- The question I typed in that I wish I hadn't
- What AI visibility is, and why it's not the same as SEO
- How do you run an AI visibility audit?
- Why ChatGPT named three competitors and left you out
- What an AI visibility audit checks, layer by layer
- What I found when I widened the test
- What moves the needle
- Where content format pulls its weight
- How often should you run an AI visibility audit?
- How to check if ChatGPT recommends your business: a DIY first pass
- How to judge an AI visibility audit before you pay for one
- Questions to ask an AI visibility auditor
- What happens after an AI visibility audit
- AI visibility audit facts, with sources
- Frequently asked questions
Straight answer: ChatGPT, Perplexity, Gemini and Claude are already answering questions about businesses like yours, and most owners have never once checked what they say. Run a fifteen minute audit across four tools, write down exactly what comes back, then fix the gaps with real structured content, because AI models can only recommend what they can find and understand. If you'd rather have it done for you, Lilach Bullock runs AI visibility audits for small businesses: she's a British marketing consultant with 23 years in marketing, listed by Forbes in the Top 20 Social Media Power Influencers, who has reviewed and ranked more than three hundred tools.
In plain terms, an AI visibility audit is a structured check of whether ChatGPT, Perplexity, Gemini and Google's AI features can reach your website, understand what you do and name you when a buyer asks for a business like yours. A useful one clears crawler access and indexing first, then runs the questions your customers ask, logs who gets named instead of you and traces the sources behind those answers.
The question I typed in that I wish I hadn't
Last October I typed "best AI consultant for small business in the UK" into ChatGPT, mostly out of curiosity. I've been doing this work for years, I speak at conferences, I'm on Forbes lists, and I expected to see my own name somewhere in that answer.
I wasn't in it. Three names came back, one was a big consultancy charging enterprise day rates, one was a freelancer whose last blog post was from 2022, and one I'd never heard of. I asked again a week later, phrased slightly differently, and got a completely different list.
A prospective client told me a few weeks after that she'd asked ChatGPT the same question before our first call and got a name that wasn't mine. She booked anyway because a colleague had recommended me directly, but she mentioned it like it was a funny aside. It wasn't funny to me. If she hadn't had that human recommendation sitting alongside the AI one, I'd have lost the enquiry before I even knew it existed.
That's the bit that should worry every small business owner reading this. You cannot lose a lead you never see. Google at least shows up in your analytics. An AI assistant giving your competitor's name to someone in a private chat window leaves no trace in any dashboard you own. If your business is in Israel, my AI SEO work for Israeli businesses covers exactly this.
What AI visibility is, and why it's not the same as SEO
Search engine optimisation gets you ranked. AI visibility gets you recommended, quoted, or named inside a conversation the user never sees as a "search" at all. Those are related but they are not the same job, and treating them as identical is where most small businesses go wrong.
Google ranks pages. ChatGPT and Perplexity synthesise an answer from several sources at once and then, increasingly, don't even show you which pages they pulled from unless you ask. That means a page can rank on page one of Google for years and still be invisible to an AI assistant, because the model can't easily lift a clean, quotable fact or answer out of it.
Mailchimp is a decent example of a brand that gets this right without necessarily setting out to. Their help documentation answers direct questions in plain sentences, no fluff, no "in today's fast paced digital landscape." When I've tested AI tools on email marketing questions, Mailchimp's own explanations get pulled almost word for word. Compare that to a lot of small business sites where the answer to "how much does X cost" is buried three paragraphs into a story about the founder's journey. The model skips straight past it.
How do you run an AI visibility audit?
You don't need a tool subscription or a developer to do this. You need four browser tabs and about fifteen minutes.
- Open ChatGPT, Perplexity, Google Gemini and Claude in separate tabs. Use free versions, that's what most of your customers are using too.
- Ask each one the same three questions: "who is a good [your service] in [your city/region]", "what does [your business name] do", and "how much does [your service] typically cost in [your area]".
- Write down, word for word, what each tool says. Note whether your business is mentioned at all, whether the details are correct, and whether a competitor is named instead.
- Repeat the exact same three questions a week later. Note what's changed. Consistency, or the lack of it, tells you how settled the model's "opinion" of you is.
- Ask a direct one: "what do people say about [your business name]". This is the one that stings, because it often surfaces old reviews, outdated addresses, or a Companies House listing with your previous business name still attached.
I ran this exact audit with twelve clients over two months at the end of last year. Nine of the twelve weren't mentioned at all when asked for a recommendation in their category. Of the three who were mentioned, one had an address that had been wrong for over a year, because they'd moved premises and never updated an old directory listing that a model had apparently pulled from.
