panel received no requests

Of roughly 38,000 domains with valid llms.txt files, 97% received no requests during May 2026 in Ahrefs’ server-log analysis. Ahrefs llms.txt server-log study.

The server logs give a concrete adoption signal: very few of the sampled files were requested. That belongs in the technical prioritization discussion.

38. 96% of observed llms.txt requests in Ahrefs’ study came from bots

Among the subset of llms.txt files that received traffic, Ahrefs attributed 96% of requests to bots and 4% to human visits. Ahrefs llms.txt server-log study.

Creating a file is one action; receiving requests is another. Ahrefs’ logs measure the latter, which matters if the team expects crawlers to use the file.

39. 19.5% of observed llms.txt fetches came from named AI bots in Ahrefs’ study

Ahrefs classified 19.5% of requests that reached llms.txt files as originating from identifiable AI tools. The study also found that 5.8% of requests came from commercial GEO or AEO tools. Ahrefs llms.txt server-log study.

If a team publishes llms.txt, inspect server logs for requests by known crawlers before devoting effort to maintaining a large file.

40. Adding schema correlated with a 4.6% relative decline in AI Overview citations in Ahrefs’ matched test

Ahrefs studied 1,885 pages that added JSON-LD structured data and compared them with 4,000 control pages. In its matched difference-in-differences analysis through March 2026, the estimated relative Google AI Overview citation change was −4.6%. Ahrefs structured-data citation study.

Choose structured data that accurately describes real page content, such as supported product information or documentation. Then test the markup and track the eligible search features.

41. Adding schema produced a 2.4% relative AI Mode citation change in Ahrefs’ matched test

In the same matched analysis, Google AI Mode citations changed by an estimated +2.4% relative to controls. The estimate did not meet the study’s significance threshold. Ahrefs structured-data citation study.

Use Google’s structured-data validation tools and Search Console enhancements to check implementation. Evaluate AI citations separately over the same period.

42. Adding schema produced a 2.2% relative ChatGPT citation change in Ahrefs’ matched test

Ahrefs recorded an estimated +2.2% relative citation change for the ChatGPT arm of its matched study, also below its statistical significance threshold. Ahrefs structured-data citation study.

Adding JSON-LD is a technical change. Product testing, original research and documentation answer a different problem: whether the page contains information worth referencing.

How to measure AI visibility without confusing the metrics

Consider a SaaS time-tracking vendor researching how it appears in AI-assisted buying journeys. I would begin with a small, repeatable set of prompts based on actual buyer questions, covering product categories, comparisons, integrations and implementation concerns.

For every prompt, save the date, assistant or search interface, the exact wording, whether the brand was named, which URLs were cited and what the answer said about the product. Repeating the same set under consistent conditions gives the team a record it can review over time.

Brand mentions, citations and referred visits answer different questions

Brand mention: the answer names the vendor. Source citation: the answer includes a link to a page on the vendor’s website. Referred visit: someone actually clicks through from an AI interface. A brand may be cited as a source without being recommended as an option, or named without receiving a link.

I’d track all three separately. A useful report might show which prompts mention the brand, which pages attract citations and which linked visits lead to product trials or sales conversations. The Semrush ghost-citations research illustrates why a named brand and a linked source deserve different fields in that report.

How I would use these findings in a B2B SaaS content program

Start by mapping the research questions buyers ask before they request a demo. Check which pages answer questions about product fit, integrations, alternatives, setup costs, implementation and customer evidence. Then examine which questions generate AI Overviews and how often the brand appears in AI-assisted discovery.

1. Audit product evidence. Check documentation, current features, plan restrictions, verified examples and customer outcomes on evaluation-stage pages.
2. Measure search visibility by query. Keep organic clicks, impressions, AI Overview exposure, cited-source links and named-brand mentions as distinct fields.
3. Inspect AI referrals. Identify which assistants refer visits, which pages they reach and which sessions result in useful actions.
4. Publish something worth citing. Product testing, primary surveys, carefully sourced reports and clear technical documentation provide specific information buyers can evaluate.

Related reading: SEO vs GEO for B2B SaaS, 2026 SaaS SEO benchmarks, and how I verify statistics before citing them.