The short version: Most small business owners have never typed their own business name into ChatGPT or Perplexity to see what comes back, and it shows. I ran the test on myself, found gaps I didn't expect, and fixed them without touching a single meta tag. The fix has almost nothing to do with SEO tricks and everything to do with being mentioned, in plain language, on sites that AI tools already trust.
The test I ran on my own business name
Back in January I sat down and asked ChatGPT, Perplexity, and Google's AI overview the same question three different ways: "who is a good AI consultant for small businesses in the UK," "recommend someone who helps small businesses with AI," and just my own name, "Lilach Bullock." Three tools, three phrasings, nine answers.
ChatGPT mentioned me by name for the third search but not the first two. Perplexity pulled in three other consultants I'd never heard of before it got to me, and the sources it cited were mostly LinkedIn posts and a G2 style review directory. Google's AI overview didn't mention me at all for the generic query, even though I rank on page one organically for a version of it.
That gap between "ranks well on Google" and "gets recommended by AI" is the bit nobody warned me about. I'd spent years building organic rankings and treated that as the whole job. It isn't anymore.
Why this is a different game to normal SEO
Traditional search ranks pages. AI chatbots recommend entities: people, businesses, brands. The tool isn't deciding which page deserves the click, it's deciding who it trusts enough to put a name to. That trust seems to come from being mentioned consistently across places the model already treats as reliable, not from how well one page on your own site is optimised.
An Ahrefs analysis of ChatGPT's citation patterns found Wikipedia, Reddit, and YouTube among the most frequently cited sources across a huge sample of AI answers, far more than most business websites. That tells you something uncomfortable: your beautifully written homepage might matter less to an AI recommendation than a Reddit thread where three strangers mentioned your company by name.
I've watched clients spend months rewriting title tags for a shift that was never going to come from title tags. The uncomfortable truth is that a lot of the AI-SEO advice going round right now is repackaged 2015 SEO with the word "AI" stapled on, and it doesn't move the number that matters, which is whether a chatbot says your name out loud when someone asks for a recommendation.
How to run this audit yourself, in one sitting
This takes about forty-five minutes and needs nothing but the free versions of the tools. Here's exactly what I do, and what I'd suggest you do before you touch anything else on your marketing to-do list this month.
- Step 1: List five real questions your customers would type, not your business name. Think "best [your service] for small business in [your city]," "who can help with [specific problem you solve]," "recommend a [your job title] for a startup."
- Step 2: Ask each question in ChatGPT, Perplexity, and Google's AI overview. Use a fresh chat each time so previous context doesn't skew the answer. Screenshot every response.
- Step 3: Note who gets named and who gets cited. Being named is one thing. Being cited as a source (with a link shown) is another and usually rarer.
- Step 4: Ask a follow-up question inside the same chat, something like "why did you recommend them" or "where does that information come from." This is the step almost nobody does and it's the most useful one, because the tool will often tell you which sites it pulled from.
- Step 5: Search your own business name plus the word "review" or "recommend" on Reddit and in Google to see what's being said about you in the wild, on pages you don't control.
- Step 6: Write down the gap between where you rank on Google and whether AI tools mention you at all. That gap is your real to-do list for the next quarter.
When I did step 4 with Perplexity, it told me flat out that two of its three sources for my competitor set were LinkedIn articles and one directory listing. Not their website. Not their case studies page. A LinkedIn post from 2023.
What changes the answer
Schema markup, FAQ page structure, "AI-optimised" content, these things help a little, mostly by making it easier for a crawler to extract a clean, quotable sentence. They don't create trust from nothing. Trust seems to come from three things I can now point to with some confidence after running this test across a handful of client accounts:
- Being mentioned by name on sites the model already trusts. That's press coverage, being quoted in a trade publication, appearing on a "best of" list someone else wrote, or a genuine mention on Reddit or a niche forum where your industry talks.