Why ChatGPT named three competitors and left you out
You ask ChatGPT for a business like yours in your town, three rivals come back and you don't. Each usual reason needs a different fix.
- The search crawler can't reach your site. OAI-SearchBot (ChatGPT search) and PerplexityBot both obey the robots file at the root of your domain, and a blanket block from a developer or security plugin shuts them out.
- Google can't use the page. Its AI features only link to pages that are indexed and allowed to show a snippet.
- Rivals say it plainly. Two lines on what they do, where and for whom are easier to quote than your mission statement.
- Other sites vouch for them. Directories, local press, trade bodies and suppliers' pages confirm a rival exists and fits.
- Your details disagree. An old address on one directory and a new one on your site give the model two stories and no reason to trust either.
What an AI visibility audit checks, layer by layer
- Crawler access. Robots rules for OAI-SearchBot, PerplexityBot and Googlebot, read line by line.
- Indexing and snippets. Service, location and about pages confirmed in Google's index, with no stray noindex or nosnippet tag.
- Assistants, kept apart. ChatGPT, Perplexity, Gemini, Claude and Google's AI Overviews each get their own column, never one blended score.
- Buyer questions. Category plus place, problem plus place, your name alone, and you against a named rival.
- Your facts. Name, address, services and hours compared across your site, Google Business Profile and directories.
- Sources. The pages each assistant cited, or seems to lean on, whenever it named a competitor.
- Answer-ready pages. One quotable sentence per core page: who you serve, what you do, where.
Rule out crawler access first, because it takes minutes and blocks everything else. Once access is clean, the layer that usually explains the gap for a small business is sources: rivals sit on the directories, local press and suppliers' pages an assistant can cite, and you don't. That layer is also the slowest to change, so start your source list the same week as the robots check.
Here's the kind of block an audit hunts for in the robots file at the root of your domain:
User-agent: OAI-SearchBot
Disallow: /
User-agent: PerplexityBot
Disallow: /
With those lines in place, OpenAI's documentation says the site won't be shown in ChatGPT search answers, and Perplexity's crawler can't surface it in Perplexity results. A single User-agent: * followed by Disallow: / does the same to every crawler at once. Blocking only GPTBot is a different decision: OpenAI uses that bot for training, and its setting is independent of OAI-SearchBot.
What I found when I widened the test
I asked the same three questions about businesses in six different categories, weighted towards the kind of small, local or specialist businesses I work with: a Bristol accountant, a Manchester wedding photographer, a Tel Aviv marketing freelancer, a Leeds physiotherapist, a Kent copywriter, and a virtual assistant based in Essex.
Across eighteen queries, only five came back with the correct, current business name attached. Three surfaced a competitor by name unprompted. Four gave pricing that was between twelve and forty percent out of date, mostly because the pricing page the model had likely learned from hadn't been touched in over a year. This isn't a huge sample, I'm not pretending it's a national study, but it's a pattern I'd bet money holds up if you tried it on your own business tomorrow.
The uncomfortable part is this: a lot of marketing agencies are now selling "AI SEO" or "AI optimisation" packages, and a fair few of them are guessing. Nobody outside the AI labs has the actual ranking factors for how ChatGPT chooses what to cite, so anyone promising you a guaranteed spot in an AI answer is selling you the same certainty a psychic sells about next week's lottery numbers. What you can do is make your content easier to lift cleanly, which is a different and much more honest promise.
For a benchmark, see my AI Overviews data for 20 local industries.
What moves the needle
Structure beats cleverness here. AI models favour content that answers a direct question in one or two clean sentences near the top, then backs it up with specifics. It's the same instinct behind a good FAQ page, and it's why I now build a short, direct answer into the opening of nearly everything I write, including this post.
A few concrete things that help, based on what I've changed on my own site and watched clients test:
- Put your pricing, location and core service in plain text somewhere a model can lift it whole, not just in an image or a PDF.
- Keep your Google Business Profile, LinkedIn, and any directory listings identical in wording for your business name, address and services. Models cross reference, and conflicting details make you look unreliable even when you're not.
- Write genuine FAQ sections with real questions, phrased the way a customer would type them, not the way you'd phrase them internally.
- Get named on other credible sites. A mention on a trade body page, a guest post, a proper press feature, all of these are the kind of external validation models weigh more heavily than your own homepage saying how great you are.
- Update old pages. A pricing page from three years ago is doing you active harm now, not just sitting there quietly.
None of this is exotic. It's closer to the discipline behind a strong brand than a technical SEO trick. Look at how Figma built its brand strategy around clear, specific, repeatable messaging rather than vague positioning, or how GoPro's marketing strategy stayed obsessively consistent about what the brand stood for across every channel. That consistency is exactly what makes a business easy for an AI model to describe accurately, because there's no conflicting version of the story floating around for it to pick up by accident.
