Two things now happen when somebody wants to know about a field. They search, and they ask. The second one is growing and it works differently, because a ranking is a list of places to look and an answer is a claim already made.
Getting named in that answer is not the same job as ranking. Some of the work overlaps. A good deal of it does not, and the parts that do not are the parts almost nobody is doing.
01What these systems are actually doing
Simplifying considerably, an assistant answering a question about a person or a field is drawing on text it has seen or retrieved, and producing a claim it can support. Two things follow from that, and they explain most of what works.
- It needs text, not design. Anything expressed only visually is invisible.
- It needs the text to be unambiguous about who you are, so it can attach the claim to the right entity.
- It weighs corroboration. A fact stated in five independent places is safer to assert than one stated on your own site alone.
- It cannot verify, so it avoids claims it cannot support, which is why thin entities get vague answers.
- Contradiction makes it hedge or pick, and when it picks it sometimes picks wrong.
02Consistency beats volume, and it is not close
This is the biggest practical difference from traditional SEO. In search, more content is generally more. In AI visibility, contradictory content is actively harmful, because it makes every claim about you less safe to state.
If your site says you are a personal branding consultant, your LinkedIn says you are a growth marketer, your old bio says content strategist and a directory says SEO expert, you have not given a model four descriptions. You have given it a reason to say something vague.
- Write your canonical description once. Role, field, city, organisation.
- Apply it verbatim to your site, your about page, your schema, every profile and every bio you send to an event or publication.
- Go back and fix the old ones you can still edit. Stale bios are the main source of contradiction.
- Send the canonical version to anyone writing about you, unprompted, so third-party records reinforce rather than dilute.
- Re-check quarterly, because drift is continuous.
Four descriptions of yourself do not make you four times as findable. They make every statement about you less safe to assert, which is how you end up unnamed.
03Write so you can be extracted
A model quoting you needs a claim it can lift. Most professional writing buries the claim: three paragraphs of context, a caveat, then the point, then more caveats. That structure is unquotable, so it does not get quoted.
- State the answer first, then support it. The inverted pyramid works here for the same reason it works in journalism.
- Use plain question headings that match how people actually ask, then answer immediately underneath.
- Keep the answer paragraph self-contained. It should make sense lifted out of the page with no surrounding context.
- Put one idea per paragraph. Paragraphs carrying three ideas get quoted wrongly or not at all.
- Include a real FAQ with real questions and complete answers, because that structure is close to ideal for extraction.
- Name things explicitly. Pronouns and this and it force a resolution step that can go wrong.
The extraction test
- Take any paragraph out of your page and read it alone
- Does it still make a complete claim?
- Is it clear who it is about, without the rest of the page?
- Could somebody quote it without needing to add context?
- If not, rewrite it so they could
04Corroboration is the hard part
A claim that appears only on your own site is a claim with one source, and one source is the weakest possible position. This is why third-party records matter more for AI visibility than for anything else in personal branding.
What helps: published interviews where you are described in text, contributed pieces under your byline on domains with history, conference and event pages, book catalogue entries, and any credible reference work that states the same facts your site does.
What helps very little: directory listings, paid placements, and profiles on platforms that generate thousands of near-identical pages. Volume without editorial judgement is close to worthless here.
05Track it properly, monthly
Fixed prompts, fresh conversations, saved verbatim. Same wording every month or the comparison is meaningless.
| Prompt | What it tests |
|---|---|
| Who is [your name]? | Whether you exist as a recognisable entity |
| What is [your name] known for? | Whether your positioning has landed |
| Who are the leading people working on [your topic] in [your city]? | Unprompted naming, the hardest and most valuable outcome |
| Who should I hire for [your service]? | Commercial intent, the one that matters most |
| What has [your name] written about [your topic]? | Whether your body of work is attached to you |
- Run all five on ChatGPT, Claude and Perplexity, in fresh conversations with no memory of you.
- Save every answer verbatim, with the model and the date.
- Score three things: are you named, is it accurate, is it favourable.
- Log any confidently stated falsehood separately and treat it as urgent, because wrong answers propagate.
- Compare month over month. The trend is the signal; any single answer is noise.
06What does not work
Ignore these
- Stuffing keywords, which reads as low quality and does not help extraction
- Publishing volume without consistency, which makes things worse
- Trying to prompt-inject instructions into a page for a model to follow
- Buying mentions on sites with no editorial process
- Claiming credentials that cannot be corroborated, because thin claims get dropped
07The realistic timeline
Descriptions tend to improve first, within a few months of fixing consistency, because that only requires the model to find clean text about you. Being named unprompted in a category question is much harder and much slower, because it requires the model to consider you among the strongest candidates in a field. Six to twelve months of genuine record building is the honest range, and for competitive fields it is longer.
What to remember
- Ranking and being named by an assistant are different outcomes won by different work.
- Consistency beats volume. Contradictory descriptions make every claim about you less safe to state, so nothing gets stated.
- Write extractable paragraphs: the claim first, self-contained, one idea each.
- A claim only your own site makes has one source. Third-party corroboration is what turns it into a fact worth asserting.
- Track five fixed prompts monthly across three assistants and treat confident falsehoods as urgent.
Questions people actually ask
How do I get ChatGPT to mention my name?
Give it clean, consistent, corroborated text to work from. Use one canonical description of yourself everywhere, write self-contained paragraphs that state a claim before supporting it, and build third-party records that repeat the same facts your own site states. There is no submission route; it is entirely a function of the text that exists about you.
Why does AI give wrong information about me?
Usually because your properties contradict each other, or because a namesake has a stronger footprint. When descriptions conflict, a model has to pick, and it sometimes picks wrong and then repeats that confidently. Fix the contradictions on properties you control first, then work on the strongest third-party sources.
Is AI visibility different from SEO?
They overlap but they are not the same. Search can show a link and let the reader judge; an assistant has to commit to a statement, so it needs unambiguous and corroborated text. Consistency matters far more, and contradictory content that would be harmless for SEO actively hurts here.
How do I track whether AI mentions me?
Run five fixed prompts monthly across ChatGPT, Claude and Perplexity in fresh conversations, using identical wording every time. Save each answer verbatim with the model and date, score whether you were named, whether it was accurate and whether it was favourable, and compare the trend rather than any single answer.
How long does it take to appear in AI answers?
Accurate descriptions of you tend to improve within a few months of fixing consistency. Being named unprompted when someone asks who the leading people in your field are is much harder, and six to twelve months of genuine record building is the honest range, longer in competitive fields.