All promptsAI visibility

Getting your business named by ChatGPT, when most of the sources are not yours

There is no rank inside ChatGPT, and for a small or local business the deciding sources are third party. Build a worklist from the citations in a real answer.

Works in
ChatGPT, Claude, Gemini
You need
Your business name, category and town · The full ChatGPT reply to a real buyer question, citations included · Every place your business is currently listed
Written for
how to get your business rank on chatgpt
98A

Scored by our own engine

This page, run through the audit we sell. Measured 4 August 2026.

Score your own page →

There is no rank to get inside ChatGPT. No assistant publishes a ranking system, exposes a position or reports a citation back to you, so the honest version of this question is which sources an answer is assembled from and whether you appear in them. The prompt below builds that list from the citations in a real answer rather than from convention.

Being an entity, not a website

For a small or local business, most of what an assistant reads about you was not written by you.

The pattern is easy to check and uncomfortable to look at. Ask a recommendation question in your category, then read the citations rather than the prose. They tend to be maps listings, review platforms, trade directories, local roundups and comparison posts, and your own site appears somewhere in the list without deciding anything.

That changes what the work is. It is less about rewriting your homepage and more about being a consistent, findable entity across sources you do not own. A business described one way on its site, another way on a directory and with an old phone number on a third is three weak signals instead of one clear one.

The part you do own is worth doing properly. Say what you do, where, and for whom in plain declaratives that survive being lifted out of the page. Publish Organization schema with a sameAs list pointing at your profiles elsewhere, so the pages that talk about you and the pages you own resolve to one thing. Check it with the structured data checker.

What can and cannot be claimed here

Crawler behaviour is documented. Selection is not.

OpenAI publishes what each of its crawlers does, including a separate agent that surfaces sites in the search features and another that fetches a page when a user asks. None of that documentation says how a source is chosen for an answer, and no operator publishes it. Anybody quoting you a mechanism, a timeline or a guaranteed mention is filling that gap with confidence rather than information.

So the prompt refuses to invent one. It works in three tiers: sources actually cited in the answer you paste, sources of the same kind that plausibly matter, and conventional advice it cannot tie to any evidence, each labelled as what it is.

What to do with the output

Start with tier one. A directory or roundup that was cited and does not list you is usually an email, not a project, and it is the highest value line the run produces.

Then make your details identical everywhere, which helps ordinary local search too and is the one action with no bet in it. For the site level plan around a single location, see the local SEO plan prompt. For what is happening underneath all of this, see how LLM SEO works, and for the labels people will sell it to you under, AEO and GEO, which is which.

The prompt 521 words
You are an entity and citations analyst working on a small business. I am
going to paste a real answer an assistant gave to a buyer question in my
category. Build me a worklist from the evidence in it. Do not give me general
advice about ranking in ChatGPT, because there is no rank to give advice about.

Open with two lines and nothing more. First, state that no assistant publishes
a ranking system, exposes a position, or reports citations to a business, so
everything below is influence rather than control. Second, say which of these
three situations I am in based on the reply I pasted: NOT MENTIONED, MENTIONED
INACCURATELY, or MENTIONED ACCURATELY. They are different problems.

Then the worklist. Work only from evidence, in three tiers, and label each tier.

TIER ONE, OBSERVED. Every source actually cited in the reply I pasted, listed
by name. For each one: is my business present on it, is my information there
correct, and can I edit it myself, get it corrected by request, or not at all.
If the reply cited a competitor page, a roundup or a directory I am missing
from, that is the most valuable line in this whole output and it goes first.

TIER TWO, LIKELY. Sources of the same kind as the ones cited, in my category
and my area, that did not appear in this particular answer. Name real, specific
ones where you can. Mark this tier clearly as inference, not evidence.

TIER THREE, UNVERIFIED. Anything you would normally recommend that you cannot
tie to a source in the reply. Say plainly that it is convention rather than
evidence. Keep this tier short and do not pad it.

Then three closing sections.

ENTITY INCONSISTENCIES. Compare my business name, category description, address
and any other detail as they appear in the reply against what I told you.
Quote each mismatch. A business described in two ways in two places is one of
the few problems here that is both cheap to fix and clearly worth fixing.

WHAT I WOULD DO FIRST, three actions in order, each with who does it, roughly
how long, and whether it is an edit I control or a request to somebody else.

WHAT THIS WILL NOT DO. Be direct. Say that none of this guarantees being named,
that answers vary between sessions and accounts, and that there is no report
anywhere confirming a change worked.

Constraints:
- Do not claim to know how ChatGPT selects or ranks sources. That is not
  published.
- Do not invent statistics, adoption figures or shares of anything.
- Do not invent a directory, review site or publication. If you are unsure a
  named source exists, put it in tier three and say you are unsure.
- Do not promise a position, a citation or a timeline.

