
How Buyers Use AI to Research Vendors
On this page
Marketing
How buyers use AI to research vendors depends on whom you ask and what you ask them. Gartner found that 45% of 645 B2B buyers used generative AI, primarily to gather vendor and product information. In a separate 6sense report, 94% of nearly 4,000 buyers reported using large language models, or LLMs. Marketing still works: it helps put useful information into the sources buyers and assistants consult before a shortlist forms.
| Survey | Who was asked | Fieldwork | Sample | What was asked, in plain words | Result |
|---|---|---|---|---|---|
| Gartner | B2B buyers | August to September 2025 | 645 | Did you use generative AI, primarily to gather vendor and product information? | 45% said they used it, primarily for that purpose |
| TrustRadius B2B Buying Disconnect | Technology buyers | Early 2026 | 1,862 buyers; report also surveyed 444 vendors | Did you use AI in your buying process? | 63% |
| G2 Answer Economy | B2B software buyers | March 2026 | 1,076 | Did you rely on AI chatbots during research, and did you start there more often than Google? | 71% and 51%, respectively |
| G2 Buyer Behavior | B2B decision-makers | April 2025 | 1,169 | Did you start research with tools such as ChatGPT more often than Google? | 29% |
| Forrester Buyers' Journey Survey | B2B buyers | 2024 | Sample not published in the blog | Did you use generative AI in any part of the purchasing process? | 89% |
| 6sense B2B Buyer Experience | B2B buyers | Dates not stated for the 2025 report | Nearly 4,000 | Are you using LLMs? | 94% |
| McKinsey AI Discovery Survey | US consumers | August 2025 | 1,927 | Do you intentionally seek out AI-powered search? | About 50% |
| Adobe companion survey | US consumers | Not stated (published March 2025) | More than 5,000 | Have you used generative AI for online shopping? | 39% |
These figures measure different populations and behaviors. Using an LLM, using generative AI somewhere in a purchase, and starting research in a chatbot more often than Google are different questions. The consumer figures also describe shopping or AI search, not B2B vendor selection. Read any percentage with its question and sample attached.
G2 and TrustRadius run review platforms, and 6sense sells intent data. All three sell to marketers. Their commercial interests do not make their findings wrong, but they are a reason to inspect the question and compare results with other sources.
Myths about how many buyers use AI
Myth: One percentage tells you how many buyers use AI
A single adoption figure describes every B2B purchase, so you can use it to set your entire marketing budget.
Gartner's survey found 45% of 645 B2B buyers used generative AI, primarily to gather vendor and product information. In Forrester's Buyers' Journey Survey, 89% used it in at least one part of their purchasing process. The 2025 6sense report says 94% of nearly 4,000 buyers were using LLMs.
What to do. Use the table as a range of evidence, not a forecast for your own buyers. Ask new leads which sources helped them build their shortlist, then compare those answers over several months. A buyer may use an LLM to prepare an internal document without asking it to recommend a vendor; those are different opportunities for marketing.
Myth: Buyers who use chatbots have stopped using Google
Once a buyer starts with an AI answer, Google and the vendor website disappear from the purchase.
In G2's March 2026 survey of 1,076 B2B software buyers, 51% said they started research in an AI chatbot more often than Google, while 61% used AI search alongside Google. In G2's April 2025 survey of 1,169 B2B decision-makers, 29% said they started with AI tools more often than Google. Gartner found 51% of buyers expected misleading information from generative AI, while 49% expected it from a sales rep.
What to do. Keep your site useful to someone arriving with a vendor name or a partially formed shortlist. Make category fit, product limits, evidence and a path to contact easy to find. G2 surveyed software buyers, so its change between surveys should not be treated as the adoption rate for every industry.
The distinction also matters for consumers. McKinsey's August 2025 survey found about half of 1,927 US consumers intentionally sought AI-powered search. Adobe's survey of more than 5,000 US respondents found 39% had used generative AI for online shopping. Neither figure means those people abandoned conventional search.
Myth: AI assistants already send most website traffic
AI referrals have replaced search as the dominant source of visits, so a small AI referral line means your marketing is failing.