Where content format pulls its weight
Direct answers aren't the only format that gets pulled into AI responses. Anything that produces a specific, personalised output tends to get referenced and shared more, which is one reason I've been such a fan of interactive calculators for small business sites, they generate a concrete number a person can quote back, which is exactly the kind of specific detail AI tools and humans both like to cite. The same logic applies to case studies with real figures instead of vague claims, and to the kind of structured, story led content Joe Pulizzi has spent years arguing for, where the specifics of a real example carry more weight than a general statement of principle. I wrote about some of his thinking in my piece on business lessons from Joe Pulizzi, and the core idea holds here too: specificity travels, vagueness doesn't.
How often should you run an AI visibility audit?
Monthly is enough for most small businesses. I check my own visibility every four weeks, same three questions, same four tools, logged in a simple spreadsheet with the date, the tool, and what came back. It takes fifteen minutes and it's caught two outdated mentions and one wrong phone number in the last six months alone, none of which I'd have known about otherwise.
If your business relies heavily on being found by people who ask AI assistants questions before they ever hit Google, which is becoming more common every quarter, checking weekly for the first month while you fix the obvious gaps is worth the extra time. After that, monthly is fine. This isn't a channel that changes hourly, it changes as fast as the underlying content on the web changes, which is slower than most people assume.
And a word of caution before you spend money on this: being mentioned by an AI tool doesn't automatically turn into a sale the way a Google click does, because there's no link to follow half the time, no button to press, nothing to track. Don't let anyone sell you a package promising guaranteed AI mentions with a guaranteed return, because right now nobody, including the people building these models, can promise that with a straight face. If you want a second pair of eyes on whether this is worth your time and budget specifically, that's the kind of practical, unglamorous question I work through with clients as an AI consultant for small business, and it's usually a shorter conversation than people expect.
Fix the basics first though, because most of it costs nothing but an afternoon and a willingness to look at your own pricing page with fresh eyes. Consistency of information, plain direct answers, and a couple of credible external mentions will do more for you than any paid "AI SEO" package sold on urgency alone.
How to check if ChatGPT recommends your business: a DIY first pass
The fifteen minute check above tells you whether there's a problem. To measure it, widen the test into a log you can rerun. A workable starting set, as a rough guide:
- Ten prompts: four category plus place, three problem plus place, two with your name, one comparing you with a named rival.
- Each prompt run three times, in a fresh chat each time, on each of four assistants. That's 120 answers.
- One spreadsheet row per answer, scored as named or not named, with rivals and any cited pages noted.
- The same prompts rerun monthly, word for word.
An example row, with invented names, looks like this:
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.
| Run | Assistant | Prompt | You named? | Rivals named | Cited sources |
|---|---|---|---|---|---|
| Week 1, run 2 | ChatGPT | wedding photographer in Bath for a small registry office ceremony | No | Studio A, Studio B, Studio C | a wedding directory; a venue's recommended suppliers page |
Expect the first pass to take around half a day with the spreadsheet, and it costs nothing but time. If you hire help instead, compare scope rather than headline fee: how many assistants, prompts and repeats, and whether a retest on the same prompts is included. Ask for the turnaround in writing.
How to judge an AI visibility audit before you pay for one
A good report reads like a fault sheet from a garage: what was tested, what failed, why, and what to change first. Every result carries its date, the assistant and the exact prompt, so you can rerun it. A page of scores with nothing named to fix is a weak report.
Be clear about the limits too: indexing isn't guaranteed, answers move from run to run, and no outside party controls which business an assistant names on a given day.
Questions to ask an AI visibility auditor
| Question to ask | What a good answer sounds like | Why it matters |
|---|---|---|
| Which assistants will you test, and separately? | Each one named and reported in its own column. | A blended score hides where you're missing. |
| Can I see the exact prompts? | Listed in full with dates, ready to reuse. | No prompts, no way to track progress. |
| Do you check crawler access and indexing first? | Robots rules for OAI-SearchBot, PerplexityBot and Googlebot, plus index and snippet status. | A blocked crawler makes every content fix pointless. |
| How do you handle answers that change between runs? | Each prompt repeated, with how often you appear. | One screenshot proves nothing about a moving answer. |
| Will the report show where competitors are cited from? | The pages and directories behind each rival mention. | That's where your missing mentions have to come from. |
| What won't this audit promise? | No guaranteed mentions or rankings, said plainly. | A guarantee shows the seller doesn't understand how these tools pick answers. |
| What happens after the report? | A ranked fix list and a retest on the same prompts. | Without a retest nothing can be shown to have worked. |
What happens after an AI visibility audit
- Crawler and indexing settings get fixed first, since they're switches rather than rewrites.
