Let's say your website has been live for eighteen months. Perhaps longer. It ranks reasonably well on Google for the queries you care about, the phone still rings, and from the outside everything appears to be in order. And yet a quiet suspicion has been growing — that something has shifted, that customers are asking questions of ChatGPT and Perplexity before they ever reach a search engine, and that your business is simply not part of the conversation.
The good news is that you do not need an agency, a dashboard, or a paid audit tool to find out. You need sixty seconds, a browser, and a willingness to look honestly at what you see.
What follows is a five-step manual diagnostic. Each step takes roughly twelve seconds. By the end you will know whether your business is being mentioned, ignored, or — most commonly — invisible for reasons that are entirely fixable. Of course, the interpretation of each step matters as much as the step itself, so I will walk through what each result actually tells you.
Step 1: Ask ChatGPT the question your customers ask
Open ChatGPT. Use the free version, no special prompt engineering required. Type the exact query a customer would type. Something like: "Best [your service] in [your city]." If you run a dental practice in Calgary, that is "best dentist in Calgary." If you sell artisanal pasta in Brooklyn, that is "best fresh pasta shop in Brooklyn."
Read the response carefully. Three things can happen, and each tells a different story.
If ChatGPT names your business by name, you are doing something right — the model has encountered enough signals about you to surface your name confidently. Take note of how it describes you, because that description is what every future customer will read.
If ChatGPT names three or four competitors but not you, the model knows the category exists in your city, has decided who the relevant players are, and has not included you in that set. This is the most common — and most painful — outcome. It is not that AI cannot find you. It is that it has found you and concluded you are not worth mentioning.
If ChatGPT gives a generic non-answer ("I recommend searching online review platforms"), the model has insufficient confidence to name anyone. This is rarer in established categories but common in narrow niches.
Step 2: Repeat the test on Perplexity
Now open Perplexity and ask precisely the same question. Why both? Because ChatGPT and Perplexity draw on different underlying indexes and apply different retrieval logic. ChatGPT leans on Bing's index plus its training data; Perplexity performs live web searches and cites its sources inline.
Perplexity is, in many ways, the more diagnostic of the two — because it shows you the citations. If your competitors appear with a little numbered footnote pointing to a Yelp page, a local "best of" list, or their own website, you can click through and see precisely where the AI is sourcing its recommendations. This is, without doubt, the single most useful exercise in this entire diagnostic.
Pay attention to two things. First, are you cited at all, anywhere in the response? Second, what kind of pages are being cited for your competitors — review aggregators, directory listings, or their own structured content? Because whatever those pages are doing, that is the format of evidence the AI considers credible.
Step 3: Check the Google AI Overview
Open Google in an incognito or private window — this matters, because your personal browsing history will distort the result. Type the same query. Look at the very top of the results page. If an AI Overview appears, it will be a paragraph or two of synthesized text, often with two or three business names tucked inside, each one a citation.
Three outcomes again. The AI Overview names you — excellent, you are part of Google's generative answer for that query. It names competitors but not you — Google has the data but is choosing them over you. No AI Overview appears at all — Google has decided this particular query does not warrant a generative answer, which is itself a useful piece of information about how Google understands the intent.
It must be said that AI Overviews are still evolving rapidly, and what triggers them changes from month to month. But the snapshot you take today is the snapshot a real customer would see today, and that is what matters.
Step 4: View the source of your own homepage
Now we move from external surfaces to your own house. Open your homepage in a browser. Right-click anywhere on the page and select "View Page Source" (or press Ctrl+U on Windows, Cmd+Option+U on Mac). A new tab opens with the raw HTML.
Use Ctrl+F (or Cmd+F) and search for the word LocalBusiness. Then search for schema.org. Then search for application/ld+json.
If none of these appear, your page has no structured data — and this is, more often than not, why AI systems cannot describe your business with any specificity. Structured data is the machine-readable version of "we are a dental practice at this address with these hours and these services." Without it, AI models are forced to infer everything from prose, and inference is slower, weaker, and easier to skip in favour of a competitor who has spelled it out properly.