Ahrefs measured AI chatbot referrals at 0.28% of total traffic across 74,752 sites in March 2026. Conductor measured 1.08% across 1,215 US enterprise domains from May through September 2025. These are different panels and periods, not a controlled before-and-after comparison.
What to do. Check your own referral hosts and lead quality, while keeping the small size of these broad panels' AI referral shares in view. A buyer can read an AI answer, search for a recommended brand, and arrive through branded search or direct traffic; that plausible path will not appear as an AI referral. The separate guide on whether SEO still works with AI answers examines the click-through studies for AI Overviews.
Google also counts clicks from AI Overviews and AI Mode within the Web search type in the Search Console Performance report. Those clicks are not separated as AI referrals there, according to Google's documentation. A referral report is useful, but it cannot count every purchase influenced by an AI answer.
Myths about what the AI reads
Myth: Your website is the only source an assistant reads
Publish a polished homepage and an assistant has everything it needs to recommend your company.
McKinsey's analysis of Google AI Overview sources found that in many cases, brand-owned sites made up only 5% to 10% of sources, while publishers, user-generated content and affiliates made up more than 65% in consumer packaged goods and financial services. In an Ahrefs study of 75,000 brands, branded web mentions had a 0.664 Spearman correlation with brand mentions in AI Overviews; correlation does not establish causation.
What to do. Check whether independent sources describe your company and its category accurately, and whether your own pages explain the same details clearly. McKinsey studied Google AI Overviews in particular sectors, so its source mix should not be presented as the mix for every assistant or B2B category. Treat it as a reason to inspect the sources in your own market.

Notice the third-party inputs around the shortlist: the brand's own site is one source among several. The mix will differ by category and assistant.
For the full method of making pages eligible and useful as sources, use SEO for AI search and AEO. If your category and comparison content is thin, Iron Goo's SEO service can help build pages intended to rank and be drawn on by the sources assistants use.
Myth: Review sites no longer matter because AI summarizes them
A buyer who gets an AI recommendation has no reason to consult reviews or compare vendors.
TrustRadius reported that review-site use among its technology buyers rose from 58% to 63%, and 74% consulted customer reviews. G2's press release says 45% of its surveyed software buyers found review-site citations the most confidence-inspiring signal in an AI response. In McKinsey's consumer survey, only 6% of AI-search users named review sites as their primary source of insight.
What to do. Keep relevant review profiles accurate and inspect the recency and substance of their reviews. Publish comparison pages that explain fit and limitations in terms a buyer can check. A review site can be an input to an answer or a later verification source even when it is not the buyer's primary source.
The figures are not competing measurements of the same behavior. TrustRadius asked technology buyers about resources used during a purchase; McKinsey asked AI-search users which source they preferred for insight. G2's review-citation result comes from a review platform's press release, so read it with that commercial interest in mind.
Myths about what still converts
Myth: Buyers accept the AI recommendation without checking
The assistant makes the decision, leaving reviews, peers and sales conversations with little influence.
TrustRadius found 94% of its surveyed technology buyers fact-check AI-generated information, and 72% always or very often do so. Gartner found 69% of its B2B buyers prefer to validate AI-generated insights with sales reps; buyers used an average of seven information sources in a recent purchase, and 51% said they were more likely to encounter misleading information from generative AI.
What to do. Give buyers evidence they can verify: current product details, clear claims, customer reviews and a sales contact who can answer specific questions. Ask sales reps to record recurring discrepancies between what prospects heard and what the product actually does. Correct the source when you can identify it, and address the question directly in sales material.

The AI answer starts a shortlist; reviews, peer conversations and a sales call help the buyer verify it. Those checks give marketing and sales opportunities to support a decision.
TrustRadius found 53% of buyers spoke with a peer during their purchase, while vendor marketing collateral ranked last among the resources buyers consulted. That does not make collateral useless: it means a claim in a brochure needs support a buyer can inspect elsewhere. Make it easy for a prospect to compare that claim with a demo, a review or an answer from a rep.
