Before you touch anything on your website, you need to know where you actually stand. And for that, you don’t need to hire anyone: you can run this audit yourself in about an hour, with free tools and a document to jot things down in.
Here’s the method we use.
Step 1: ask the way your customer would ask
Open ChatGPT and Perplexity (both, because they don’t answer the same way) and put the questions someone who needs you — but doesn’t know you — would actually type.
The typical mistake is asking about your own brand: “What do you know about [your brand]?”. That only tells you whether the model has heard of you, not whether it recommends you. Ask about the problem instead:
- “Which [your service] would you recommend in [your city]?”
- “I have [the problem you solve] — who should I contact?”
- “What are the best [your sector] companies for small businesses in [your area]?”
- “I’m comparing a few options for [your service] — what should I look for?”
For each question, note down: do you show up?, who does?, and how is each one described?. That last column is the most revealing one — it tells you what you’re actually associated with.
Repeat this in incognito mode or logged out. If the model has memory of your past conversations, the results will be skewed in your favour.
Step 2: check whether it knows you by name
Now ask directly about your business: “What is [your brand]?”, “What does [your brand] do?”.
There are three possible outcomes, and each one means something different:
- It describes you accurately → you’re in good shape, your information is consistent and easy to find
- It doesn’t know you → there’s little material about you out there, or it isn’t crawlable. This is the most common case, and the easiest to fix
- It describes you wrong → 🚨 the most urgent case. It confuses you with someone else, credits you with services you don’t offer, or works from outdated information. That’s reaching potential customers right now
Step 3: check whether your site is machine-readable
Three concrete checks:
a) Is your content actual text? Open your site and try selecting the text of your services and pricing with your mouse. If you can’t select it because it’s an image, a model can’t read it either. This is surprisingly common on sites built with visual page builders.
b) Does it show up without JavaScript? Right-click the page and choose “View Page Source”, then search for a paragraph of your text. If it’s not there, your content is being generated afterwards by JavaScript, and many crawlers will never see it.
c) Do you have structured data? Run your site through Google’s Rich Results Test or validator.schema.org. It’ll tell you whether you have markup for your organisation, services, or FAQs. If nothing comes back, this is one of the highest-return fixes you can make.
Step 4: check whether you’re letting AI crawlers in
Check your yourdomain.com/robots.txt file. Some security plugins and some hosts block crawlers like GPTBot, PerplexityBot, ClaudeBot or Google-Extended by default.
If they’re blocked, there’s nothing left to optimise — you’ve shut the door before you even started.
Worth flagging: there’s a legitimate call to make here. Blocking them means your content won’t be used to train models, but it also means you won’t get cited. That’s a business decision, not a technical setting — just make it deliberately, not by default.
Step 5: get your external sources to agree
Models don’t only read your website. Check that these all match, exactly:
- Google Business Profile: name, address, phone number, category, opening hours
- Social profiles: LinkedIn, and whatever else you use
- Industry directories you’re listed in
- Reviews: how many, and what they say
Any contradiction between these sources and your website produces an answer that’s hedged, or flat-out wrong.
Step 6: look at who does get recommended
Go back to the answers from Step 1 and study whoever shows up. What do they have that you don’t?
It’s almost always one of these: content that answers specific questions, a presence in industry directories or press, a solid volume of reviews, or a website that clearly explains who it’s for and what problem it solves.
Don’t copy them — just spot the pattern.
How to read the result
If you run this audit and barely show up anywhere, that’s not a disaster — it’s the norm right now for almost any small business. The upside is that your competitors are in the same boat, and the room for improvement is huge.
Save your notes with the date. Repeat the same questions in three or six months and you’ll have a real measurement of whether the work paid off — which, for now, is the only serious way to track this, since no official tool does it for you.
And if you’d rather we did it for you
You can absolutely run this audit yourself — that’s why we’ve published the whole thing. We run it more systematically: more queries, a comparison against your competitors, a full technical review of structured data and accessibility, and a prioritised plan for what to fix first.
If you’d like us to run it on your site, tell us about your business. And if you’d rather do it yourself with this guide, go for it — what matters is knowing where you stand before your competitors work it out.
In the next article: exactly what to change on your site so AI actually understands it.
Want to apply this in your business? See how we do it in GEO (ranking in AI search engines) and customer acquisition with AI, or tell us about your case and we will reply within 24-48 working hours.