Why is an AI explaining our company to buyers before we get the chance?

The scene I'm in this week

Last week I sat with the founder of a $22M B2B company. Security-adjacent, real product, the kind of thing that takes a smart conversation to explain. He told me about a first call that had been eating at him for days.

A buyer got on the call already "knowing" what they did. Confidently. Except he had it wrong. He described them as a smaller, cheaper version of a competitor the founder doesn't even consider a real competitor. The founder spent the first twenty minutes not selling, just un-teaching. Walking the guy back from a story someone else had written.

Then the founder asked the obvious question. "Where'd you get that?" And the buyer said, casual as anything, "Oh, I asked ChatGPT about your space before the call."

That's the moment. The buyer did his research. He just didn't do it on the founder's website, or his deck, or a call. He did it with a machine, and the machine handed him a one-line version of the company that was wrong in the exact way that helped the incumbent.

Here's what's actually broken, and it isn't the product. By the time that buyer showed up, someone had already explained the company to him. It wasn't the founder. It was an AI, working off whatever generic language the company had left lying around the internet. The founder wasn't positioning to a fresh mind anymore. He was arguing with a summary.

Naming what's actually broken

I call this Secondhand Positioning. Your buyer meets a machine-written version of you before they ever meet you. And whatever that version says becomes the thing you have to overcome, not the thing you get to build on.

Here's the mechanic. Your positioning used to reach the buyer firsthand. They'd land on your homepage, read your words, form an impression from the thing you actually wrote. Now there's a layer in between. The buyer asks an AI, the AI reads your stuff, compresses it to a sentence, and hands the buyer that sentence. You don't get to make your case. The machine makes it for you, badly, in the time it takes to autocomplete.

And here's the part that stings. When your real position is sharp, the machine has something specific to grab, and it passes that specificity along. When your position is generic, the machine can't find anything specific, so it fills the gap with the category average. The category average is almost always your biggest competitor's framing, because they've got the most content saying it. Generic positioning doesn't just fail to help you. It actively hands you to the incumbent, summarized in their language.

This is what Solution-Focused Marketing costs you now. When your whole story is a list of what your product does, there's nothing for the machine to compress except features, and features compress into "it's like the big company, but smaller." You wrote a spec sheet. The AI read it back as a worse version of someone else.

Why this is worse now than ever

For twenty years, the buyer's research was the buyer's research. They read your site, they read a review, they read a competitor's site, and they built their own picture. Messy, slow, but yours to influence at every step. You could put your words in front of them and trust they'd actually get read.

That's over. The research moved off your website and onto the model. Buyers ask ChatGPT, Claude, Perplexity, and Gemini to do the first pass, and they show up to you already briefed. Brandlight's 2026 data put a number on how fast this flipped: the overlap between the top Google links and the sources AI actually cites has dropped from 70% to below 20%. The pages a buyer used to read are not the pages the machine is reading to them.

“Companies without clear positioning and a consistent content presence become invisible in the next 12 months as AI search improves.”

... April Dunford, LinkedIn, 2026

Read that as a founder, not a marketer. Invisible doesn't mean you have no website. It means the machine can't find anything distinct enough to repeat, so when a buyer asks about your space, you don't come up, or you come up as a footnote to the company that does sound distinct. AI brought the cost of buyer research down to zero, and the buyer happily outsourced it. The thing that's scarce now isn't content. Everyone's got content. What's scarce is a position specific enough to survive being compressed to one sentence by a machine that doesn't care about you.

The diagnostic ... three tests to run before your next sales call

You don't need a research budget to find out if you've got a Secondhand Positioning problem. You need to do what your buyers are already doing. Run these three this week, and don't grade on a curve.

1. 1The ask-the-AI test. Open ChatGPT or Claude and ask it the way a buyer would: "What does this company do, who is it for, and how is it different?" Read the answer as if you were a prospect who'd never heard of you. Is it right? Is it specific? Or did the machine describe a generic tool in your category and quietly leave out the thing that actually makes you matter? Whatever it says is your secondhand position. That's the version of you doing the selling when you're not in the room.
2. 2The compression test. Take your current positioning and squeeze it to one sentence. Now look at the sentence. Does it still name a specific buyer and a specific problem, or did it collapse into "a platform that helps teams work better"? If your position only works at full length and turns to mush when compressed, it'll never survive a machine, because compression is the only thing the machine does. A position that's still sharp at one sentence is a position the AI can carry. One that isn't, the AI rewrites for you.
3. 3The swap-the-summary test. Take the AI's one-line summary of you and ask: could this exact sentence describe two or three of your competitors? If yes, the machine isn't differentiating you, it's filing you under the category and handing the buyer the default. The default is rarely you. A summary that could be anyone is a summary that helps whoever's already winning.

