The short version
Before your first call, most B2B buyers have already asked an AI assistant who to consider. Half of software buyers now start their research in a chatbot more often than in Google.
The AI answer is now the biggest single influence on who makes the shortlist. If you're not in it, you're often not in the deal at all.
AI assistants name vendors for mechanical reasons: a checkable number on the page, a plain statement of who you serve, and listings on directories and review sites. Most firms have none of these.
Buyers still come to the first call to check what the AI told them. Plan the call around that, not around a pitch they've already read a summary of.
A head of operations needs a new accounting partner. Before she visits a single website, she opens ChatGPT and asks for the best options for a firm her size. Then she asks it to compare the top three, what each costs, and what to watch out for. Twenty minutes later she has a shortlist, a view on pricing and a list of questions.
She books one call. By the time you meet her, the AI has already introduced you, or left you out.
What an AI research session looks like
The scale of this is no longer a forecast. G2's 2026 survey of 1,076 B2B software buyers found that 51% now start their research with an AI chatbot more often than with Google, up from 29% a year earlier, and 71% rely on AI chatbots at some point in the process. Buyers of services are moving the same way. Gartner's 2026 survey of 646 B2B buyers found 45% had used AI during a recent purchase.
A typical session runs in four steps:
Ask for options. "Who are the best providers of X for a company like mine?" The answer usually names three to five firms.
Compare them. Comparing vendors' strengths and weaknesses is the most common use of AI in software research, according to G2.
Pressure-test the favourites. Pricing, risks, what other customers say, what to ask on a first call.
Go deeper on one or two. Some buyers run a deep research report or a long thinking-mode prompt before they contact anyone.
Which assistant they use matters less than you'd think. ChatGPT is the most used for B2B software research by a wide margin, but Perplexity, Gemini, Copilot and Claude all work the same basic way: when the question needs current facts, they search the web, read the pages that answer it and summarise what they find.
The shortlist forms before you know the deal exists
The AI answer isn't one input among many. It's now the biggest one.
It also changes outcomes. In G2's research, 69% of buyers said an AI chatbot led them to a different vendor than they first planned, and a third picked a vendor they had never considered before. Your own website ranks below review sites, level with market research firms, as an influence on the shortlist.
None of this shows up in your reports. The buyer never clicked anything, so the research session is invisible to your analytics. All you see is a "direct" enquiry, or nothing at all because you weren't named. That's the same dark social problem that undercuts every multi-touch attribution report, with a new and bigger source.
What we found when we asked ChatGPT 124 buyer questions
For a B2B accounting client, we ran 124 buyer questions through ChatGPT across six markets between July and September 2026. Each question was the kind a founder asks before hiring a firm, run in a fresh, unpersonalised chat with the client never mentioned in the prompt.
Three things stood out.
The vendor seat is usually empty. In 91 of 124 answers, ChatGPT named no service provider of any kind. In 49, it cited no source at all and answered from its own training. In most categories, nobody owns these answers yet.
Firms got named for mechanical reasons. The providers that did appear had published a checkable figure in a plain sentence, such as how many clients of a certain type they serve. Our client's one mention came from the only page on its site with a specific number on it.
One word changes which market the buyer is shown. Asking for a "bookkeeping service" and asking for an "accounting firm" returned completely different lists, with no overlap. The words on your pages have to match the words your buyers type.
This was one client in one category, so treat the exact numbers as one data point. The pattern behind them is the useful part, and it matches what we see across other accounts.
How to get into the AI shortlist
None of this needs a new team. It needs your existing pages to say things a model can use.
Put checkable numbers in plain sentences. "We work with 120 law firms across the US" gets quoted. "We work with leading firms" doesn't. Models quote what is quotable.
Say exactly who you serve and what you work with. Name the industries, company sizes, markets and platforms in plain text on your own site. When a buyer asks "which firms work with HubSpot", the answer goes to whoever wrote that sentence down.
Get listed where models look. Review sites, partner directories for the platforms you work with, and industry associations. G2's buyers rank review sites second only to AI chatbots as a shortlist influence, and the assistants draw on them too.
Answer the questions nobody answers. If an AI answers a common buyer question with no source, the first credible page that answers it well has no one to displace.
Use your buyers' words, not your own. Check which term returns your competitors, then use that term on your service pages.
Let the crawlers in. Check that your robots.txt allows the AI crawlers, such as GPTBot and PerplexityBot, and that your pages render as plain HTML they can read.
Our guide to answer engine optimization covers each of these in more depth. And because AI answers only name firms buyers can find evidence for, this work sits in the creation half of demand generation, not the capture half.
// The Rule
If a page has no number a model can quote, it has nothing a model will quote.
Check where you stand in 30 minutes
You don't need a tool to get a first read on your LLM visibility:
Write down 15 to 20 questions your buyers ask before they hire someone like you. Use their words, not yours.
Run each one in a fresh, temporary chat with personalisation turned off, so your own history doesn't skew the answer. Never mention your company in the prompt.
For each answer, note two things: were you named, and was your site cited as a source? Also note which competitors and which websites were.
Keep the questions exactly the same and run them again every quarter. Changing the wording makes before-and-after comparisons meaningless.
The competitors and sources that keep appearing tell you exactly where to get listed and what kind of page to publish.
Change your first call to match how buyers arrive
Buyers who research with AI don't skip the call. They use it differently. Gartner found 69% of B2B buyers turn to sales reps to validate what AI told them, and buyers are split almost evenly on whether generative AI or a sales rep is more likely to mislead them. They arrive with a summary of you they half-trust and want to check.
So the discovery call should change:
Start by asking what they've already looked at. "What have you read about us, and what did the AI tools tell you?" You'll hear the version of you they're comparing.
Correct the record early. If the AI got your pricing, your focus or your size wrong, fix it in the first ten minutes, before it shapes the rest of the conversation.
Skip the overview they've already seen. Spend the time on what no summary can give them: specific outcomes for clients like them, how the first 90 days work, and what it costs.
Expect more people on the call. Each member of the buying committee may have run their own AI research, and their summaries won't match. Part of your job is reconciling them.
Want to know what AI tools say about you before your next buyer asks? In a 30-minute call we'll run your buyers' questions live and show you who gets named instead of you.
Frequently asked questions
How do B2B buyers use AI to research vendors?
B2B buyers use AI assistants such as ChatGPT to list possible vendors, compare their strengths and weaknesses, check pricing and risks, and prepare questions before contacting anyone. G2's 2026 research found 51% of B2B software buyers now start their research in an AI chatbot more often than in Google.
How does a company get recommended by ChatGPT?
AI assistants tend to name companies whose own pages state specific, checkable facts in plain sentences, such as the number and type of clients served, the platforms supported and the markets covered. Listings on review sites, partner directories and industry associations add further evidence the assistant can draw on.
Do AI shortlists actually change which vendor a buyer chooses?
Yes. In G2's 2026 survey, 69% of software buyers said an AI chatbot led them to choose a different vendor than they first planned, and about a third chose a vendor they had not previously considered.
How can a business check whether AI assistants mention it?
The simplest method is to run 15 to 20 real buyer questions in fresh, unpersonalised AI chats without naming the company, then record whether it was named and whether its website was cited. Running the same questions each quarter shows whether visibility is improving.
Do buyers who research with AI still want a sales call?
Most do. Gartner's 2026 research found that 69% of B2B buyers turn to sales reps to validate AI-generated insights, which makes the first call less about introducing the company and more about confirming or correcting what the buyer has already read.