Why does ChatGPT recommend our competitor when our product is better?

The scene I'm in this week

This week a founder told me something almost in passing, and it's been rattling around my head since. He runs a $30M B2B SaaS company. Real product. Customers who renew. And he'd just lost a deal he was sure was his.

Not to a bigger competitor. To a smaller one he'd barely heard of. When his rep finally got the buyer on a call, the shortlist was already set. Three names. His wasn't one of them.

He asked where the shortlist came from. The buyer shrugged and said, "I asked ChatGPT who the best vendors were for our situation, then asked a few follow-ups."

Read that again. The buyer built the finalist list inside a chat window before a single human at his company knew the deal existed. No bounce, no lost-deal alert, no form abandoned. His name just wasn't in the room.

That's the quietest way a B2B company loses now. And he thought he had a lead-gen problem. He doesn't. He has a being-describable problem.

Naming what's actually broken (the SEO world just said it out loud)

Here's the part that should make every founder sit up. The people who profit from the old game just told you the game changed.

At WordCamp Europe this year, four SEO practitioners sat on a panel and agreed on something that breaks twenty years of their own playbook. Pam Aungst Cronin of Pam Ann Marketing said it flat out.

“Brand is the new backlink. That is what really you need to think about.”

... Pam Aungst Cronin, SEO panel, WordCamp Europe 2026

Jovana Smoljanovic Tucakov argued you have to run SEO and brand strategy as one motion. Alex Moss of Yoast pushed on clarity and reducing ambiguity for the machines. David Cuesta made the case that smaller brands win by earning consensus through consistency.

Sit with the source for a second. These are link people. The discipline that spent two decades teaching you that authority equals who points at you just told a room full of marketers that links are losing to brand. When the people selling the old moat tell you the moat moved, the debate's over.

Why this is worse now than ever

Here's the shift underneath all of it. AI brought the cost of content to zero. Anyone can generate a hundred posts this afternoon. Volume isn't the moat anymore. Perspective is. A clear, specific point of view is the one thing the machine can't manufacture out of thin air, because it has to read it somewhere first.

And AI search collapsed the funnel. Buyers used to scan ten blue links and click around. Now they ask one question and get one answer, or a tidy list of three names. There's no page two to claw onto. You're in the answer or you're invisible, and the loss is silent because it happens in a conversation you were never part of.

That changes what "getting found" even means. Ranking fourth on Google still earned you a click. Being the model's fourth-favorite earns you nothing. The recommendation is winner-take-most, and the model picks its winner on a signal most founders have never deliberately built.

Run this diagnostic on your own homepage

The panel got the what right and stopped one layer short of the why. Brand to a language model has nothing to do with your logo or your color palette. The model can't see any of that. What it can see is consensus.

Before ChatGPT recommends you, it triangulates. It reads your homepage, your LinkedIn, the Reddit threads, the comparison pages, the review sites. Then it asks one question of all of it: do these sources agree on who this company is for and what problem it solves? If yes, there's something safe to repeat, and you get named. If they contradict each other, there's nothing it can say with confidence, so it reaches for the competitor it can describe in one clean sentence.

Backlinks measured who vouched for you. Brand measures whether the internet can agree on what you actually are. That's a story problem, not a technical one. Run the Three Questions Test on your own homepage, the way a machine would:

1. 1Who do you serve? Stated specifically enough that a model could name the buyer out loud. "Modern teams" and "growth-stage companies" don't count. "Revenue-cycle leaders at 200-bed hospital systems" does.
2. 2What problem triggers the search? The actual pain that sends someone looking, not your feature list. If your page describes what your product is before it names what breaks without it, the machine has nothing worth repeating.
3. 3What's your point of view? The opinion that makes you describable as something other than "another platform." A model can only recommend you for a category it can place you in. Give it one.

What I see across 100+ B2B homepages

We run this diagnostic constantly, and the failure mode is almost always the same. The homepage says "AI-powered platform for modern teams." The comparison sites, written by people who actually had to categorize the company, call it a legacy tool with a fresh coat of paint. The Reddit thread describes it as the thing one annoyed customer complained about. Three sources, three different stories, zero consensus.

We score that homepage on what we call a NarcScore, which measures how much the page talks about the company versus the customer's problem. High-NarcScore pages read like a company talking to a mirror. They list what the product does and never name who it's for or what breaks if you don't use it. A mirror gives the machine nothing to repeat, because there's no buyer and no problem in the copy, only features. And features are the one thing every competitor in the category also claims.

That's the AI-Parmesan trap. Sprinkling "AI-powered" on a page that was already unclear is parmesan on a weak narrative. It makes the dish look finished and changes nothing underneath. The model reads right through it and recommends the company whose story it can actually say back.

A real example

One healthcare-IT company we worked with had the classic setup. A hero section full of platform language, AI citations going to competitors, and no engine able to say who they actually helped. We didn't touch their schema first.

We rewrote the hero to name a specific buyer and a specific trigger, the kind of sentence a real customer would use to explain why they went looking in the first place. Then we made that same story consistent across the surfaces the model reads most. Same answer on the homepage, the LinkedIn page, the comparison content.

Within about a month, Perplexity started citing them by name, with brand attribution, on queries they'd been invisible for. That's directional, not a guaranteed number, and how fast it moves depends on how crowded your category is. But the mechanism is reliable. The model started recommending them because it finally had a coherent story to repeat.

What this means for you

The SEO industry will now sell you "AEO services," and most of it will be schema markup plus more content. That's link-building wearing a new costume. It treats a clarity problem as a technical one, and it'll cost you a quarter to find out it didn't move the needle.

The real work sits upstream of the tools. One coherent story, true to what you actually do for a specific buyer, codified so every page and every person and every third-party surface repeats it the same way. That's the Magnetic Messaging Framework, and the reason we then build it into an AI Brand Twin is so the story holds at scale, across every channel and every agent, without drifting back into mirror-talk.

Brand is the new backlink. We'd say it a little differently. The clear story is the new backlink, and it always was. The SEO world just caught up to a truth that was sitting upstream of the tactics the whole time. This is just truth.

Three things you can do this week:

1. 1Run the cover-the-logo test on your homepage. Hide your name and show it to someone outside your company for five seconds. Ask them who it's for and what problem it solves. If they hesitate, the machine hesitates too.
2. 2Audit the three surfaces the model reads most ... your homepage, your founder's LinkedIn, and the top comparison page for your category. Do they tell the same story about who you serve? Where they disagree is where your recommendation leaks.
3. 3Pick the one question your buyers actually ask AI ("who's the best [category] for [our situation]") and make sure a clear, sourced answer in your own words exists somewhere the model reads. One clear answer beats a hundred vague pages.