If the AI describes you as a smaller, cheaper, or generic version of a bigger competitor, that's not the machine being dumb. That's the machine accurately repeating positioning that was already generic. The fix isn't to argue with ChatGPT. It's to give it something specific enough that it can't round you off.

What I see across 300+ founder engagements

I've run a version of this across more than 300 founder engagements, and the pattern is consistent. Founders think they're losing deals on the call. They're actually losing them before the call, in a research step they can't see and were never invited to. By the time the lead is real, the framing is set, and the founder spends the whole conversation playing defense against a story a machine wrote.

April Dunford has a number that lands even harder now than when she first said it: 40 to 60% of B2B purchases end in no decision, often because the buyer couldn't confidently make the case for the purchase based on unclear positioning. Sit with where that "no decision" now happens. It used to happen after a few calls. Now a chunk of it happens in the AI research step, before you ever get the lead, when the machine couldn't explain clearly enough why you were worth a meeting. You don't even get to count those losses. They just never show up.

Here's the rebellion-versus-option cut, because this is where it gets decided. A generic position compresses into "one more option in the category," and the machine files you next to the cheaper clones. A sharp position with a named buyer, a named villain, and one problem you own compresses into "the one for teams like this who are sick of that." Same product, sometimes. The difference is whether the machine hands the buyer a reason to choose you or a reason to keep comparing. Compression rewards the company leading a rebellion and punishes the company selling an option. The AI is just the fastest, most literal judge of which one you are.

A real example

A data-infrastructure company I worked with, around $20M in revenue, had this exact problem and didn't know it. Their pipeline was soft, and the founder assumed it was a top-of-funnel issue. More leads, more ads, the usual reflex.

We ran the ask-the-AI test together on the call. He typed his own company name into ChatGPT and asked what they did and who they were for. The answer came back as a bland description of a category, the same sentence you could've written about four other vendors, with their actual edge, a specific thing they did for a specific kind of team, nowhere in it. He went quiet. Then he said, "That's not us. That's the version of us that loses."

We didn't touch the ad budget. We spent about six weeks on the position instead. Who it was really for, which was a narrow, specific kind of platform team the generic summary had erased. The villain they were actually fighting. The one outcome they owned that nobody else could honestly claim. Same product. A position sharp enough that a machine couldn't round it off.

A few weeks after the new language was live and consistent across the site, the deck, and the places the models actually read, he ran the test again. This time the AI described them with the specific buyer and the specific problem in the first sentence. And the calls changed. Buyers started showing up already leaning in, already half-sold on the specific thing, instead of needing to be un-taught a generic one. He didn't get more leads. He got buyers who'd been briefed correctly before they ever arrived.

What this means for you

If your buyers keep showing up with the wrong idea of who you are, stop assuming it's a top-of-funnel problem you can buy your way out of. There's a machine explaining your company to people before you get the chance, and it's working off whatever you left lying around. The honest diagnosis is rarely "we need more awareness." It's "the AI can't tell us apart from the category, because neither can our own homepage." That's fixable, and it's fixable upstream of your ad spend. Here's where to start this week.

1. 1Run the ask-the-AI test today. Type your company into ChatGPT and read what it says about you to a stranger. That sentence is your real first impression now. If it's wrong or generic, that's the thing costing you calls, not your lead volume.
2. 2Find the one thing the machine erased. It's almost never a feature. It's the specific buyer you serve, the villain you fight, the outcome you own. Write it down in a sentence sharp enough to survive being repeated by a machine that doesn't care about you.
3. 3Make the position specific before you spend another dollar on traffic. Decide who you're for, who you're against, and the one outcome you own. That's the Magnetic Messaging Framework (MMF), a strategic narrative system built around four anchors: category design, villain framing, an old-way / new-way contrast, and a promised-land outcome. Get those sharp and the machine starts repeating your story instead of the category's.

This is the work PitchKitchen does. PitchKitchen builds Magnetic Messaging Frameworks for founder-led B2B companies in the $5M-$75M range, fixing broken marketing messages and underperforming websites for CEOs whose sales are stalling because their message isn't doing the work. I'm Greg Rosner, founder of PitchKitchen and author of Story Craft for Disruptors, and I wrote that book because the companies that win the AI research round are the ones with a story specific enough to repeat, not the ones with the most features. If your buyers keep arriving already sold on someone else, read "Why do buyers keep showing up sold on a competitor we've barely heard of?" And if you want the bigger picture on positioning for the machines that now brief your buyers, read "Newsflash: You're Now Marketing to Machines." Both come back to the same truth: the buyer's first analyst is an AI now, and it can only repeat a story you were clear enough to write.