2026-08-03 · Jack Stovell
How to get seen in ChatGPT — the mechanics, not the myths
When someone types your name into ChatGPT, one of two things happens. Either the model answers from what it absorbed during training — which for most working professionals is nothing — or it runs a web search and reads a handful of pages in the seconds before it answers. That second path is the one you can actually influence, this month, without waiting for a model retrain.
So the question "how do I get seen in ChatGPT" reduces to a plainer one: when the search happens, is there a page about you that the crawler can fetch, parse and safely quote?
What the crawler actually sees
OpenAI fetches pages with its own crawlers, and they behave like fast, literal readers. No login. No JavaScript execution worth relying on. No patience for a page that returns an empty application shell and hydrates in the browser.
That rules out most of the places your professional identity currently lives:
- **LinkedIn** serves crawlers a login wall. Your profile is invisible.
- **Social bios** are a sentence, shared across a billion accounts, on domains the model weighs for the platform, not for you.
- **A JavaScript-rendered portfolio site** often returns a near-empty HTML shell. The crawler reads your `<div id="root"></div>` and moves on. (This failure is common enough that we wrote a [separate guide on it](/guides/js-shells-soft-404s-how-sites-lie-to-crawlers), and our checker tests for it.)
What actually gets quoted
Pages that get used in AI answers share a boring set of properties. Real HTML with the content in the source. A stable URL that has only ever meant one thing. Facts stated as facts — names, roles, dates, organisations — rather than adjectives. Machine-readable identity markup (schema.org Person or Organization) so the model can tell you apart from the other people with your name. And links out to your other surfaces — GitHub, your site, your company — so the model can corroborate.
None of this is exotic. It's the same discipline as classic SEO, minus the tricks: there is no keyword density game with a model that reads the whole page.
The steps, in order of effect
1. **Have one canonical page about you** that a crawler can read with zero JavaScript. Test it: `curl` your page and look for your actual biography in the response. If it isn't there, nothing downstream matters. 2. **Say checkable things.** "Founder of X since 2024, previously Y" gives a model something to repeat. "Passionate innovator" gives it nothing. 3. **Mark up your identity** with schema.org JSON-LD, including `sameAs` links to every other profile you control. This is how models disambiguate people. 4. **Keep robots.txt out of your way.** Several hosts now block AI crawlers by default — silently. Check yours before assuming anything else on this list is working. 5. **Date your facts and keep them current.** A page with a visible "last verified" date reads as maintained, and stale facts are worse than absent ones.
Where hilyt fits
A hilyt profile is that canonical page, built so you don't have to assemble the parts yourself: server-rendered HTML for crawlers, Person or Organization JSON-LD with your `sameAs` links, dated claims with sources you can verify with one click, and a plain-text card at `/<username>/llm.txt` any model can read in a single fetch. It's free, and you can see a draft of yours in about thirty seconds before creating anything.
One honest caveat: nobody can guarantee a model mentions you — anyone selling that guarantee is selling something else. What you control is whether, at the moment the search runs, there's a page worth quoting. That part takes about ten minutes.