- Consistency of the same facts everywhere. Same job title, same one-line description of what you do, same city, repeated across your site, your LinkedIn, your directory listings, and any press mentions. Inconsistent facts seem to make models hedge or drop you entirely rather than guess.
- Clear, direct answers on your own pages that a model can lift as one clean sentence. Not clever copy. A flat, quotable statement near the top of the page that answers the exact question someone would type.
That last point is exactly why I now open every long piece I write with a short, bolded, direct answer before the article gets going, the way this post starts. It's a small structural change and it changes what a model can pull.
The content that earns mentions
This is where a lot of small business owners waste effort. They chase mentions by pitching journalists cold, which rarely works below a certain size of business. What worked better for me was building content worth talking about, the kind of thing other people reference because it's useful, not because you asked them to.
I think about brands that get cited constantly without begging for it. Look at how GoPro built its marketing strategy around customer-generated proof rather than its own claims, that's the same principle: the mentions come from what other people say about you, not what you say about yourself. Or how Joe Pulizzi's approach to content marketing was built on giving away useful material until people started citing it as the reference point in the industry, that's exactly the behaviour that gets picked up by AI models now, years later.
Interactive tools do this well too. A calculator or a quiz that solves one specific problem tends to get linked from forums, roundup posts, and Reddit threads in a way a normal blog post rarely does, which is worth reading about if you haven't seen how interactive calculators convert and get shared. Same logic applies to a consistent newsletter, the sort of steady, recognisable output Mailchimp built its brand strategy around, because a newsletter with a stable name and format is easy for other people, and models, to reference back to consistently.
What I did after the audit
Three things, none of them technical. I updated every bio I could find (LinkedIn, guest post author boxes, three old directory listings) so the one-line description of what I do was word-for-word identical. I pitched two trade newsletters with a specific stat from my own client work rather than a generic pitch, and one of them ran it. And I rewrote the top of my five most-visited pages to open with a direct, bolded answer to the exact question that page targets, instead of a warm-up paragraph.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Six weeks later I reran the same nine searches. ChatGPT mentioned me for two of the three phrasings instead of one. Perplexity's citation list for my name included my own site for the first time instead of only third-party mentions. Small movement, but it's the first movement I've seen from anything other than raw time and organic backlinks, and it took about a day and a half of actual work.
None of this replaces the slower work of building a reputation over years. But if you run a small business and you've never checked whether AI tools even know you exist, that's the gap worth closing before you spend another penny on ads or another month rewriting page titles nobody's model is reading closely.
If you want a second pair of eyes on this, an AI consultant for small business can usually spot the gap in an afternoon, because most of it is pattern recognition once you've seen it done a few times.
Frequently asked questions
How do I know if ChatGPT is recommending my business?
Open a fresh chat and ask the exact question a customer would type, not your own business name. Try three or four phrasings. If your business doesn't appear in any of them, ask a follow-up question about where the recommendations came from, the tool will often tell you which sites it trusts.
Does schema markup help AI chatbots recommend my business?
A little, mostly by making it easier for a crawler to pull a clean sentence from your page. It doesn't create trust on its own. Being mentioned by name on sites the model already trusts, like press coverage, directories, or genuine forum mentions, tends to matter far more.
Is this different from normal SEO?
Yes. Normal search ranks pages based largely on your own site's signals. AI recommendations lean more on whether your name and facts appear consistently across other sites the model already trusts, so a page one Google ranking and an AI mention don't always move together.
How often should I check this?
Once a quarter is enough for most small businesses. Run the same five questions each time, screenshot the answers, and compare. The change is usually slow, so monthly checks mostly just show you noise.
Related reading: The AI Subscription Stack That's Quietly Draining Small Business Budgets in 2026 and How to Build an AI-Powered Calculator That Brings You Leads.
If you want the full breakdown, here is everything I know about AI marketing.
Before you change anything, free AI Visibility Checker so you have a baseline to measure against rather than a guess.
