Your Buyer Asked ChatGPT Before They Answered Your Email
TL;DR. G2's March 2026 survey found that 51% of B2B software buyers now start their research with an AI chatbot more often than with Google. So your outbound lands with someone who will check you in an AI answer before they reply, or instead of replying. Write every email to survive that check: one claim they can verify, nothing your website contradicts, and a reason to talk to a person that the AI can't answer.
How B2B buyers use AI search, in numbers
A year ago, AI search was something buyers tried. Now, by G2's count, it's where half of them start.
- Half of buyers start in a chatbot more often than in Google. G2 surveyed 1,076 B2B software buyers in March 2026. 51% now start their research with an AI chatbot more often than with Google, up from 29% in G2's 2025 report (G2, April 2026).
- Nearly everyone uses it. Forrester reports that 94% of business buyers now use AI when they buy, and that generative AI and conversational search are now named "a more meaningful or important source of information than any other source, far outpacing vendor websites, product experts, and sales" (Forrester, January 2026).
- They still come to you late. 6sense's 2025 Buyer Experience Report puts the buyer's first contact with a seller at 61% of the way through the journey, and found buyers started 79% of the engagements (6sense, November 2025).
Add it up. Most of the buying happens before anyone talks to you, and more and more of it happens inside an AI answer.
What buyers do with the AI's answer
They act on it. In G2's 2026 report, 69% of buyers said an AI chatbot surfaced information that led them to choose a different vendor than they expected (G2, The Answer Economy).
They don't fully trust it either. In the same G2 report, 64% said they run into inaccuracies often or very often. And Gartner found that 69% of B2B buyers prefer to validate AI-generated insights with sales reps (Gartner, May 2026).
So the AI is the first filter, and a person is the final check. Your outbound sits in between.
Your cold email is now a prompt
Say a VP of Operations gets your email on a Tuesday morning. Before she decides whether to reply, she pastes your company name into ChatGPT and asks what you do and who you compete with. Then she opens your website in another tab.
Three things happen in about a minute:
- The AI describes you, accurately or not.
- Your website confirms your email, or contradicts it.
- She decides whether you're worth thirty minutes.
Your email wasn't skipped and it wasn't answered. It was checked. That's the new job of a cold email: be worth checking, and pass the check.
What changes in the email
Make one claim she can verify
Not five. One specific, checkable point about her business or yours.
Vague claims fail the check because there's nothing to check. "We help operations teams work smarter" returns a shrug from any AI. "Running two warehouses means two sets of carrier contracts to reconcile. Is that on your team or on finance?" rests on one fact she can confirm in ten seconds, and it ends on a question only she can answer.
That's also why personalization has to be real. A mail-merge line about her "impressive growth" doesn't survive a reader who is checking. (More on that difference.)
Say nothing your website contradicts
In a separate 2025 survey, Gartner found 69% of B2B buyers report inconsistencies between what's on a sales organization's website and what its sellers tell them (Gartner, June 2025).
In the old world, that gap cost you on the second call. Now it costs you before the first reply. If your email promises something your homepage doesn't say, or names a customer type your site never mentions, the check fails and she moves on.
Give her a reason only a person can answer
The AI can tell her what you do, roughly. It can't tell her what you'd do about her specific problem, or why you wrote to her this week and not last quarter.
That part is the reply. A question about her situation, built from something real about her company, is the one thing an AI answer can't settle for her.
Don't make claims you can't source
If the AI can't find your "3x pipeline" anywhere but your own email, that's a credibility problem, not a hook. Every fact in an outbound email should trace to a source the reader could find. We hold our own work to that rule: every fact about a prospect that lands in one of our deliverables has to trace to a primary source or to the prospect's own confirmation.
What changes off the email
Be in the answer. G2 found 85% of buyers think more highly of a vendor when AI includes them in an answer, and one in three bought from a vendor they'd never heard of before (G2, April 2026).
You don't need a new acronym to act on that. Three plain moves:
- Say plainly what you do and for whom, on the pages buyers and AI tools read first. If a person can't summarize your homepage in one sentence, neither can a model.
- Keep the same facts everywhere. Your site, your LinkedIn, your review profiles and your outbound should agree on what you sell, who it's for and what makes you different.
- Earn mentions outside your own site. Reviews, comparisons, podcasts and customer posts are what an AI answer leans on when it describes a vendor.
If you want help with this, it's part of our GTM consulting: how your company shows up when buyers ask ChatGPT, Perplexity and Google's AI Overviews, and the plan to fix it.
What to stop doing
Some outbound habits were always weak. A buyer who checks your email in an AI answer makes them fail faster.
Name-dropping customers the buyer can't find. If your email says "teams like Acme and Globex use us" and neither appears on your site or in any review, the check turns up nothing, and nothing reads as made up.
Flattery dressed as research. "I was impressed by your growth" says nothing she can check and nothing she didn't know. It's the opposite of a verifiable claim. (How we research prospects.)
Follow-ups that repeat the first email. If the second and third emails restate the first, each one is a fresh chance to fail the same check. Each touch should bring a new, checkable reason to talk.
A website that lags your pitch. If your outbound sells a use case your homepage doesn't mention, the buyer's first lookup will tell her you're something else. Fix the page before you scale the email.
None of these are new mistakes. They just used to cost you a reply. Now they cost you the buyer's trust before she's met you.
What doesn't change
Buyers still want a person at the end. The 69% in Gartner's May 2026 survey who validate AI insights with a rep are telling you where the conversation still happens.
AI shortened the path to the shortlist. It didn't remove the conversation. Cold email's job is to be the reason that conversation happens with you, and not with whoever the AI mentioned first.
The bottom line
Write every outbound email as if the reader will fact-check it in an AI answer before replying. More and more, they will. The emails that win are specific enough to verify, consistent with everything else about you, and built around a question only a person can answer. (Why more AI email won't help.)
Frequently asked questions
Do B2B buyers really use ChatGPT to research vendors?
Yes. G2's March 2026 survey of 1,076 B2B software buyers found 51% start research with an AI chatbot more often than with Google (G2, April 2026).
Does AI search make cold email less useful?
It changes the job. The email is now often the reason a buyer looks you up. An email that's specific and verifiable makes the lookup work in your favor; a vague one gives the buyer nothing to confirm.
How do I find out what AI tools say about my company?
Ask them the questions your buyers would ask: what you do, who you compete with, who you're best for. Then compare the answers with your website and your outbound copy, and fix whatever disagrees.
Want to see researched-grade treatment on one of your accounts?
Pick an in-market account that's currently slated for the template. Send me the company name and I'll show you what the research pass produces. No charge, no obligation.
Book a 30 minute walkthrough