2026-08-03 · Jack Stovell

An AI-readable profile for consultants

An AI-readable profile for a consultant is a public page that states, in plain crawlable text, the specific problems you solve, for whom, and the dated work that proves it. When a buyer asks an AI assistant who can help with a problem in their market, the assistant builds its shortlist from pages it can read in that moment. If your niche exists only in a slide deck, a LinkedIn profile or your own conversations, you are not in the running.

How does an AI assistant decide which consultants to recommend?

It searches, fetches the pages it finds, and shortlists the candidates it can describe with confidence. The discovery moment looks like this: a buyer asks an assistant "who can help us reduce churn for a B2B SaaS company in the Nordics", the assistant runs searches, reads what comes back, and names people whose pages plausibly answer the question. There are two ways to be known to a model — baked into its training data, or found by live retrieval — and almost no consultant is famous enough for the first. Retrieval is the arena, which is good news: it is the path you can influence this week. How answer engines pick between candidate sources is its own subject — see how AI decides which sources to cite — but none of the picking happens if there is nothing to read.

How do you make sure AI recommends you for your niche?

You cannot make sure. Nobody controls model output, and anyone selling a guaranteed recommendation is selling something they do not have. What you can do is make certain that when the shortlist forms, you are findable and verifiable. The lever is phrasing: your niche has to exist on the page in the language clients use when they describe the problem. "Strategic advisor to ambitious organisations" matches no query anyone types; "pricing strategy for seed-stage B2B SaaS" matches a real one. Go through your page and check that every engagement type you want to be found for appears as a plain sentence a buyer's question could land on — the market, the problem, the kind of client. This reads as less impressive than positioning language. It is what retrieval actually matches on.

What separates you from everyone else claiming the same niche?

Corroboration. Any consultant can write "I help companies with X" — self-description is the weakest class of evidence a machine can find. The strongest is client-facing artefacts that exist outside your own site: a conference talk with your name on the programme, an article in an industry publication, a case page a client links to, a podcast appearance. Each is a page someone else published that confirms you do the work you claim. Cross-checking is cheap for a machine, and when an assistant does it, those artefacts are the difference between "this page says" and "multiple sources agree". If you have them, link every one from your profile. If you do not, earning two or three is worth more than any rewrite of your own copy.

Does a dated track record matter more than good copy?

Yes. "Twenty years of experience" is an adjective wearing a number; an engagement history with years, sectors and outcomes is checkable structure. AI systems preferentially repeat claims that look verifiable — a role, an organisation, a date, a source. So write your track record as dated entries rather than prose: what you did, for what kind of client, when, with a link where one exists. A hilyt profile is built around exactly this unit — claims with source URLs and dates, rendered on a plain-text card like i.hilyt.it/jack/llm.txt — but the principle holds on any page you maintain yourself.

The generic layer matters too, but it is the same for every profession and covered in depth elsewhere: one canonical page that is open and crawlable, schema.org Person JSON-LD to separate you from other people with your name, and a bio that says the same thing on every surface it appears. The full walkthrough is how to create an AI-readable profile of yourself.

What won't this get you?

A recommendation on demand. Being readable and corroborated removes the reasons an assistant cannot surface you; it does not decide that it will. The deciding factors — retrieval ranking, competition for the query, model behaviour — sit outside anyone's control, and we publish our own citation misses rather than pretend otherwise. The honest goal is narrower and still worth having: when a buyer's question creates a shortlist in your niche, you are on the field, verifiable, and easy to describe accurately.

Preview your profile in 60 seconds