How to Find Out Which Competitors AI Is Recommending

Most coaches and consultants name their competitors wrong. They list the people they see on LinkedIn or the names that come up at conferences. Then a buyer opens ChatGPT, asks a discovery question, and gets recommended three people the coach has never heard of.

Key takeaways

• Your real AI competitive set is the recurring list of names that appear when you ask ChatGPT, Claude, Perplexity, and Microsoft Copilot the discovery questions a buyer would actually type.
• That list is almost never the same as the competitors you track on social media or industry roundups.
• You need to run the same 8 to 10 questions across all four engines, write down every name returned, and look for who shows up 3 or more times.
• The names that recur are your real competitors. The names that show up once are noise. The names that show up everywhere are the citation pattern you need to match.
• A niche site can outrank a much bigger brand in AI answers because AI engines reward specific, clearly-structured pages, not size. BakingSubs has earned 162,500 Microsoft Copilot citations, with 112,500 of those landing in just the last three months, without being a famous brand.

Why your gut list of competitors is almost always wrong

You think your competitors are the people you see on LinkedIn. AI engines do not look at LinkedIn the way you do. They scan structured pages, author bios, and topical coverage, then pick the experts whose content best matches the buyer's question.

This means the person you watch every week may not be the person AI recommends when a buyer asks "who is the best leadership coach for first-time tech founders." The recommendation often goes to someone with less social presence but a tighter, clearer site.

Take an illustrative scenario. Yusuf is an executive coach in Austin who works with engineering managers stepping into VP roles. He has watched the same three coaches on LinkedIn for two years, assuming those were his competitors. When he ran his first set of AI discovery questions, none of those three names showed up. Instead, the same two coaches appeared in 7 of his 10 queries: one based in Toronto whose blog covers IC-to-manager transitions in depth, and one in Berlin who had written 14 posts about engineering leadership reviews. Yusuf had never heard of either of them. They were his real AI competitive set.

The 4 step process to surface your real competitors

Start by writing down the questions a buyer would type into an AI engine when they are looking for someone like you. Not branded questions. Not questions that contain your name. The questions a stranger asks when they have a problem and want a recommendation.

Then run those questions through each of the 4 engines and record what comes back. Here is the full process.

Step 1: Write 8 to 10 discovery questions. These should match what a buyer types right before they hire someone. Examples for a health coach: "who is the best functional medicine coach for women with hashimoto's", "what coach helps with perimenopause weight gain", "best health coach for autoimmune issues near me". Notice each one names a specific problem and a specific person. Vague questions like "best health coach" return vague answers.

Step 2: Run each question through ChatGPT, Claude, Perplexity, and Microsoft Copilot. Use fresh chat sessions so memory does not skew results. Copy the names returned into a spreadsheet. Note which engine returned which name. Perplexity will often show source URLs. Save those too because they reveal which pages got cited.

Step 3: Count the frequency. Sort the spreadsheet by name. Anyone who shows up 3 or more times across your 8 to 10 questions is a real AI competitor. Anyone who shows up once or twice is probably a fluke. Anyone showing up in 6 or more questions is dominating your niche in AI search and is the model to study.

Step 4: Look at why they got cited. Click through to the source pages each engine pulled from. You are looking for: a clear author bio with credentials, a single specific topic the page is about, structured FAQ sections, and an obvious answer to a discovery question in the first paragraph. These are the patterns ChatGPT actually looks for when recommending experts.

What recurring names tell you about your positioning

When the same competitor appears across all 4 engines for 7 of your 10 questions, two things are true. First, they have built strong topical authority on the exact problem your buyers are searching for. Second, the AI engines have decided they are the canonical answer for that niche, and they will keep getting cited until someone publishes content that beats them on specificity.

That is the opening. AI engines do not lock in a single expert forever. They pick whoever has the clearest, most specific page right now. A niche site that publishes 20 posts about one narrow problem can displace a generalist who has been around for a decade. This is the whole basis of the Citation Cluster Method: match the recurring competitor's topical coverage, then go deeper on a sub-angle they have not covered well.

If the recurring competitor is a generalist coach who covers 12 different topics, your move is to pick one of those topics and own it. If they are already a niche specialist, your move is to go more specific, by sub-population, geography, or stage of the problem.

What to do when no one recurs (the empty niche case)

Sometimes you run the 8 to 10 questions and nothing recurs. Different names show up in different engines. No clear pattern. This is the empty-niche signal and it is the best outcome you can find.

