Key Takeaways
- Ask the questions your buyers ask, not the ones you would ask. The source list changes completely between a category question and a hiring question, and only one of those two matters to your pipeline.
- Record the source, not just the names. Who got named tells you the score. Which website the answer leaned on tells you where the work has to happen.
- Run every question more than once. Assistants vary between runs. A domain that shows up once is noise, and a domain that shows up in most runs is a fixture worth pursuing.
- Expect a short list to do most of the work. A handful of source families usually accounts for the bulk of the citations in any one niche, which is good news, because the target is small.
- Turn the list into a place to be described. Being cited is not the goal. Being described on a site that gets cited is what puts your name into the answer.
When an assistant answers a question about your niche, it is often leaning on a small set of websites. Those sites are doing the describing, and whoever they describe is who gets named.
Most people never look at that list. They read the answer, notice they are not in it, and go rewrite their homepage, which is the one place the answer was not built from.
This is how to find the list yourself in an afternoon, what to write down while you do it, and how to turn a column of domains into something worth acting on.
How do you find out which websites AI cites in your niche?
Ask the assistants that show their sources the questions your buyers actually ask, then record every domain they link to and count the repeats. That is the whole method, and it needs no software.
It works because citations are visible on some tools by design. Perplexity's help center describes an approach built around retrieving and citing current sources, which means the source list is right there in the answer rather than being something you have to infer.
Google's guidance on AI features in Search describes AI Overviews as carrying links to supporting pages for the same reason. Between those two, plus ChatGPT when it searches, you can see the sources behind a large share of the answers your buyers are getting.
The method, step by step
Write the questions first, run them across two or three tools, record every cited domain, then repeat the whole set a few days later. The repetition is not optional, because a single pass measures the assistant's mood as much as your niche.
Write twenty to twenty-five questions in your buyer's words, spread across the range: what the service is, how to choose, how much it costs, and who to hire for a specific situation. Include at least one with a city in it. Do not write them the way you would describe your own work, because your buyers do not use that vocabulary.
Then run each question on a tool that shows citations, and open the sources rather than trusting the labels. A domain that looks like an authority is sometimes a thin listing page, and a domain you have never heard of is sometimes the piece the whole answer was built from. What a citation does and does not mean is in what it means when AI cites a website.
What to record for every answer
Four things: the question, whether you were named, which source the answer leaned on, and how many times you have seen that source before. The fourth column is the one that turns a pile of notes into a finding.
Source families rather than named websites, because we have no measured domain list we could publish honestly yet. The six rows are illustrative, written to show the shape of the record rather than to report a result. In this example the directory family was seen four times, the association and the trade publication three each, the coach's own site and the review platform twice each, and the forum thread once.
Record whether you were named even when the answer is no, because the pattern of where you are absent is more useful than a single count. Being absent from every category question and present on two narrow ones is a completely different situation from being absent everywhere.
How to read the list once you have it
Sort by how often each source reappears, then split the top of that list into three kinds: places that describe businesses, places that explain the category, and places that rank or list providers. Each kind needs a different move from you.
| Kind of source | What it gives an assistant | Can you appear on it | Your move |
|---|---|---|---|
| Directory or listing | Structured details it can cross-check | Yes, usually by claiming a profile | Claim it and write a real description, not a category |
| Professional association | Credential claims it can trust | Yes, if you qualify for membership | Join and complete the public profile fully |
| Trade publication | Quotable authoritative phrasing | Sometimes, as a source or contributor | Offer a specific, checkable point of view |
| Review platform | Independent evidence that clients exist | Yes | Ask clients to describe the problem you solved |
| A provider's own site | A clear self-description worth quoting | Yes, this one is yours to write | Write the page that answers one question completely |
| Forum thread | Ordinary language about real situations | Rarely, and never by self-promotion | Read it for vocabulary, then use those words |
The last row is the cheapest thing on this page. Threads where buyers describe their own problem are a free vocabulary list, and using their phrasing on your pages is the single easiest way to become matchable. What that source mix looks like across the coaching market is in the websites AI cites in the coaching niche.
What to do with the sites you find
Get described on the top three, in the words your buyers use, and leave the rest alone. A long list is a distraction, and the tail almost never repays the effort of chasing it.
Being described matters more than being linked. A profile that says you coach nonprofit directors in Denver does more than a link from a page that says nothing about you, because the first one gives an assistant a fact it can match and repeat, and the second gives it nothing to say.
Then re-run the same questions a month later and compare. You are watching for two things: your name entering answers where it was absent, and your own site starting to appear in the source list. The first is the outcome and the second is the leading indicator. The method for tracking that without buying anything is in how to check if AI recommends you.
Our free check runs 25 real buyer questions from your niche at $0 and emails a plain-English readout within the hour, including who is being named in the answers you want. It runs as soon as you ask for it.
Questions we hear the most
How do you find out which websites AI cites in your niche?
Ask assistants that show their sources the questions your buyers ask, then write down every domain they link to. Repeat each question several times and count how often each domain reappears, because one run is not a measurement.
Which AI tools actually show their sources?
Perplexity shows citations for almost every answer, and Google AI Overviews carry links to supporting pages. ChatGPT shows sources when it searches the web, and shows none when it answers from what it already learned.
How many questions do I need to run to get a useful list?
Twenty to twenty-five questions covering the range from category questions to hiring questions, each run a few times. Fewer than that and you are looking at variation between runs rather than a pattern.
Do the cited websites differ between AI tools?
Yes, quite a lot. Engines retrieve differently and weight sources differently, so the overlap between two tools is partial. Domains that appear on every tool are the ones worth spending effort on.
Should I try to get a link from every site on the list?
No. A link is not the point. What matters is being described on those sites in the words your buyers use, which is often a profile, a listing, or a quote rather than a link.
What if my own website never appears as a source?
That is the common case and it is fixable. It usually means no page of yours answers a specific question directly enough to be worth quoting, rather than that anything is technically wrong with your site.
How does the AnswerHalo free check compare to doing this by hand?
It runs the same idea for you at $0. We ask ChatGPT 25 real buyer questions from your niche and email a plain-English readout within the hour: who got named, and the first thing to fix. It runs as soon as you ask for it.
What does the paid audit add on top of that?
The $500 audit lands by this time tomorrow and widens the picture: more questions, more assistants, the source list behind the answers, and a prioritized set of fixes. Everything arrives by email, with no calls at any stage.
Your buyers are already asking. Find out what AI tells them.
The free check asks ChatGPT 25 real buyer questions about your niche and sends you the report within the hour, at $0. If it shows you are already getting named everywhere, we will say so in plain words.
Get my free check →