How 200 B2B buyers actually use AI to pick software

Close-up of a keyboard with a prominent 'buy' button

Watching a software buyer evaluate your product in real time is nearly impossible. You can run win/loss interviews, check demo request patterns, and monitor G2 traffic, but you rarely get to see the actual conversation that shaped their shortlist.

Kevin Indig analyzed survey results from 200 participants across 13 questions about their AI interactions during software buying decisions, using data from Wynter. Participants shared their first prompts and follow-up sequences. The result is one of the clearest windows yet into how B2B buyers actually use AI to evaluate vendors.

The Methodology

The survey covered 200 responses across 13 questions. All participants had SaaS or software industry profiles. Demo and signup shares were 35.2% among Claude users (50 of 142), 31.2% among ChatGPT users (39 of 125), and 33.8% among Gemini users (26 of 77). All three groups had a median of 3 to 5 follow-up messages.

A few limits worth keeping in mind: the survey is self-reported, prompt quotes may be recalled rather than copied verbatim, and a reported demo or signup does not confirm a completed purchase. The data also cannot establish whether an AI-introduced vendor was ultimately chosen.

AI Reopens the Shortlist

Before the AI chat, 158 of 200 respondents (79%) already had at least one vendor in mind. They weren’t starting from scratch. They had done some research, maybe asked a colleague, maybe run a quick Google search. They came in with opinions.

Then AI added to the list. 99 of those 158 respondents (62.7%) added a new vendor to their shortlist through AI. One engineer evaluating AI visibility tools came in with 2 or 3 candidates and walked out with more. After getting suggestions that didn’t fit, they pushed back in a follow-up: “I care about product level insights, not just brand level.”

a person holding a cell phone in their hand

Among the 40 respondents who started without any vendors in mind, AI added a shortlisted option for 33 of them (82.5%). But only 40% of that group reported choosing a vendor, compared with 69% of the group that already had candidates. The group that arrived with context moved further through the buying process.

The operator implication: if your brand isn’t surfacing in AI answers before buyers are ready to buy, you’re not just missing awareness. You’re getting cut from a shortlist you were never on.

The Follow-Up Questions That Actually Decide Deals

Buyers don’t just ask one question and close a tab. 41.5% of participants sent 3 to 5 follow-up messages after their first prompt. One-third sent 6 or more.

The most common follow-up topics in this sample:

  • Pricing: 57.5% asked about it
  • Vendor comparisons: 54.5%
  • Integrations: 46%
  • Setup and implementation: 40%

One buyer’s chosen vendor appeared in the first AI answer, but they still followed up with: “Eliminate any vendors that don’t have an API integration with Marketo.” Showing up first doesn’t mean you survive the filter round.

Respondents who sent 1 to 2 follow-ups covered an average of 2.4 topics. Those who sent 10 or more covered an average of 5.0 topics, including asking AI to recommend a single vendor outright. Longer conversations covered more ground.

On demo and signup activity: respondents who sent 6 or more follow-ups reported a demo or signup more often than those who sent 1 to 5 (41.8% vs. 32.3%). The rate peaked at 6 to 10 follow-ups, then fell in the 10+ group. The survey can’t establish that longer conversations caused more action, only that they correlated.

What Happens After the Chat

AI didn’t replace the rest of the research process. It extended it. After the AI chat:

  • 145 of 200 respondents (72.5%) visited vendor websites
  • 96 (48.0%) ran a Google search
  • 62 (31.0%) asked peers
  • 58 (29.0%) searched Reddit
  • 46 (23.0%) checked G2 or other review sites

The AI conversation fed into other channels, not away from them. Buyers who used AI still Googled, still visited your site, and still asked for social proof.

MacBook Pro on top of brown table

Platform fragmentation also matters here. 67.5% of participants used more than one AI platform for their research. Claude led at 71% usage, ChatGPT came in at 62.5%, Gemini at 38.5%, and Google’s AI Overviews or AI Mode at 15%. A buyer might ask Claude for an initial shortlist, then run the same question through ChatGPT to cross-check it.

What This Means for Solo Operators and Small Teams

If you sell software, even as a solo founder with a simple SaaS product, this data changes how you should think about visibility.

First, being mentioned early matters but doesn’t guarantee survival. 78 of 126 respondents who chose a vendor (62%) picked one from the first AI answer. But 75 of those 78 still sent follow-up questions. Early mention earns a seat at the table, not a closed deal.

Second, your pricing page, integration documentation, and setup guides are now part of your AI discoverability stack. Those are the exact topics buyers ask about in follow-ups. If AI can’t accurately answer those questions about your product, you get filtered out.

Third, brand awareness before the research phase matters more than it did. Buyers who already had vendors in mind were more likely to choose a vendor (69% vs. 40%) and more likely to report a demo or signup (36.1% vs. 25.0%) than those who started cold. Getting into someone’s mental shortlist before they open an AI chat is still valuable.

Indig frames it through a pipeline: retrieval, then citation, then trust. Getting retrieved by the AI isn’t enough. Getting cited isn’t enough either. The buyer still has to trust what they see and act on it.

Stay on top of AI & Automation with BizStack Newsletter
BizStack  —  Entrepreneur’s Business Stack
Logo