ChatGPT, Claude and Perplexity recommend the experts they can describe clearly for the exact question a buyer asks. To be one of them, describe yourself the same way everywhere you appear, be specific about who you help and with what, publish answers to the questions buyers already ask you, and make sure your expertise and client proof live in places AI can read, such as YouTube, your website, podcasts and public reviews.
How buying used to work, and what changed
If you needed a consultant a few years ago, you'd probably go to Google and open 10 or 15 tabs. You'd read websites and reviews, watch a few videos and check a LinkedIn profile or two. Then you'd build your own shortlist. The research was yours, so an expert with a strong website and a clear profile had a fair chance to make a good impression.
Now the buyer asks the assistant and the assistant does the research. It reads the websites, the reviews and the video transcripts, then comes back with one or two names. People trust those answers because they talk to these tools 10, 15, 20 times a day. A recommendation from something you speak to that often feels a lot like a recommendation from a friend. That's why I call AI the new word of mouth.
LinkedIn and Google still matter. What's new is that a lot of recommendations now happen in a room most experts haven't walked into yet.
The "who would you recommend?" conversation is moving into AI
AI tools aren't only recommending big brands. I heard about someone who asked ChatGPT where to buy a specialist honey. It recommended a small local beekeeper and gave his phone number. When the man rang, the beekeeper was floored, because he'd never had a customer find him through ChatGPT before.
You don't need to be famous to be the answer. You need to be clearly described.
Find out whether your name comes up when your ideal client asks ChatGPT, Claude or Perplexity for someone like you. Record what you find so you have a baseline. Prompt 1 below does this, and the tracker keeps the record.
In my business, I've recorded a baseline of what the main AI tools say about me. I'm checking it month to month to see whether the description changes, whether I get recommended, and whether any of it leads to sales conversations.
AI builds the shortlist, so how the internet describes you matters more than how good your content is
You've probably invested in a good website, worked on your LinkedIn profile and collected testimonials. All of that still matters, but only if somebody gets there. If the AI builds its shortlist and you're not on it, buyers never see the good work you do.
If a stranger read your LinkedIn profile and then your website, would they come away with the same description of you? If you describe yourself one way on LinkedIn and six other ways everywhere else, a machine has no clear idea what to recommend you for.
Write one positioning statement and use it everywhere you describe yourself: your LinkedIn headline and About section, your website, speaker bios and guest profiles. Prompt 3 checks how consistent you are right now.
Buyers type their whole situation, which is good news for specialists
People used to type something like "business consultant UK" into Google. With AI, they describe everything.
That one question carries the size of the business, the problem, the channel and the country, and the AI tries to match all of it. If your positioning is "I help businesses grow", there's nothing for it to match. Vague experts don't get recommended. Specific ones do.
Write down the detailed questions your ideal client would type or say to ChatGPT. Then check whether your content answers them.
I've stopped describing myself as a LinkedIn strategist. I now say I help established experts become visible, trusted and chosen. LinkedIn is still a big part of that, but my positioning now speaks to the buyer's problem rather than the platform.
Specific questions need specific answers
You already know most of the questions your clients ask. You hear them on discovery calls and coaching calls, and they turn up in your DMs. Most of them have never been answered anywhere public.
Block out an hour and write down every question a prospect might have, at three stages:
- They don't know they have a problem yet.
- They know the problem and are weighing it up. What's keeping them up at night?
- They're comparing you with a competitor or another kind of solution.
Most experts get to 50 or 60 questions. Each one can become a video, an article, a post or a podcast episode. AI can help you draft the answers, and then you rewrite them in your own voice with your stories and experience woven in. Prompt 4 pulls the questions out of your call transcripts.
I used AI to go through 105 of my sales calls to see why some people buy and others don't. It changed how I think about content, and the video this guide is based on started that way.
Watch how I did itI Used AI to Analyse 105 Sales Calls. Here's What I Found
Where your expertise lives decides whether AI can find it
If everything you know lives in your LinkedIn feed, you're exposed. AI tools can change which sources they trust very quickly.
YouTube is a source AI already draws on, and a video of you answering the questions clients ask is exactly the kind of content that turns up in AI answers. Other people talking about you carries weight too, because it isn't only you saying you're good at what you do. Podcast interviews, speaking slots, panels and guest articles all count. So do clients describing your work in their own words.
Add at least one source beyond LinkedIn that you own or that others publish. Then move your client results out of places AI can't see, like your DMs or a gated community, and onto public pages.
