How do we position an AI product?

You position an AI product by leading with the problem you end and the buyer you built it for, then describing the AI as the mechanism that makes your answer possible, specifically enough that a skeptical buyer can check it. "AI" belongs in the second sentence of your story. The first sentence is the buyer recognizing their own problem.

That order matters more every quarter. Every competitor in your category now says AI, and a word everyone uses stops sorting anyone. The people doing the sorting increasingly aren't people, either. Your buyer asks ChatGPT or Perplexity for a shortlist, and the engine can only repeat what your site makes clear.

Why is positioning an AI product harder than positioning other software?

Three things work against you at once. The word is free: any vendor can type "AI-powered" into a hero, so the claim proves nothing on its own. The category is unsettled: buyers often don't know what to compare you to, which means they default to whatever list an engine hands them. And the buyer is wary: they've sat through demos that looked brilliant and died in the pilot.

The usual response is to say AI louder. We call that AI-Parmesan: sprinkling "AI-powered" on top of a narrative that was weak before the AI showed up. It feels like positioning. It reads like everyone else. I covered when the word still earns its place in Is calling our product AI-powered still helping us, or is the word costing us trust?, including a deletion test worth running on your own hero.

What does the data say about how AI products position themselves?

In our 2026 Healthtech Messaging Index we scored 302 healthtech homepages on 19 criteria. For 105 of them, our scorer filed the product in an AI category: AI medical scribes, clinical AI for revenue cycle, AI patient communication. Criterion 14, the AI-Parmesan Index, asks whether the page says what the AI actually does. Here's how the two groups compared.

That's correlation, and the companies with specific claims probably do other things well too. The pattern is still hard to ignore. When the page says what the AI does, the company is far easier to tell apart from its neighbors, and more of them show up when an engine is asked who to recommend.

The top-scoring company in the index shows what specific looks like. Theator doesn't lead with AI at all. It leads with a problem surgical leaders know: operative reports written from memory, hours after the procedure. Then it names the mechanism: the surgical video that already exists becomes a structured report. A buyer can check every part of that sentence.

How do I position an AI product, step by step?

1. 1Start with the old way. What does your buyer do today to live with this problem, and what does it cost them? April Dunford makes competitive alternatives the first step of positioning for a reason, and it matters double for AI products, because the real alternative is usually a spreadsheet, a person or a habit, rather than another AI vendor. I walked through her method in Competitive alternatives: how to run Dunford's first step.
2. 2Name the buyer as a role. "Health systems" is a market. "The VP of revenue cycle who owns denial rates" is a buyer, and an engine can match that buyer to a question.
3. 3Describe the AI as input, action and output. What does it read, what does it do, and what does the buyer get? "Reads every prior-auth request, flags the ones likely to be denied, so your team works the risky 15% first" is a claim. "AI-powered prior authorization" is a label. If you deleted the word AI and the sentence still says something, you've got a mechanism.
4. 4Pick a category the buyer already searches for, then say how you're different inside it. Inventing a category is expensive. If buyers type "AI medical scribe," use it. More on that choice in What category should my B2B product be in?.
5. 5Add proof that costs you something to claim: a named customer, a measured result, a pilot length you'll commit to. How do we prove our AI is real and not just marketing? goes deeper on that.

Should "AI" be in our category name?

Only when your buyer searches that way. Some categories have settled around the word: buyers really do type "AI medical scribe" or "AI SDR." In those markets, leaving AI out makes you harder to find. In most markets, though, your category is the job the buyer is hiring for, and AI is how you do it. A buyer looking for denial management software wants fewer denials. They'll care that you use AI the moment you show them how it gets them there.

A quick test: write your category the way a buyer would type it into a search box. If the word AI shows up naturally, keep it. If you had to add it, it's decoration.

How do I make sure AI engines describe our AI product correctly?

The engines briefing your buyer read your site, your reviews, your LinkedIn and the lists other people publish. If your homepage says "the AI-powered platform for healthcare," the engine files you next to a few hundred companies that say the same thing, and it has no reason to pick you. That's what the Entity Distinctiveness gap in the table is measuring: 92% of the specific group could be told apart from their competitors, against 43% of the rest.

Two things fix it. The first is the specific claim from step 3, said the same way everywhere, so the engine meets one story instead of five. The second is a single documented source for that story. At PitchKitchen that's the Magnetic Messaging Framework, turned into an AI Brand Twin that your team and their AI tools write from. When your reps, your site and your AI-drafted content all describe the mechanism the same way, the engines start to repeat it back. We go further on this in What changes about B2B positioning when AI is doing the buyer research?.

Want to see where your homepage stands? Run the free Brand Signal Score. It scores your page on the same 19 criteria we used for the index, including the AI-Parmesan Index and Entity Distinctiveness, and it shows you which lines an engine can actually lift.

What does a positioned AI product sound like?

Here's a hypothetical before and after for a revenue cycle company. Before: "The AI-native platform transforming revenue cycle management." After: "Denied claims cost a mid-size health system millions a year, and most teams find out after the money is gone. We read every claim before it's submitted, flag the ones likely to be denied and tell your team why, so they fix the risky ones first."

The second version names a buyer, a problem, an old way and a mechanism. It also gives an AI engine something specific to repeat when a VP of revenue cycle asks who can help with denials. Would you agree that's the sentence you want an engine saying about you?

Your model will improve every quarter, and so will your competitors'. The story around it is the part a buyer can't get anywhere else, and it's the only part an AI engine can read.