My business, category and area: [YOUR BUSINESS NAME, CATEGORY AND TOWN OR CITY]
The question I asked: [THE BUYER QUESTION YOU TYPED]
The full reply: [PASTE THE FULL REPLY INCLUDING EVERY CITATION AND LINK]
Where I am already listed: [LIST EVERY PLACE YOUR BUSINESS IS LISTED, OR WRITE "not sure"]

What to change

Everything in square brackets is yours to replace. Nothing else needs editing.

[YOUR BUSINESS NAME, CATEGORY AND TOWN OR CITY]
Exactly as you trade, including the legal suffix if you use it. "Marsden Joinery Ltd, fitted kitchens, Bristol" gives the entity check something to compare against. A vague category gives you a worklist for a business that does not exist, and the area is what makes tier two nameable at all.
[THE BUYER QUESTION YOU TYPED]
The question as a stranger would ask it, not your keyword. "Who makes fitted kitchens in Bristol" or "best fitted kitchen company Bristol". Ask the one your buyers would ask, because the sources an assistant reaches for change completely between a recommendation question and a factual one.
[PASTE THE FULL REPLY INCLUDING EVERY CITATION AND LINK]
The whole answer with the citation list intact, not a summary. The citations are the entire evidence base for tier one, and pasting the prose without them turns the worklist into guesswork dressed up as analysis.
[LIST EVERY PLACE YOUR BUSINESS IS LISTED, OR WRITE "not sure"]
Maps listing, review platforms, trade bodies, chambers, supplier locators, old directories. One per line, imperfect is fine. This is what the entity check compares against, and "not sure" is an honest answer that turns the first action into an audit of where you already appear.

How to run it

  1. 01
    Ask the recommendation question in a temporary chat

    Use a fresh or temporary session with search enabled and no memory, so you get the answer a stranger would get rather than one shaped by your history. Ask the question a buyer would ask, and copy the entire reply with its citations.

  2. 02
    Ask it two more times and keep all three

    Answers vary between runs. Three replies show you which sources appear consistently, and those are the ones worth working on. A source that appeared once may have been a coincidence, and building a plan on one reply is building on a sample of one.

  3. 03
    Write down where you are already listed

    Maps, review platforms, trade bodies, supplier locators, anything with your name on it. Include the stale ones with the old phone number, because those are exactly what the entity section is for and they are usually the reason a business is described two different ways.

  4. 04
    Run the prompt and start at tier one

    Tier one is the only part built on evidence. A directory or roundup that was cited and does not include you is the single most actionable line the output will produce, and it is usually an email rather than a project.

  5. 05
    Fix the entity mismatches before writing anything new

    Make the name, category, address and phone identical everywhere you appear. It is dull, it is cheap, and it helps ordinary local search at the same time, which makes it the one item on the list with no bet attached to it.

  6. 06
    Declare who you are on your own site, then re-ask in a month

    Publish Organization schema with a sameAs list pointing at your profiles elsewhere, so the pages talking about you and the pages you own resolve to one entity. Check it with the free structured data tool, then ask the same three questions again from fresh sessions and compare which sources are named.

Questions people ask

How do I get my business to rank on ChatGPT?

You cannot rank in the sense you rank on Google, because there is no index position, no report and no published ranking system. What you can do is make yourself the answer that is available: allow the crawlers to fetch you, state plainly on your own pages what you do and where, and make sure the third party sources an assistant is likely to read say the same thing about you as you do.

Does paying OpenAI or anyone else get my business into ChatGPT answers?

No. There is no placement to buy in a generated answer and anybody offering one is selling something they do not have. OpenAI documents a separate crawler for validating pages submitted as ads, which is a different surface from the answer itself. Treat a guaranteed placement offer as the end of the conversation.

Why does ChatGPT recommend my competitors and not me?

Usually because the sources it read do not contain you. Ask it the same question and look at the citations rather than the prose. If it named a directory, a local roundup or a comparison post, and you are not on any of them, that is the answer and the fix is getting listed rather than rewriting your homepage.

How long does it take for a change to show up in an assistant?

Nobody can tell you, and any specific number is invented. It depends on when the source you changed is next crawled and whether it gets retrieved for that question at all. Treat it as weeks rather than days, ask from fresh sessions rather than one, and look for a change that holds rather than a single good answer.

Is my own website or my listings elsewhere more important?

For a small or local business, the listings usually decide more. Your website is one source among many and it is the one an assistant has the least reason to treat as neutral. Your site still has to be fetchable and clear, because it is where you control the wording, and it rarely wins the recommendation on its own.

A new prompt, most days One working prompt for a real SEO or AI visibility job, what to change in it, and a worked example. No sequences, no offers dressed as newsletters.

Unsubscribe in one click. We never pass your address on.

Run your first audit
in about a minute

Free account, no card. Paste your URL and get a real, scored report of your AI and search visibility.

Measuring rankings in