Why doesn't ChatGPT mention me when someone asks for the best consultant?

The advice you found was written for companies, which is why it does not work

Almost every guide to being mentioned by AI assumes the thing being mentioned is a company, and a company has a property that you do not have. It owns a domain that functions as its identity. Structured data on that domain, a crawlable site architecture, a file at the root describing the business, review-site profiles and vendor directories all point at one address, and that address is the entity.

A person has no equivalent anchor. Your website is a site about you, which is not the same thing as being you in the way a company domain is the company. You could delete it tomorrow and the model's picture of you would barely shift, because that picture was assembled from bylines, interview transcripts, speaker rosters, directory entries and other people's mentions, most of which sit on servers you have never had access to.

So the company playbook produces a specific, frustrating result for individuals. You do everything on the list, your own pages become immaculate, and the answer to who is the best consultant in your field still contains four other names and not yours. Nothing on that list addressed the actual constraint.

Two different pipelines decide whether your name appears

There are two mechanisms and they have almost nothing in common. The first is trained memory, which is what the model absorbed from text before its training cut-off. The second is retrieval, where the product runs a search at the moment of the question and writes its answer from the pages it fetches. Advice that does not say which one it is addressing cannot be acted on, because the two respond to opposite inputs.

The two pipelines, and why the same tactic can be useless for one and decisive for the other.

Run both tests before you spend a rupee or an hour. If the offline answer omits you and the online answer names you, your problem is historical text volume and the fix is slow. If both omit you but the online answer names people whose visible credentials are no better than yours, your problem is retrieval and the fix is measured in weeks.

Three keys, and most consultants hold one

A model produces a person's name when three conditions are true at the same time. Each one has a distinct failure mode, and the failures look different enough that you can identify which key is missing from a single answer.

The order matters and it is not negotiable. Corroboration for an unresolved name simply strengthens somebody else's entity, which is the quiet reason a scattergun publicity push can produce genuine coverage and zero change in what a model says about you.

What names a company versus what names a person

The two source lists overlap far less than the shared vocabulary suggests. Below, the same sources are scored for a company and for an individual, with the key each one supplies and whether you can publish it yourself.

A source map for a named individual. Read the last two columns together: the strongest person sources are mostly the ones you cannot publish.

Count the strong rows in the person column. Seven of them are sources you cannot publish, which is the entire reason the company playbook misfires. For a business, visibility work is largely a publishing problem. For an individual it is largely an access problem, and the two require different budgets, different skills and different patience.

Your profile still matters, but not for the reason most people assume. It is the resolution anchor that the other sources point back at, which is a specific job with specific requirements, examined in what AI reads on your profile.

You are competing to be an item in a list, not to have the best page

Look closely at the question you are worried about. Who is the best consultant for X is list-shaped. It asks for names in a set, not for a description of one person. A model answering a list-shaped question draws on list-shaped sources, and if your name has never appeared inside somebody else's list of names, you are not in the candidate pool at all.

This is the distinction that explains most of the confusion. A page about you is evidence that you exist. A list that contains you is evidence of standing. Only the second can answer a superlative, because a superlative is a comparison and a page about one person contains no comparison.

• Roundup articles where a writer names several practitioners, which is why being useful to journalists outperforms pitching yourself to them.
• Awards shortlists, including the ones you do not win, because the shortlist page carries every name on it.
• Conference speaker rosters and panel line-ups, which are permanent, indexed pages listing people beside a topic.
• Professional association and accreditation directories, where the category is in the page title and your name is in the body.
• Contributor and expert panels attached to publications, where a masthead page lists people with their specialisms.
• Podcast guest indexes, where a show archive becomes a list of names attached to a subject.
• Book contributor and anthology pages, where several named people sit under a single category heading.

The namesake problem, which companies never have

A company can register a trademark, buy the exact-match domain and own its name in a way you cannot. You share your name with strangers, and if one of them is more prominent, the model's entity for that string belongs to them. Everything you publish then reinforces a record filed under somebody else's identity.

The repair is unglamorous and it works. Choose one canonical form of your name and never publish under another. Include the middle initial if you need it, keep or drop the accent consistently, and stop alternating between the short form your friends use and the long form on your passport. Then place a disambiguator immediately adjacent to it everywhere: the city, the category, or the firm.

Test it in one prompt. Ask a model to tell you about your full name plus your category. If it returns the other person, or blends both of you into one biography, resolution is your problem and no amount of publishing will fix it until the name itself is disambiguated. The same failure shows up in ordinary search, which is why when nothing comes up for your name and this problem are usually the same problem wearing two coats.

The answer you already get tells you which key is missing

You do not need an audit to diagnose this. Ask one question, read the shape of the reply, and the missing key is named for you. The shapes are surprisingly consistent across products.

Six answer shapes and the diagnosis each one carries.

The final row is the one people skip. Once a model does name you, it names you using somebody else's sentence, and that sentence becomes your public description whether or not you would have written it that way.

Run the check monthly, and record it properly

This has to be a repeatable measurement rather than an anxious spot check, because a single answer from a single session tells you almost nothing. Models sample from a distribution, so the same question asked twice can produce different names. What you are tracking is a rate, not an event.

One practical note. Keep the sheet even when the answers are bleak, because the value arrives in month four when a shape changes and you can point to exactly which piece of work landed between the two dates.

What will not move it, said plainly

Several popular tactics have no mechanism behind them for an individual, and knowing which ones are inert saves the budget for the ones that are not.

• Posting more on LinkedIn, on its own. Posts are excellent for the humans who see them and they are a poor entity source, because they are not indexed the way an article on a publication is and they do not corroborate anything.
• A file on your own domain that describes you to model crawlers. Even where it is read, it sits on a server you control, which means it can describe an entity and can never corroborate one.
• Keyword stuffing your bio with your category ten times. Repetition on one page you own is not the same signal as repetition across pages you do not own.
• Telling the model about yourself inside a chat. That context lasts for the conversation and changes nothing for the next person who asks.
• A single press release on a distribution wire. It creates many near-identical copies of one source, and near-identical copies are closer to one source than to many.
• Buying followers or engagement. Neither is visible to the mechanisms described here at all.

Meanwhile the work does have a real timeline, and it is honest to say that the timeline is long. Resolution can be fixed in a week. Category wording can be fixed in an afternoon. Corroboration is the slow one, because it depends on other people publishing, and other people publish on their own schedule.

That is an argument for doing the fast keys immediately and treating corroboration as a standing quarterly commitment rather than a campaign. It is also an argument for not letting this work displace the things that produce enquiries this month, which is why your profile as a conversion step stays the higher priority while the entity work compounds quietly underneath it.