How to Get Found on ChatGPT as an Expert

ChatGPT now decides who gets the call before the buyer ever visits your site. When someone types "best executive coach for first-time founders" or "trauma-informed therapist in Austin," ChatGPT picks three to five names. If you are not one of them, you do not exist for that buyer.

The good news: the way ChatGPT picks is mechanical, not magical. Once you understand the five ways it finds experts, you can audit your own site and fix the gaps in a weekend.

Key takeaways

ChatGPT discovers experts through five separate channels: live web browsing, training data inclusion, custom GPT recommendations, connector and plugin mentions, and structured data cues. Most expert sites are visible through zero of them.
The single biggest reason coaches and consultants get skipped is not bad SEO. It is structural ambiguity. ChatGPT cannot tell who you are, who you help, or whether you are a real person or a brand page.
BakingSubs earned 144,321 Microsoft Copilot citations across a quarter using the same patterns that work on ChatGPT, with no ads, no backlinks, and no social.
A real person with a clear niche beats a polished agency site nine times out of ten in ChatGPT recommendations, because language models are trained to surface humans for human-shaped questions.
The fastest way to find your gaps is to ask ChatGPT the eight questions your buyers would ask, then read the answers like an audit, not a horoscope.

Why ChatGPT recommendations matter more than Google rankings now

Buyers who used to type "executive coach for tech founders" into Google now type the same thing into ChatGPT. They get a short list of names, a quick paragraph on each, and a recommendation about who fits their situation best. They click one or two. They never see the search engine results page.

This is happening across every expert-led category. Mediation, therapy, coaching, fractional CMO work, niche consulting, expert-led ecommerce. Anywhere a buyer used to research a person, an AI search engine is now doing the shortlisting.

The shift matters because Google rewarded volume and backlinks, while ChatGPT rewards clarity and specificity. A site that ranked page one on Google for ten years can be completely invisible to ChatGPT today. I see it happen most often with established consultants whose sites were built between 2015 and 2020. They are well-known. They have referrals. They have no ChatGPT presence. Their younger competitors do.

If you are still thinking about this as "SEO for AI," you are framing it wrong. It is closer to being a recommended vendor in a peer's slack channel. The question is not "did you rank?" The question is "did the model decide you were the right person to name?"

The 5 ways ChatGPT actually finds experts

ChatGPT does not have one discovery mechanism. It has five, and they work differently. Most experts try to optimize for one and ignore the other four, which is why their visibility stays flat.

1. Live web browsing citations

When ChatGPT has browsing turned on (which is the default for most paid users in 2026), it does a real-time search and pulls in pages it can quote from. These pages need to load fast, be easy for the model to parse, and contain a clear answer near the top.

This is the channel most people think of when they hear "AI search." It is also the easiest to win, because the bar for what counts as a citable page is lower than the bar for ranking on Google. A clear H1, a direct answer in the first paragraph, and a credible author bio will often beat a page with 47 backlinks but a vague intro.

2. Training data inclusion

Some of what ChatGPT "knows" about you comes from the data it was trained on, which includes a frozen snapshot of the web. If your site, your name, and your specific niche appeared in that snapshot in a way the model could pattern-match, you get pulled into answers even when browsing is off.

You cannot control what is already in the training data. You can control what goes in next time. Every time OpenAI refreshes its training corpus, sites that have published consistent, specific, well-structured content about a clear topic get rolled in. Sites that published five vague posts in 2022 do not.

3. Custom GPT recommendations

This one most experts have never thought about. There are tens of thousands of custom GPTs in the GPT store, many of them built around specific problems ("Find me a financial planner," "Help me pick a leadership coach," "Marketing strategy for SaaS founders"). When users ask these custom GPTs for recommendations, the GPTs draw from both their custom instructions and live web data.

If a custom GPT's creator has built it around a niche you serve, and your site has the structural clarity to be picked up, you can become a recurring recommendation inside that GPT. I have seen consultants get a steady trickle of qualified inbound from a single custom GPT that recommends them by name in its default response. They did not build the GPT. They just made themselves easy to recommend.

4. Connector and plugin mentions

ChatGPT now connects to other tools through connectors, plugins, and the broader integration layer. When a user asks ChatGPT to "find experts on X" through a connected service, the model leans heavily on directories, marketplaces, and structured listings.

If you are listed in a well-structured directory in your niche, with a clear bio and a specific positioning statement, you increase the chance that a connector-driven query surfaces you. This is the closest thing to old-school directory SEO that still works. The catch: only specific, niche-focused directories help. Generic "find a coach" sites with thousands of unfiltered listings do not.