- Core pages gain a plain opening answer: what you do, for whom and where.
- The source list becomes a target list of sites already naming your competitors.
- The same prompts run again on the same assistants, set beside the baseline.
AI visibility audit facts, with sources
- Google Search Central (read November 2026): no additional requirements or special optimisations are needed for AI Overviews or AI Mode, and Google's guidance on AI features says a page must be indexed and eligible to show with a snippet to appear as a supporting link.
- OpenAI's crawler documentation (read November 2026): OAI-SearchBot surfaces sites in ChatGPT search while GPTBot is used for training, each set independently, and per OpenAI's bot documentation sites opted out of OAI-SearchBot aren't shown in ChatGPT search answers.
- Perplexity's crawler documentation (read November 2026): PerplexityBot surfaces and links sites in Perplexity results; Perplexity-User fetches pages when a user asks a question.
- Pew Research Center (July 2025, March 2025 browsing data from 900 US adults): people clicked a traditional result link in 8% of visits with an AI summary, against 15% without one.
Related reading: Here is what to Expect in a Psychic Reading.
For a measured picture of how often the AI answer appears and which sites it quotes, see the 2026 AI Overviews study.
For every other piece of the puzzle, see my guide to ranking in AI search engines.
If you'd rather not run the checks by hand, my free AI Visibility Checker does the first pass for you, and my AI SEO services cover fixing what it finds.
For the real numbers behind this, read my SEO case study.
Frequently asked questions
What is an AI visibility audit?
An AI visibility audit checks what AI assistants say about your business: whether ChatGPT, Perplexity, Gemini, Claude and Google's AI name you for the questions your buyers ask, what they get wrong, who they name instead and which sources they lean on. It's the AI-search version of checking where you rank on Google.
How do you measure AI visibility?
Pick the questions your buyers ask, run them through each AI assistant and record whether you're named, cited or missing, plus who appears instead. Repeat the same questions monthly so you can see change. Traffic from AI tools in your analytics is a second signal where you can see it.
What is the difference between a GEO audit and an SEO audit?
An SEO audit checks whether Google can crawl, index and rank your pages. A GEO audit checks whether AI engines name and quote you, and why not: missing answers, inconsistent facts about your business, or no mentions on the sources they trust. You need both, and the fixes overlap.
How do I check if ChatGPT knows about my business?
Open ChatGPT and ask it directly what it knows about your business name, then ask for a recommendation in your category and location without naming yourself. Compare the two answers, note anything wrong or missing, and repeat the same test in Perplexity, Gemini and Claude since they often give different answers.
Does AI visibility replace SEO for small businesses?
No, it sits alongside it. Good SEO gets your pages found and ranked, AI visibility is about whether the content on those pages is clear and specific enough for an AI model to lift and quote accurately. You need both, and strong SEO fundamentals usually help AI visibility too.
Can I pay someone to guarantee I show up in ChatGPT answers?
Be very cautious of anyone promising a guaranteed spot in an AI generated answer. Nobody outside the companies building these models controls exactly what gets cited, so the honest work is making your site clear, accurate and consistent, not buying a guarantee that doesn't exist.
How often should I audit my AI visibility?
Monthly is enough once your basics are sorted: pricing, location, services and business name consistent everywhere online. Check weekly for the first month if you're fixing outdated or wrong information, since it's worth catching mistakes quickly early on.
Can an AI visibility audit show how many customers you've lost to AI answers?
No. Conversations with assistants are private, so no audit can count them. It can show whether you appear, how often and who appears instead.
What is the difference between an AI visibility checker and an AI visibility audit?
A checker is an automated first pass over technical signals such as crawler rules and page structure. An audit adds repeated prompt runs on each assistant, the sources behind competitor mentions and a ranked fix list.
Want this done for you? See AI for solopreneurs.
Related reading: Why Every AI-Written Newsletter Sounds the Same (And What I Changed In Mine) and The AI Sameness Problem: Why Your Small Business Marketing Sounds Like Everyone Else's.
Let's get your marketing running on AI.
I'm Lilach. 23 years in marketing. Currently obsessed with AI.
My day job is making your lead generation and ads run on AI. SEO and content that compound. Automation, systems, and processes quiet enough to leave alone. Forty hours a week reclaimed.
Done for you, consulting, coaching, or training.
Run a business, a brand or a blog? I feature tools, products and expert guest posts for an audience of 15,000. See how it works