Your Buyers Already Asked ChatGPT About You

Your prospect read about you this morning. You weren't in the room, and neither was your website.

Here's how it goes now. A VP of Operations has a problem and a budget. Before she opens a tab with your name in it, she types "best tools for X for a 200-person company, with rough pricing" into ChatGPT. She gets back 5 names, a paragraph on each, and a table. You're in the table or you aren't. If you are, the paragraph about you was written by a model from whatever it could find.

This is normal buyer behavior now. In G2's March 2026 survey of 1,076 software buyers, 51% said they start research in an AI chatbot more often than in Google, up from 29% a year earlier. The number that should bother you more: 69% ended up picking a different vendor than they'd planned because the chatbot recommended it.

G2 sells visibility to software vendors, so it has a stake in this story. Treat the precision as marketing. The direction holds up elsewhere: Responsive's survey of 350 buyers found GenAI had overtaken search for a quarter of B2B buyers, and 80% of tech buyers used it at least as much as search.

The summary gets written without you

A model answering a buying question can't see your sales deck. It reads what's public: your pricing page, your docs, your changelog, G2 and Capterra reviews, Reddit threads, and the "X vs. Y" page your competitor published 2 years ago.

When something's missing, it guesses. Or it fills the slot with a vendor who did publish the answer.

"Contact us for pricing" reads to a model as no information. Docs behind a login read as no docs. A product page full of "AI-native platform for modern teams" reads as nothing at all, because there's no fact in it to repeat.

Ask about your own company and the misses tend to be specific. A pricing tier you retired 2 years ago, quoted as current. An integration you never built, because a competitor has it and the model blended you together. A complaint from an old Reddit thread presented as how the product works today. Each one traces back to a page somebody can read, which is the good news. Pages can be fixed.

And the buyer often never checks. When Google shows an AI summary, users clicked a regular search result on 8% of visits, against 15% when there was no summary, per Pew's tracking of 900 US adults. About 1% clicked a link inside the summary itself.

Ahrefs is blunter: an AI Overview now correlates with a 58% lower click-through rate for the top-ranking page. So the summary is the first impression. For a lot of buyers, it's also the last.

What to make readable

Give the model facts it can quote.

Publish real pricing. Even a range. "Starts at $500 a month, most customers land between $2k and $5k" beats a form every time. Enterprise deals can still be custom; say so on the page.

Open your docs. Keep admin guides and anything security-sensitive behind auth. Public docs on what the product does, what it connects to, and how setup works are sales material. Treat them that way.

Write the comparison page yourself. If buyers ask "you vs. Competitor A," somebody's answer is getting read. Make one of them yours, and be honest about where you lose. Models and buyers both notice when a comparison is rigged.

Say plainly who you're for. "For finance teams at 50 to 500 person SaaS companies who've outgrown spreadsheets" is a sentence a model can repeat. "The future of finance" gives it nothing to work with.

Get reviews where models read them. G2's own data says review-site citations are buyers' top trust signal in AI answers. 20 recent reviews beat 200 from 2021.

How to check what they say about you

Do this once a month. It takes an hour.

Open ChatGPT, Perplexity, Gemini and Google's AI Mode. Ask what your buyers ask: "best [category] for [your customer]," "how much does [you] cost," "[you] vs. [competitor]," "problems with [you]." Use a fresh, logged-out session.

Write down 4 things: whether you showed up, what it said you cost, what it said you're bad at, and which sources it cited. The citations are the useful part. They tell you exactly which pages are doing the talking for you.

Then fix the source and leave the model alone. If Perplexity keeps citing a 2023 Reddit thread about a bug you fixed long ago, publish a clear changelog entry and a support page that says so.

Running the queries is the mechanical 70% of the 70/30 split, and an agent can do it every week. Deciding what you'll publish about pricing, and admitting who you lose to, is the 30% that needs you.

What not to buy

Someone will pitch you "AI SEO," "GEO" or "LLM optimization" on a monthly retainer. Some will promise placement in ChatGPT answers. No agency controls that. The models don't sell placement, and their behavior shifts every few months.

Be skeptical of anyone selling guaranteed citations, or a plan to mass-produce hundreds of thin pages for AI crawlers to find. The first is a promise they can't keep. The second can make your brand look like spam to the same models you're courting.

The traffic is real, and small. Microsoft Clarity found AI referrals were under 1% of traffic across 1,200+ publisher sites, yet converted to sign-ups at 1.66% against 0.15% for search. Few visitors, but warm ones. That's worth a week of work on your pricing page and docs. A retainer can wait until you've done that week.

My advice on these proposals: before you sign anything monthly, publish your pricing, open your docs and run the monthly check above for 2 months. If the answers still look wrong after that, at least you'll know exactly what you're paying someone to fix.

The model is going to describe you either way. Give it something true to say.