5. Structured data cues

This is the hidden layer. ChatGPT and the underlying language models read structured data on your site, which is the hidden tags that tell AI engines what each part of a page is. The most important ones for experts are Person schema (telling the model you are a real human), Service schema (telling it what you actually do), and FAQ schema (giving it ready-to-quote answers).

Most expert sites have none of this. They have a homepage with a photo and a paragraph. ChatGPT looks at the page and cannot confidently tell whether "Sarah Chen" is a person, a brand, a fictional character, or a domain name. So it skips you and recommends someone whose page makes the answer obvious.

What ChatGPT looks at on your actual website

When ChatGPT decides whether to recommend you, it weighs four signals harder than the rest. Get these right and you get cited. Get them wrong and the model treats you as background noise.

Who you are, stated plainly. Your homepage and about page need to say "I am a [role] who helps [specific audience] do [specific outcome]" in language a 12-year-old could parse. Not "leveraging holistic frameworks for transformative growth." Just the actual sentence.

A specific niche, repeated consistently. ChatGPT loves specificity. A coach who says "I help second-time founders rebuild after a failed exit" will beat a coach who says "I help leaders grow" every time, because the first one is matchable to a buyer's exact question and the second one is matchable to no one's.

Author signals on every piece of content. Each blog post should have a visible author byline, a short bio block at the bottom, and ideally a Person schema tag in the page source. ChatGPT increasingly weights "is this a real human with a track record" as a primary filter, especially for advice-shaped queries.

Topical depth in one place. A site with 30 posts on one specific topic crushes a site with 200 posts on 40 topics. ChatGPT pattern-matches "this site is the go-to source on X" through density, not breadth. This is the whole basis of the Citation Cluster Method: pick one narrow topic, publish a dense cluster of related posts around it, and let the model conclude you own that niche.

Why most expert sites are invisible to ChatGPT (the real reasons)

Most expert-led sites fail the ChatGPT test for boring structural reasons, not strategic ones. I see the same five patterns over and over.

The first is a homepage that talks about transformation without saying who is being transformed or into what. ChatGPT cannot recommend a vibe. It can recommend a person who helps a specific group with a specific problem.

The second is a blog full of "5 mindset shifts" posts that could appear on any coach's site. There is nothing in the writing that locks the model into recommending the author specifically. Compare this to a post titled "What to do when your CFO quits two weeks before a board meeting." That one is unmistakable. Only one type of person writes that post.

The third is the missing About page or the About page that reads like a press release. ChatGPT needs a clear, first-person, plain-language story about who you are, where you trained, and what you have done. Bullet points are fine. Vague self-praise is not.

The fourth is the technical layer. No Person schema. No FAQ schema on FAQ pages. No structured Service descriptions. The page looks fine to a human and is invisible to the model. There is a longer breakdown of what helpful content actually means in AI search that gets into this in more depth.

The fifth, and the one nobody wants to hear, is that the site has not published in 14 months. AI search engines pattern-match on freshness for advice-style queries. A site that last published in 2024 looks dormant to the model, regardless of how good the archive is.

The 8-question diagnostic: run this in 20 minutes

Here is the exact diagnostic I use to figure out where someone stands on ChatGPT visibility. It is the same one that powers the AI Visibility Check, and you can run a version of it yourself right now.

Open ChatGPT. Turn browsing on. Ask it these eight questions, one at a time, using your actual niche and geography. Read the answers carefully.

"Who are the top [your role] who work with [your specific audience]?"
"I am looking for a [your role] who specializes in [your specific problem]. Who should I consider?"
"What are the credentials I should look for in a [your role] for [your audience]?"
"Compare the top three [your role] who work with [your audience] in [your geography or remotely]."
"Is [your name] a good [your role] for [your audience]?"
"What do people say about working with [your name]?"
"What is [your name] known for?"
"Who else does work similar to [your name]?"

The pattern of answers tells you where you sit. If your name does not appear in questions 1 through 4, you are invisible on discovery queries. If ChatGPT cannot answer questions 5 through 8 about you with specificity, your site has not given the model enough to work with. If your competitors show up by name in questions 1 through 4 and you do not, you have a competitive visibility gap, not just an awareness gap.

I cover the full eight-question methodology and how to read the four possible outcomes in the Visibility Check writeup. The short version: most expert sites fail questions 1 through 4 hard and pass questions 5 through 8 only partially. That combination means you are findable if a buyer already knows your name, and invisible to every buyer who does not.

