Do you show up when someone asks ChatGPT about you?

Whether a person appears in AI answers depends far less on their own profile than on crawlable content other people have written about them. In measured studies, roughly eighty-four per cent of AI citations come from earned editorial media. BrandLoop measures that threshold as a separate GEO score and lists what is missing.

Being named when someone asks an AI engine about you is not a continuation of classic search optimisation. The mechanism is different, and on one point it points the other way entirely.

ChatGPT cannot read LinkedIn directly

LinkedIn’s robots.txt names around twenty-five crawlers. GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot and Google-Extended are not among them. Googlebot is named and allowed. So AI engines reach LinkedIn content only indirectly, through surfaces that rely on Google’s index.

The consequence is clean: your LinkedIn profile is an identity anchor, not a source. Turn off the public profile setting and even that indirect path closes. How the same profile fares in recruiter search is a separate question, covered in the piece on how headhunters search.

The lever is not what you write, it is what is written about you

A study of more than twenty-five million cited links found that eighty-four per cent of citations in AI answers come from earned editorial media. Paid content accounted for three in a thousand. Whatever you publish on your own site, this line does not move until somebody independent writes about you.

You still need a page

A meta-analysis across more than fifty studies put URL accessibility and search rank at the top. This is a threshold: with no crawlable page there is no signal to carry. Login walls, JavaScript-only rendering and a noindex tag each close it.

Format makes a measured difference

That last line matters: keyword stuffing was the only method measured to reduce visibility. Some habits carried over from classic search optimisation actively hurt here.

Advice with no evidence behind it

• llms.txt: Google’s John Mueller stated plainly that no AI system currently uses it. Server logs across roughly nine hundred domains over seven months recorded zero requests from the major AI crawlers.
• schema.org markup: a study comparing 1,885 pages against 4,000 controls found the measured effect indistinguishable from zero, and slightly negative on one platform. Treat it as thirty minutes of hygiene, not as a lever.
• Rank tracking: ask the same question a hundred times and the odds of two answers returning the same list are below one in a hundred. Tracking rank is noise. Share of appearances across many repetitions is a real measurement.

What belongs on your own page

The page needed to clear the threshold is not a portfolio; it is an entity page. Its single job is to hold sentences an engine can quote when describing you. In practice that means four things.

• A one-sentence definition: “Name is the person who does X in Y.” At the top, meaningful on its own.
• A few verifiable, concrete records: a talk, a publication, a product, a role — with dates and places.
• Links to your other profiles: LinkedIn, a company page if there is one. Entity matching runs through these.
• Plain answers to the questions people ask. Query-answer matching is the highest-scoring measured factor, so headings should be shaped like the questions people type.

There is a technical floor too

Writing a page and having it read are not the same thing. Most AI engines run separate crawlers, and confusing them is a common error: blocking the training bot does not end your visibility, blocking the search bot does. The detail is in our piece on where ChatGPT reads from, which also shows that the most commonly missed blocker is not robots.txt but the content delivery network.

Measuring visibility, step by step

GEO work without measurement is work without a result. There is a simple way to measure your own:

• Prepare three question types: a direct question with your name, an advice question in your field, and a “who are the people in X” list question.
• Ask each on every engine at least ten times, in a fresh session with history off.
• Count how often you appear, not where you rank. Ask the same question a hundred times and the odds of two answers returning the same list are under one percent; a single measurement is noise.
• Measure your languages separately. Source overlap between two languages has been measured at only 22%.
• Repeat quarterly and record the percentage. The benchmark is your own previous measurement.

How editorial coverage actually happens

If the large majority of citations come from editorial content, that is where the work is. There is no shortcut, but there is a known route: produce something concrete a journalist or trade publication can use — an original number, a case, a measurement. General opinion pieces do not get cited; data does.

Starting small also works: industry newsletters, professional associations, university pages, podcast show notes. All of these are crawlable and count as independent sources.