Myth: Sales has plenty of time to shape the shortlist
Marketing can wait for a buyer to contact sales before explaining why the company belongs on the shortlist.
The 6sense 2025 B2B Buyer Experience Report says first contact occurred 61% of the way through the journey, down from 69%, and buyers initiated 79% of engagements. It also says 94% ranked a preferred vendor before engaging sellers, and that favorite won about 80% of the time. The report says buyers' use of LLMs has not changed their reliance on vendor content or third-party experts.
What to do. Give a prospect enough information to assess fit before a call, then make the call useful for validation. Sales can still affect the outcome, but the 6sense findings suggest that much of the work of earning a place on the shortlist happens before a seller sees the opportunity. The company sells intent data, and its report page does not state the fieldwork dates.
G2 found 69% of its surveyed software buyers chose a different vendor than initially planned because it appeared in a chatbot recommendation; one in three bought from a vendor they had not previously heard of. That is a reason to make your company understandable to sources buyers consult, not a guarantee that an assistant will recommend it.
Myth: AI traffic always converts better, and analytics shows all its influence
Every AI-referred visit is unusually valuable, and the analytics channel report captures every lead influenced by AI.
Adobe's analysis, based on more than one trillion visits to US retail sites, found AI-referred visitors were 43% less likely to convert in July 2024 and 9% less likely in its March 2025 report than other visitors. They also had a 23% lower bounce rate. Later Adobe reports say AI-referred retail visits converted better than non-AI traffic in 2026.
What to do. Measure conversion for your own AI-referred sessions and compare similar periods and outcomes. Adobe's results concern retail visits, so they cannot establish a conversion rate for a B2B sales pipeline. A low bounce rate is also not a purchase or a qualified lead.
Use a required free-text How did you hear about us? field on the contact form, and ask the same question on the first sales call. Tag answers mentioning an assistant, review site or peer, while retaining the original wording. Compare those answers with your analytics report without assuming either source tells the whole story.
Self-report: 6 AI answers / 34 answered leads = 17.6% Analytics: 1 AI-referred converting session / 40 inbound leads = 2.5%
In this hypothetical month, the company receives 40 inbound leads. Its form records 14 answers naming Google, nine naming a referral or peer, six naming an AI assistant, five naming a review site and six blank answers. Analytics records one AI-referred session that converted. The 17.6% and 2.5% figures use different denominators and measure different things: a reported influence among answered leads and an observed referral among all leads.
The gap is a prompt for investigation, not proof that five extra leads came from AI. Buyers may remember one source and omit another, and the blank responses leave part of the picture unknown. The guide to measuring marketing without a data team explains how to combine self-reported attribution with a branded-search series.
Track branded queries in Search Console each month, including known brand misspellings, and compare the series with non-branded queries. Branded search is a directional proxy for demand, not a count of AI mentions: Ahrefs found branded search volume had a 0.392 correlation with brand mentions in AI Overviews, with no causation established. Run a fixed sample of 10 to 20 category prompts monthly in ChatGPT, Gemini, Copilot and Perplexity; record whether your company appears and which sources are cited. Answers can vary between runs, so the record is a sample rather than a census.
What to do first
- Add the required free-text source question to your contact form and ask it again on the first sales call. Keep the original answers and tag mentions of AI assistants, peers and review sites.
- Check your site for the information a shortlisted buyer needs: category fit, product details, limitations and a clear way to contact sales. Give reps a place to record claims prospects want verified.
- Review the accuracy and recency of your relevant review profiles, then inspect the independent sources that appear when you ask assistants about your category.
- Build a monthly measurement sheet with branded Search Console queries, non-branded queries, leads and referrer hosts such as
chatgpt.comandperplexity.ai. Segment by the actual referrer host, not text found in campaign parameters. - Save a fixed set of 10 to 20 buyer prompts and run them monthly across the assistants your buyers mention. Record appearances and cited sources alongside the date, so a changed answer can be checked against the next sample.
Put your next steps in order
Build on what you have just read with a free ebook on semantic SEO and AI visibility. Get worked examples, decision rules, checklists and a do-this-week action in every chapter, plus a six-week plan for your site.

