How B2B buyers research vendors in ChatGPT, and what your target accounts see

Two large buyer surveys published this year describe the same shift. G2 surveyed 1,076 B2B software buyers and decision-makers in March 2026 and found that 51% start their research with an AI chatbot more often than with Google, up from 29% in its 2025 report, and that 71% use AI chatbots at some point in their research. G2 also reports that AI chatbots are now the top source influencing buyer shortlists, ahead of review sites, analyst firms and vendor websites.

Semrush surveyed 643 US B2B professionals in March and April 2026, kept 622 valid responses and analysed the 519 who use AI for work. Among them, 66% regularly use AI to research products, vendors or solutions, 92% say AI shaped their shortlist and 83% say it influenced their final vendor decision. ChatGPT was the most used tool for product research (71%), followed by Google Gemini (61%) and Microsoft Copilot (45%).

Forrester's State of Business Buying 2026 frames the same trend from the buyer's side: nearly all business buyers (94%) report using AI during their buying process, while also checking AI output against peers, product experts and analysts because AI answers can be incomplete or unreliable.

The Semrush data is useful because it breaks AI use down by buying stage and task. We mapped it to what an enterprise marketing team can influence.

The prompting behaviour matters as much as the stage. In the same survey, 61% of AI users describe their specific use case or problem, 56% ask for direct vendor comparisons and 45% include constraints such as budget, features or compatibility. A page that only says "leading platform for the enterprise" has nothing to match.

One finding should change how ABM teams plan content. Semrush reports that only 7% of AI users notice a vendor in an AI response because they recognize the brand. The factors that draw attention are a close use-case match (53%), a clear and detailed description (50%) and clear benefits or outcomes (38%).

For target accounts this is both a risk and an opening. A well-known incumbent can be described generically while a smaller vendor with a precise answer for the buyer's industry and constraint gets the attention. That is the basis of account-based GEO: writing for the questions specific segments and roles ask, not for the category in general.

AI answers rarely end the research. Semrush found that after AI mentions a vendor, 71% visit the vendor's website, 63% search Google for the company, 46% compare against alternatives and 38% check G2 or similar review sites. G2 reports that citations from software review sites are the top trust signal that increases buyers' confidence in an AI answer.

So the website still decides what happens next. When a buyer from a target account arrives from ChatGPT with a specific question in mind, the landing page has to confirm the answer quickly and offer a next step that suits early research. Demandbase reported on August 12, 2026 that monthly ChatGPT-referred visits to B2B websites in its dataset rose 303%, from roughly 645,000 in June 2025 to 2.6 million in June 2026, so this traffic is no longer a rounding error.

Visits are only part of the picture. Many buyers read the answer and never click, which is why we measure AI visibility for target accounts with prompt testing by segment as well as referral data.

For account-based programs, the shift moves part of the first impression out of your control. Before a tier 1 account sees an ad or a sales email, someone on its buying committee may already have asked an assistant for a shortlist and read a summary of your strengths and weaknesses. 6sense's 2025 Buyer Experience Report, vendor research, found 94% of buying groups ranked preferred vendors before first contact.

Sales teams feel this as prospects arriving with fixed views, sometimes based on outdated facts. Give sellers what AI tools currently say about you and your main competitors for their segments, so they can confirm what is right and correct what is wrong early in the first conversation.

1. Write down 20 to 40 prompts your target segments would ask at each stage, using the language of your sales calls, not your positioning deck.
2. Run them in ChatGPT, Gemini, Copilot, Perplexity and Google AI Mode, and record whether you appear, how you are described and which sources are cited.
3. Fix the facts on your own site first: what you do, for whom, integrations, security, pricing model and proof, each stated plainly on a crawlable page.
4. Publish use-case and comparison pages for the industries and constraints your tier 1 and tier 2 accounts share.
5. Correct or add the third-party sources the answers cite, starting with review sites and partner directories.
6. Check crawler access: OpenAI documents that OAI-SearchBot is used to surface websites in ChatGPT search results, so blocking it removes you from those answers.

Our generative engine optimization service runs this cycle monthly for enterprise B2B teams.

If you want the prompt set built for your own target accounts, talk to our team.