I was only on LinkedIn. Now I'm making long-form YouTube videos, I have a podcast, and I'm going after guest spots and stages on purpose. I'm also pulling client results out of my DMs and my community so they can be seen.
The prompts
Copy these into ChatGPT, Claude or Perplexity. Swap anything in [square brackets] for your own details. Run Prompt 1 in more than one tool, because they often give different answers.
1. The recommendation test
Shift 01 · run monthlyI [describe your ideal client's situation in their words, e.g. run a 20-person professional services firm]. [Describe their problem, e.g. Most of our clients come from referrals but that's slowing down.] Who could help me [the outcome they want] through [the channel or method, if they'd name one]? I'm based in [country or city]. Please give me the names of specific people or businesses you would recommend, and say why you recommend each one.
Write it the way your client would say it, not the way you'd describe your service. Don't include your own name.
2. The description check
Shift 01What do you know about [your full name], [your business name]? Tell me: - who they help - what they help with - what they're known for - where you found this information If you aren't sure about something, say so rather than guessing.
This tells you how AI describes you when it does know your name, and which sources it's leaning on.
3. The consistency audit
Shift 02Below are the ways I describe myself in different places. Read them as if you were a stranger, or an AI tool deciding whether to recommend me. LinkedIn headline: [paste] LinkedIn About section: [paste] Website homepage or About page: [paste] Speaker or guest bio: [paste] [Any other place, e.g. Skool, podcast bio]: [paste] 1. In one sentence, who do I help and with what, based on all of these together? 2. Where do these descriptions contradict each other or point in different directions? 3. Write one positioning statement, under 30 words, that I could use consistently across all of them. Keep my wording where you can.
Use the statement it gives you as a starting draft, then edit it until it sounds like you.
4. The buyer question extractor
Shift 04Below are transcripts or notes from my [discovery calls / sales calls / coaching calls / DMs] with potential clients. I help [who you help] with [what you help with]. Pull out every question these people asked me, or clearly wanted answered, in their own words. Group them into three stages: 1. Not yet aware they have a problem 2. Aware of the problem and weighing it up 3. Comparing me with another person or another kind of solution For each question, note how many times it came up. Don't invent questions that aren't in the material. [Paste transcripts or notes here]
If you record calls with a notetaker, paste a few transcripts at a time. Remove anything confidential first.
5. Answer drafter
Shift 04I'm [your name], and I help [who you help] with [what you help with]. Here is a question my clients often ask: "[Paste one question from Prompt 4]" Help me plan a short answer I could turn into a video, article or post. Give me: - the direct answer in two or three sentences - the points I should cover - where I should add my own story or a client example (leave these as gaps for me to fill) - a title that uses the words a buyer would type into ChatGPT or Google Don't write the full piece. I'll write it in my own voice.
Keeping the stories as your gaps to fill is what stops the answer sounding like everyone else's.
Your AI visibility tracker
Run Prompt 1 once a month in each tool you care about and log what happens here. After a few months you'll see whether you're moving onto the shortlist. Your entries save in this browser only. Use "Copy as table" to paste them into a spreadsheet.
| Month | Tool | Result | Question | Recommended instead / description | Enquiries |
|---|
Questions people ask
Can a small business or solo expert really get recommended by ChatGPT?
Yes. AI tools recommend whoever they can describe most clearly for the question asked. The beekeeper example in Shift 1 is a one-person business that got recommended with his phone number. What counts is being specific and consistent about who you help.
How do I check whether ChatGPT recommends me?
Use Prompt 1. Ask the detailed question your ideal client would ask, including their situation, problem, channel and location, and leave your own name out. Note whether you appear, who appears instead and how you're described. Do it monthly and log it in the tracker.
Does LinkedIn still matter if buyers are asking AI?
Yes. AI tools read LinkedIn and Google too. The risk is having all your expertise in one place. Adding YouTube, your own website, podcasts, speaking and public client reviews gives AI more to draw on.
What kind of content helps AI recommend an expert?
Answers to the specific questions buyers already ask you on calls and in DMs, written or recorded in your own voice with your own examples. Most experts can list 50 to 60 of these in an hour.
Is being mentioned by AI the goal?
Being mentioned gets you onto the shortlist. The goal is being the one they pick from it, so the trust you build in your content and conversations still decides the sale.
What to do next
Run Prompt 1 today and log the result. Then pick whichever of these fits where you are.
Watch me run this test on myselfI Asked AI to Recommend an Expert. It Didn't Recommend Me.
This guide is based onYour referrals are moving to AI: 5 Shifts for coaches and consultants