It means your niche has no dominant AI-cited expert yet. The engines are guessing. Whoever publishes the first clearly-structured cluster of content on that problem becomes the default recommendation. This is what happened with BakingSubs in its niche. There was no entrenched citation winner, so a niche site built around 1 specific topic could rack up 162,500 Microsoft Copilot citations to date, with 112,500 of those in the last three months as the engines settled on it as the canonical source. The full case study walks through what that looked like.

If you run your questions and see an empty niche, move fast. Empty niches do not stay empty. The first specialist site to publish a tight cluster will lock in the citation pattern.

How to set up the reconnaissance the right way

A few rules to make the process reliable.

Use fresh chat sessions or incognito mode. ChatGPT and Claude both have memory that personalizes answers. If you are signed in, you will get answers tuned to your past conversations, not what a buyer sees.

Ask the same exact question across all 4 engines, word for word. Different phrasings return different competitors. You want apples to apples.

Do not include your own name or brand in any question. You are trying to see what a buyer sees, not what an engine returns when prompted with you.

Save the source URLs Perplexity shows you. Those are the exact pages the engine used to build its answer. Read them. They are your roadmap for what to publish.

Repeat the whole exercise every 90 days. The recurring names will shift as new content gets published and indexed. The 90-day refresh is also how you measure whether your own work is moving you into the recurring list.

The shortcut: let the Visibility Check do this for you

Running 8 questions across 4 engines and tagging every name takes 2 to 3 hours if you do it carefully. The AI Visibility Check does the same thing automatically. It runs 8 discovery-intent questions per engine, sorts the recurring names into a real competitive set, and tells you which of the 4 outcomes you are in: Invisible, Mixed, Winning, or Empty Niche.

The output you get is the same thing you would build by hand in a spreadsheet, except it surfaces the patterns you would miss when reading 32 answers in a row. It also tells you whose pages to study, which is the next move after you know who your real competitors are.

If you want to do the reconnaissance yourself first, do it. The manual version teaches you what AI engines pay attention to. The automated version saves you the next 90-day refresh.

Frequently asked questions

How many discovery questions do I need to run to get a reliable competitor list?

8 to 10 is the floor. Below 8 and you will see too much noise to spot a pattern. Above 12 and you start asking questions a real buyer would not type, which pulls in irrelevant names. Use questions that name a specific problem, a specific population, or a specific geography.

Should I worry about competitors who only show up on one engine?

Not really. A name that appears only on Perplexity but not on ChatGPT, Claude, or Copilot is usually getting cited for a single high-ranking source page. That can shift overnight. Focus on the names that recur across at least 3 of the 4 engines. Those are stable citation patterns.

What if the same competitor shows up for half my questions but I have never heard of them?

That is the most common and most useful finding. It usually means they are a specialist who has gone deep on a narrow sub-niche while you have been watching the wrong people on social media. Read their site carefully. Look at how their pages are structured, what topics they cover, and how their author bio is written. That is the model. Why ChatGPT recommends your competitor and not you walks through what those pages tend to have in common.

Does the size of the competitor's business matter?

No. AI engines do not measure traffic, revenue, or team size. They measure how clearly a page answers a buyer's question and how much topical coverage the site has on that exact problem. A solo coach with 30 tight, specific posts on one niche regularly beats a 50-person agency with a generic blog. This is why AI often recommends the smaller competitor.

How long until I start showing up in the recurring list myself?

Plan on 8 to 16 weeks from the day you publish the first post in a tight topical cluster, assuming the cluster is structured the way AI engines reward. The 4 engines index and re-weight at different speeds. Perplexity and Copilot tend to move faster than ChatGPT. The full mechanism is in the Citation Cluster Method writeup.

What to do this week

Pick 8 questions a buyer types when they are looking for someone like you. Open fresh sessions in ChatGPT, Claude, Perplexity, and Microsoft Copilot. Run all 8 questions through all 4 engines and write down every name that comes back. Sort by frequency. The names that recur 3 or more times are your real competitive set.

If that sounds like 3 hours you do not have this week, run the AI Visibility Check instead. It produces the same competitive set in about a minute and tells you which of the 4 visibility branches you are in. Either way, the goal is the same: stop guessing who you compete with in AI search, and start working from the actual list buyers see.