The proof: what 144,321 citations actually looked like

I built BakingSubs as a niche site in baking ingredient substitutions. In 12 months, with no ads, no backlinks campaign, and no social media presence, it earned 144,321 Microsoft Copilot citations across a quarter and over 5,000 daily Google clicks at peak. Every single citation came from the same patterns I just described: clear structural signals, a tight topical cluster, plain-language answers near the top of every page, and author depth on a narrow subject.

The reason the BakingSubs story matters for you is that baking substitutions and expert services are structurally identical from an AI engine's point of view. Both are advice-shaped queries. Both reward specificity. Both punish sites that try to be everything to everyone. The full BakingSubs case study walks through the exact content cluster, but the headline takeaway is simple: a tightly clustered, plainly written, author-credentialed site beats a polished brand site every time.

Here is the part most people miss. BakingSubs is a niche site, not a brand. It is well-established within its niche, but it is not a household name. If a niche site can earn 144,321 citations in one engine, an expert with real credentials and a clear specialty can absolutely earn meaningful visibility in their own category. The mechanism is the same.

What to do this week

If you have 90 minutes this week, here is what gets you the most visibility lift per hour of effort.

Rewrite your homepage opening to state, in one sentence, who you are, who you help, and what specific outcome you deliver. No metaphors. No "transformation." Just the literal sentence.

Add a Person schema block to your About page. If you use WordPress or Webflow, there are free plugins that handle this. The tag tells ChatGPT and the other engines that this page is about a real human, which changes how they weight everything else on your site.

Pick one specific question your buyers ask, and write a 1,500-word post answering it in plain language. Put the direct answer in the first paragraph. Add an author bio at the bottom. Publish it.

Run the 8-question diagnostic above. Save the answers in a doc. You now have a baseline.

For a deeper structural rebuild, the Citation Cluster Method writeup walks through how to plan a 20-to-30 post cluster that gets a site cited consistently across all four engines. And if you want the audited version of this diagnostic with specific fixes mapped to your site, the AI Citation Playbook is built around the same framework.

Frequently asked questions

How long does it take to get found on ChatGPT as a new expert?

For live web browsing citations, you can show up within a few weeks of publishing a tightly focused, well-structured cluster of posts in a narrow niche. For training data inclusion, it takes longer, because you have to wait for the next refresh cycle. Most experts I have seen go from invisible to showing up in browsing-based recommendations within 8 to 14 weeks of starting a focused content cluster.

Do I need backlinks to be recommended by ChatGPT?

No. BakingSubs earned 144,321 Microsoft Copilot citations with no backlink campaign at all. ChatGPT and the other engines weight structural clarity, topical density, and author signals far more than link count. Backlinks still help, but they are nowhere near the top of the priority list for expert-led sites. If you are spending time on link outreach instead of publishing focused content, you are optimizing for the old game.

Is ChatGPT visibility different from Google AI Overviews?

Yes, though the underlying signals overlap. ChatGPT pulls heavily from its training data and live browsing, while Google AI Overviews lean more on Google's own ranking signals plus structured data. A site optimized for ChatGPT will often show up in Overviews too, but the reverse is less reliable. Each engine has quirks, which is why running the 8-question diagnostic in each one separately is worth the time.

What if my niche is too small for ChatGPT to know about?

A small niche is an advantage, not a problem. ChatGPT recommends specific names for specific problems, and the narrower the problem, the easier it is to dominate the recommendation. The risk is not "my niche is too small." The risk is "my niche is too vague." A coach for "ambitious professionals" has no chance. A coach for "engineering managers being promoted to director for the first time" has a clear lane.

Should I be on every AI engine or focus on ChatGPT first?

ChatGPT has the largest share of expert-discovery queries right now, so it makes sense as the first priority. But the patterns that get you cited by ChatGPT will also get you cited by Claude, Perplexity, and Copilot, because the underlying signals are similar. Once you have the structural fundamentals in place, your visibility tends to grow across all four engines together. The Perplexity-specific playbook and the ChatGPT-specific recommendation factors cover the per-engine quirks.

The cheapest thing you can do today is run the 8-question diagnostic on your own name and niche, and write down what you find. Most experts have never done this once, which means they are guessing about how AI engines see them. Twenty minutes of looking will tell you more than a year of opinions, and it will point you straight at the fix that matters most. From there, you decide whether to handle it yourself or pull in the AI Citation Playbook to skip ahead.