Every business I've worked with can name its competitors. Ask the question in a workshop and the list arrives inside ten seconds: three or four names, usually the same ones the sales team hears in deals, usually companies that look almost exactly like the business asking. It's a comfortable list. It's also, increasingly, not the list that decides anything, because the buyer has stopped building it. Something else builds it now.

The person evaluating you doesn't open five tabs and make a spreadsheet any more. They ask an assistant "who are the alternatives to X?" and get a table back — four columns, a row per company, a tidy little verdict at the bottom. You're either in that table or you aren't, and the position you occupy in it was decided by a model reading whatever the web happened to say about you.

Your competitive set used to be something you argued for. It's now something you're assigned.

Assigned, not argued for

For most of the last two decades, positioning worked because you controlled the frame. You told the market who you were like and who you weren't, and you spent money making that framing stick. If a buyer put you in the wrong bracket, that was a failure of your marketing, and you could go and fix it: a better landing page, a sharper pitch, a comparison of your own.

That whole mechanism assumed a human doing the comparing, one who could be persuaded. The machine assembling the table can't be persuaded. It can only be informed. It reads what exists — your site, your reviews, the directories, the forum threads, the half-accurate roundup some affiliate site published in 2023 — and from that it infers a category for you. If the material is thin, it infers anyway. It always produces an answer, and the answer arrives with the calm authority of something that has read everything.

So the question stops being "how do we beat our competitors?" and becomes something less flattering: who's deciding that these are our competitors at all?

Why the machine reaches for substitutes

When a model is asked for alternatives to a company, it does the easiest available thing. It looks for companies that resemble that company. Same words on the tin, same category label, same shape. It gathers substitutes.

I've written before about the difference between alternatives and substitutes: substitutes do the same thing in a different form, alternatives do a different thing that serves the same purpose. The distinction matters here more than it ever has, because the model is structurally biased towards the first kind. Purpose is hard to infer from a web page. Form is trivial. So you get sorted by form, and dropped into a row next to four companies you'd never have chosen to stand beside, compared on the dimensions they happen to compete on rather than the ones you're actually good at.

If your whole edge is that you solve a problem differently, if you're an alternative rather than a substitute, then being flattened into a substitutes table is the specific thing that kills you. You're now being judged on features you deliberately don't have, by a comparison you never agreed to enter.

The small business gets sorted hardest

Big brands survive this. They've got enough written about them that the model has real material to work with: analyst notes, press, documentation, a decade of coverage. The picture it builds is roughly right, because the evidence base is thick.

The small business has almost nothing. A website, some social, a handful of reviews. When the evidence is thin, the model falls back on the category label, the one word in your headline that sounds most like a market, and sorts you by that. It's a bit like being judged entirely on the sign above the door. You could be doing something genuinely unusual inside, and it wouldn't matter, because nothing you've published gives the machine any way to know.

Worse, you'll never see it happen. Nobody sends you the comparison table. There's no bounce rate for a shortlist you were left off, no analytics event that fires when an assistant places you in the wrong bracket and moves on. You simply notice, some months later, that the enquiries feel thinner and the ones arriving are all asking about the wrong things.

Writing for the thing that sorts you

None of this calls for a new marketing budget. It calls for saying, in plain and unmissable language, the things you've always assumed people would work out on their own, because the reader who used to work it out has been replaced by one who won't infer anything you haven't said.

  1. State the purpose, not just the product. Not "we're a project management tool" but "we exist so a twelve-person agency stops losing money on scope creep." Purpose is what the model can't guess and what places you among alternatives rather than substitutes. If it appears nowhere on your site, it doesn't exist.
  2. Name what you replace, and name what you don't. Say plainly who you're a better fit for and who you're a worse fit for. It feels like a marketing sin, handing business away in writing. It's actually the strongest signal you can send, because a model repeats explicit claims far more readily than it constructs implicit ones, and the disqualification is what makes the qualification credible.
  3. Go and read your own table. Ask three assistants who your alternatives are. The answer is free, it takes four minutes, and it's the closest thing you'll get to seeing your business the way your next buyer will. Most owners have never done it, and are somewhat appalled when they do.

I'll admit I don't know how durable any of this is. The models change every few months, and a tactic that shapes the table today may count for nothing by Christmas; nobody can honestly tell you what the stable version of this looks like. But the customer who used to browse your site now increasingly arrives as something AI-shaped, and pretending otherwise doesn't make the table go away. It just means you had no say in it.

The takeaway

The competitive set was never really yours to control, but you at least used to get a vote. Now the vote is cast somewhere you can't see, by something that will only ever repeat back what it can find. If you've never told the world what you're actually for, something else will decide, and it'll do so in a table where you're compared on everything except the reason anyone should choose you.