If you do see a JSON-LD block, read it. Is the business type correct? Is the address there? Phone number? Opening hours? Services? Reviews? A half-filled schema is not the same as a complete one. For the practical mechanics of fixing this, see our walkthrough on schema markup for small businesses in 15 minutes.
Step 5: Check your robots.txt for AI crawlers
This is the step almost nobody runs, and the one that catches the most people out. In your browser's address bar, type your domain followed by /robots.txt — for example, yourbusiness.com/robots.txt. A plain text file will load.
Look for lines that mention GPTBot, PerplexityBot, ClaudeBot, Google-Extended, or CCBot. There are two patterns to recognise.
If you see User-agent: GPTBot followed by Disallow: /, you are explicitly blocking ChatGPT from crawling your site. This was a fashionable default in 2023, when many CMS templates added it automatically. If your site is older than eighteen months, there is a non-trivial chance this is sitting in your robots.txt without your knowledge — which means ChatGPT has been forbidden from learning anything about you, for years.
If GPTBot does not appear at all, the crawler is allowed by default. That is the desired state. The same logic applies to the other AI user-agents: silence means permission, an explicit Disallow: / under their name means the door is closed.
Of course, allowing the crawler is necessary but not sufficient — the crawler still has to find a reason to remember you. But blocking it guarantees invisibility, and that is the easiest mistake to fix.
What the five steps tell you, taken together
Each step isolates a different layer of the problem. ChatGPT and Perplexity show you whether the AI knows you exist. Google AI Overview shows you whether you are part of the synthesized answer Google is now serving. Your page source shows you whether your own site is giving the machines anything to read. Your robots.txt shows you whether you have been blocking them without realising it.
The diagnostic does not produce a single tidy verdict. It produces a map. If Steps 1 and 2 fail but Steps 4 and 5 are healthy, the problem is external — your structured data is fine, the crawlers are allowed in, but the evidence elsewhere on the web (reviews, directory citations, third-party mentions) is too thin to push you into the model's working memory. If Steps 4 or 5 fail, the problem is internal, and you can fix it this afternoon.
For a fuller treatment of why so many established sites fall into this trap, the companion piece on why your website is invisible to ChatGPT and Perplexity goes deeper into the underlying mechanics.
Frequently asked questions
Why does ChatGPT name some businesses and not others in the same city? AI models are not running a fresh search every time. They surface what their retrieval and training data already consider authoritative. Businesses that appear consistently across review platforms, directories, local "best of" lists, and that publish machine-readable structured data on their own site, accumulate enough signal weight to be remembered. Businesses with only a website and no schema accumulate almost none.
Can I just ask ChatGPT to add my business? No. ChatGPT has no admin interface, no submission form, no "claim your listing." The only path into the model's working memory is improving the public signals it draws from — reviews, citations, structured data, mentions in indexed third-party content. The good news is that these signals are entirely within your control.
How often should I re-run this 60-second test? Once a month is more than enough. AI models update their indexes on rolling cycles, and visible changes typically take several weeks after the underlying signals improve. Running the test weekly will only frustrate you.
What if my robots.txt blocks GPTBot — will allowing it suddenly make ChatGPT recommend me? On its own, no. Allowing the crawler removes a barrier, but the crawler still needs a reason to cite you. Think of it as unlocking the door — the visitor still has to want to come in. Pair the unblock with proper schema and a few authoritative third-party mentions, and you give the crawler something worth indexing.
I have done the test and I am invisible. Where do I start? Start with whichever step failed most clearly. If robots.txt is blocking GPTBot, unblock it today. If your page source has no schema, install LocalBusiness JSON-LD this week. If the external surfaces are the weakness, the work is slower — patient accumulation of reviews, citations, and authoritative mentions. The plain-English overview in what is GEO for business owners is a sensible next read.
Or skip the manual version — licheo.com/seo-standings runs the full check in one click.