When a prospective client asks ChatGPT "best accountant for my limited company near me," the assistant doesn't browse your site like a human does. It fetches the raw page, reads what it can, and decides whether it has enough to name you. If something quietly stops it reading you, you're not in the running, no matter how good your firm is. The good news: you can check the main failure points yourself, for free, in about ten minutes. Here's how, in plain English.
These five checks mirror what Renownly's own Foundations Check tool inspects automatically. Doing them by hand won't be as thorough, but it'll tell you whether you have an obvious problem, and most firms find at least one.
The 10-minute self-check
// Check 1
The "blank to a crawler" check
This is the highest-impact one, so it's first. Some modern websites look perfect in your browser but serve a near-empty page to a simple crawler, because the real content only appears after JavaScript runs. Many AI crawlers read the raw response and may see almost nothing. You can't be recommended on content a crawler never received.
In our June 2026 study, this wasn't rare:
6 of 34 firms (18%) served a homepage that read as completely blank to our automated reader: a success-type response with no readable text and no title. Human visitors saw a normal site; a simple crawler saw an empty shell.
// Source: Renownly, UK Accountancy AI-Visibility Foundations Report, 3 June 2026. Convenience sample, n=34. We flag this conservatively as a readability risk, not a confirmed hard block.
// Check 2
Can the AI crawlers even reach you? (robots.txt)
Your site has a small public file that tells automated visitors where they may go. A previous developer or a security plugin can quietly add a line that tells the AI crawlers to stay away, often without anyone realising.
User-agent: GPTBot
Disallow: /That Disallow: / means "block everything." If you see it against any AI crawler, or against * (all agents), that's a flag to fix. A Disallow: on only admin areas like /wp-admin/ is normal and harmless. If you have no robots.txt at all (a "404"), crawlers are allowed by default, which is fine.
// Check 3
Do you tell AI "I am an accountant"? (schema)
Structured data (schema) is invisible code that states your business type, location and contact details in a format machines read reliably. Without the right type, an assistant has to guess you're an accountancy firm from your prose.
This is the most common gap we see. In our study, only 4 of 24 readable firms (17%) carried an AccountingService-type schema. If you're missing it, our copy-paste AccountingService guide gives you a ready-to-fill block.
// Check 4
Is your name, address and phone machine-readable? (NAP)
Your contact details are almost certainly on the page for humans. The question is whether they're also in your structured data, where an AI can read them reliably and cross-check that two listings are the same firm.
In our study, almost everyone had NAP somewhere on the page, but only 2 of 24 readable firms (8%) carried it inside structured data an AI could trust. The details are there for people; they're often missing from the machine-readable layer.
// Check 5
Do you have an llms.txt? (the easy edge)
An llms.txt is a short, plain-text summary you publish to help AI assistants understand what you do, an emerging standard, not yet essential, but an easy edge almost nobody has taken.
In our study, only 4 of 34 firms (12%) published one. It's a low-cost way to hand assistants a clear, on-your-terms description of your firm. (Practising what we preach, this very site publishes an llms.txt you can look at as an example.)
What your results mean
If you passed all five, your foundations are in good shape, better than most firms in our sample. If you flagged one or more (most firms do), you've found exactly where an AI assistant may be losing you, and every one of these is fixable.
Here's the honest framing. These checks tell you whether AI assistants can read and identify your firm. They don't, on their own, tell you whether an assistant actually names you when a client asks, that depends on more signals (directory presence, reviews, earned mentions) and on the assistants themselves, which vary their answers run to run and change without notice. Getting the foundations right stacks the odds in your favour. It can't guarantee an outcome, and you should be wary of anyone who says it can.
// If you found a problem
The quick triage
- 1Blank to a crawler? Talk to your web developer about server-side rendering or pre-rendering. This is the one most worth fixing.
- 2AI crawler blocked? Ask whoever maintains your site to remove the Disallow: / line for that crawler. Usually a one-line change.
- 3No AccountingService schema? Use our copy-paste schema guide, then validate it.
- 4NAP not structured? The same schema block fixes this, get your name, address and phone into the JSON-LD.
- 5No llms.txt? Lowest priority, but an easy add. A short plain-text summary of your firm at the root of your site.
Want a second opinion that's faster and more thorough than doing it by hand? That's exactly what our free Snapshot is for, it runs these checks automatically, plus actually tests whether AI assistants name you for the questions your clients ask.