2026-08-03 · Jack Stovell

How to create an AI-readable profile of yourself

To create an AI-readable profile of yourself, publish one page you control at a stable URL: server-rendered HTML carrying your name, a one-line description of what you do, a few dated facts, and links to the profiles that prove your identity. Add schema.org Person JSON-LD with a sameAs array so machines can tell which same-named person you are, then link the page from everywhere you already exist. That is the whole recipe — the rest of this guide walks through each step.

What does AI-readable actually mean?

AI-readable means an AI system can fetch your page with a plain HTTP request and understand it without executing JavaScript, logging in, or guessing who it describes. Four properties do almost all of the work:

Server-rendered HTML. Most AI crawlers fetch raw HTML and do not execute JavaScript, so a client-side-rendered page looks near-empty to them. Semantic HTML for AI readability covers what good markup looks like.
A stable URL you control. Retrieval systems reward pages that stay put; a URL that moves or dies takes your identity record with it.
Machine-readable identity. Structured data that states, without ambiguity, which person this page is about.
No login wall. If a crawler gets redirected to a sign-in page, your profile does not exist as far as that system is concerned.

The login-wall point is why social profiles fail this test. LinkedIn walls off most profile content and blocks most crawlers, so an assistant asked about you usually cannot open the one page you keep most polished.

What goes on the page?

The page needs exactly what a stranger — human or machine — would need to identify you and describe you accurately: name, what you do, evidence, and links.

Your name, as the page title and main heading, in the form you want to be found under.
A one-liner: what you do, for whom, where. This is the sentence assistants will paraphrase, so write it the way you want it repeated.
Dated facts. "Lead engineer at Example Ltd since 2021" beats "experienced engineer". Dates let a model reconcile old and new information instead of blending them.
Links that prove identity: your site, GitHub, published work, employer pages — places that corroborate your claims and are unambiguously the same person.
A visible last-updated date, so a reader and a model can both weigh freshness.

Keep it short and factual. A profile page is a dossier, not a landing page — adjectives give a model nothing to quote.

How do you make it machine-readable?

Add schema.org Person JSON-LD to the page: a small script block that states your identity in the vocabulary machines already parse. Here is a complete minimal example you can paste into your page's HTML and edit:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Person",
  "@id": "https://janedoe.example.com/#person",
  "name": "Jane Doe",
  "url": "https://janedoe.example.com/",
  "jobTitle": "Freelance data engineer",
  "sameAs": [
    "https://github.com/janedoe",
    "https://www.linkedin.com/in/janedoe",
    "https://bsky.app/profile/janedoe.example.com"
  ]
}
</script>

The sameAs array is the disambiguation mechanism: it ties this page to other profiles that are verifiably you, which is how a machine separates you from everyone else with your name. Field-by-field detail and a fuller example live in the Person schema guide.

You can also publish a plain-text llms.txt card alongside the page — a proposed convention for a machine-first index. Adoption is early and no major AI vendor has committed to reading it, but it costs almost nothing and doubles as dense, unambiguous documentation of who you are; what llms.txt is and how to write one makes the honest case.

If you would rather not hand-build any of this, a hilyt profile ships all of it — edge-rendered semantic HTML, Person JSON-LD with sameAs, dated claims with source links, and an llm.txt card — free, with no login wall. i.hilyt.it/jack is a live example.

How do AI systems find it?

AI systems find your page the same two ways search engines do — by following links and by reading indexes — so the job is to link it from everywhere you already exist.

Put the URL in the website field of every profile you hold: GitHub, LinkedIn, X, Bluesky, community accounts, your email signature.
List the page in your site's sitemap so crawlers discover it without depending on links alone.
Use the same name and the same one-liner everywhere. Consistency is the signal machines use to merge scattered mentions into a single identity.

It helps to know there are two separate paths by which an assistant knows you: training data baked into model weights before a cutoff, and live retrieval, where the assistant searches and reads pages in the moment. Retrieval is the path you can influence this week, and a linked, consistent, crawlable page is exactly what it rewards. Influence over training accrues slowly, from the same open material.

Discovery is not instant, either way. Crawling and indexing take days to weeks, and a page that gets updated and linked over time earns more trust than one published once and abandoned — which is another argument for the visible last-updated date.

How do you know it worked?

Test the page the way a machine reads it, not the way your browser does.

Fetch it as a bot: request the URL with an AI crawler user agent set (curl with a changed user-agent header is enough) and confirm the response contains your actual content, not an empty JavaScript shell or a block page. Check if AI can read your website walks through this properly.
Run the free AI visibility checker. It tests your robots.txt AI stance, detects JavaScript shells, checks identity markup, llms.txt and sitemap, and runs a live model probe against your name.
Ask the assistants. With web browsing on, ask ChatGPT, Claude and Perplexity who you are, and see whether your page shows up in the citations.

One honest caveat: AI-readable is necessary, not sufficient. Removing the reasons you can't be read is the part you control; whether a given answer then cites you is decided by retrieval, ranking and model behaviour. But those systems can only choose from pages they can read — so this is the step everything else depends on.

Create your AI-readable profile