First, the thing most guides get wrong
An assistant answering a question about you is usually not reciting training data. It is searching, reading a few pages, and summarising them. That is why the answer changes month to month, and why a page published last week can show up in it.
Which means this is winnable. It also means the lever is not "get into the model". The lever is: be the clearest, most consistent, most quotable source about yourself that a crawler can reach.
Step 1. Find out what they say now
Open ChatGPT, Claude and Gemini. Start a new conversation in each, turn memory off, and ask the same question: who is [your full name]? Then ask a harder one: is [your name] credible in [your field]?
Save the answers. This is your baseline, and it is the only honest way to tell later whether anything changed. Do this before you touch a single page. Almost nobody does, which is why almost nobody can prove their work.
If you would rather work from a written checklist, the free ten minute self audit sets out the exact questions to ask each assistant and what every kind of answer tells you.
Step 2. Check you are not blocking the crawlers
Many sites quietly block the exact crawlers they want. Open your own robots.txt and look for these names. If any is disallowed, that assistant cannot read you.
- GPTBot, OAI-SearchBot, ChatGPT-User for ChatGPT.
- ClaudeBot, anthropic-ai for Claude.
- Google-Extended for Gemini and Google's AI answers.
- PerplexityBot, CCBot for Perplexity and Common Crawl.
This takes five minutes and is the single most common reason a professional is invisible to assistants while ranking perfectly well on Google.
Step 3. Stop contradicting yourself
A model comparing your website, your LinkedIn and a directory listing will find three different job titles, two cities and a subscriber number that does not match. Faced with disagreement, it hedges, and a hedge reads as doubt.
Pick one description of what you do and one location, and make every source say it identically. Not similarly. Identically. This is unglamorous and it is the step with the largest effect.
Step 4. Give the crawler structured data
Publish Person schema on your site: your name, your role, your location, and a sameAs list pointing at every profile that is genuinely yours. That last part is what lets a machine connect a LinkedIn profile, a YouTube channel and your website to one human being instead of guessing.
If you hold certifications, mark each one up with its issuer and its verification link. A credential an assistant can check is worth more than ten it cannot.
Step 5. Write in sentences a model can lift
Assistants quote. They do not paraphrase well, and they avoid material they cannot condense safely. So write short declarative sentences that are true on their own, out of context.
"Board certified in interventional cardiology, practising in Houston since 2008" can be quoted. "A passionate, patient centred approach to cardiovascular wellness" cannot, because it says nothing checkable. Marketing language is invisible to machines for the same reason it is unconvincing to people.
Step 6. Get one source that is not you
This is why assistants say "self reported". If every claim traces back to your own website, a careful model attributes it to you rather than stating it. One credible third party changes the grammar of the answer.
A professional association listing, a university page, a conference bio, a published paper, a genuine press mention. It does not need to be the New York Times. It needs to not be you.
Step 7. Re-check, and expect it to take a while
Repeat step 1 every few weeks with the same questions. Live search picks up changes in days once your pages are indexed. Anything drawn from training data moves on the model's schedule, not yours, and nobody selling you this service controls that.
Questions people actually ask
- How do I show up in ChatGPT search results? The same way you show up anywhere a machine reads you. Make the facts about you exist in a form it can check, make every source agree, and let the crawlers in. There is no separate switch for ChatGPT search. It reads the open web and whatever its search partner returns, so most of the work that makes you findable on Google is the same work that makes you show up in ChatGPT.
- How long does it take to appear in ChatGPT results? There is no fixed timetable. Assistants that search the live web can pick up a change within days once the page is indexed. Answers drawn from training data can lag by months, and no supplier controls that schedule.
- Can anyone guarantee what ChatGPT says about me? No. You control the inputs: whether the facts exist in a readable form, whether they agree with each other, and whether the crawlers are allowed in. The model decides the sentence.
- Why does ChatGPT say my achievements are self reported? Because the only source for them is you. When a claim appears solely on your own site, an assistant attributes it to you rather than stating it as fact. Independent coverage is what removes the hedge.
- Is this the same as SEO? No. Search ranks pages and shows a list. An assistant returns one answer with no second result underneath it. Search rewards links and authority. Assistants reward being legible and consistent.
What nobody can promise you
Not a specific sentence. Not a position. Not a date. Anyone who offers those is describing something they do not control, and you should treat it the way you would treat a doctor guaranteeing an outcome.
What is controllable is everything above: whether the facts exist in machine readable form, whether they agree, whether the crawlers are allowed in, and whether a single source other than you confirms the record. Do those and you have done the whole of the available work.
If you would rather not do it yourself, this is what I do for physicians, attorneys, academics and executives.