G2: 51% of B2B Buyers Now Start Research in an AI Chatbot

What happened

In its new report The Evaluation Maze, G2 surveyed more than 1,000 B2B software buyers and decision-makers and interviewed 50 sales and marketing leaders. The headline finding: 51% of buyers now open ChatGPT, Claude, Gemini, or Google AI Mode before they open Google search, and 80% have used one of those tools to source software recommendations within the last two years. The report describes a discovery phase that has collapsed from what used to be several hours of website browsing and analyst-report reading into a single prompt. That shift compounds an earlier finding from Gartner that 69% of B2B buyers now validate the AI-generated shortlist with a rep, and from Forrester that 70 to 80% of the buying journey is completed before sales contact. In practice: the AI answer sets your shortlist, the rep gets a call only to confirm what the AI already said.

Why 51% of B2B buyers starting in an AI chatbot matters now

If half of buyers no longer touch a search engine until after they have three candidate vendors in mind, the traditional funnel top has moved off Google entirely. SEO teams optimizing for organic rankings on category keywords are now optimizing for a shrinking slice of buyer discovery. The category page that used to be the entry point is now a validation page — buyers land on it after ChatGPT already named you, or they never land on it at all. Content investment has to shift accordingly, and pricing, comparison, and category framing pages need to be legible to LLMs first, humans second.

1. 01
   Audit whether AI chatbots name you for your category

   Open ChatGPT, Claude, Gemini, and Google AI Mode. Ask each: 'What are the best tools for [your category]?' If you are not named in the first three answers on any of the four engines, you have a discovery-phase gap. Log the wording of the answers — that is now the copy your prospects read before touching your site.
2. 02
   Rewrite pricing and comparison pages for LLM extraction

   The pages LLMs cite for shortlist decisions are pricing, comparison, and integration pages. Make each one machine-legible: use tables, structured H2s per feature, and price numbers in plain text rather than images. Add a one-paragraph summary at the top of each page that answers 'what is this and who is it for' in fewer than 40 words.
3. 03
   Prepare reps for the validation call, not the demo

   First calls are now confirmation calls. Train reps to ask 'which AI tools did you use to find us and what did they say' in the opening two minutes. That single question surfaces the shortlist framing you were compared against, and lets the rep correct any AI hallucinations about your product before pricing enters the conversation.

By the numbers: 51% of buyers start research in an AI chatbot (G2 2026). 80% have used ChatGPT or Google AI Mode to source software recommendations in the last two years. 69% of buyers validate the AI answer with a sales rep before committing (Gartner).

What to do this week

Run your own category through ChatGPT, Perplexity, and Google AI Mode this week and record whether you are named, how you are described, and which competitors appear first. Send that transcript to your sales team along with a two-line rebuttal for every mischaracterization. Then pick the one comparison page on your site that most needs restructuring for LLM extraction — usually the '[you] vs [top competitor]' page — and rewrite the first 40 words to answer 'what is this and who wins for what' in a single scannable paragraph.