Of the 21 contractor-specialist firms whose homepages our automated reader could open, only 4 (19%) carried the machine-readable "this is an accounting business" markup that AI assistants lean on to classify and recommend a firm.
And across the full sample of 28 firms, 7 (25%) served a simple automated crawler either a blank shell or nothing readable at all. In a niche whose buyers live in AI chat windows, a quarter of specialist firms are effectively invisible to the kind of fetch an AI assistant performs, and most of the rest read to an AI as a generic website, not an accountant.
We ran our own free Foundations Check, the same tool behind our free Snapshot, across the public websites of 28 UK accountancy firms that explicitly specialise in contractors, IR35 and limited-company (PSC) clients. This is a small convenience sample (n=28), not a census, and it measures foundations readiness, not whether any AI engine actually names a given firm. The numbers below are reported exactly as measured, with their denominators and every limitation stated. We have deliberately kept the report aggregate and anonymised: we name no individual firm with a shortcoming.
Why this matters for contractor accountants
Contractors are an unusually AI-native buyer group: digitally confident, relentless researchers, and exactly the kind of people who now ask ChatGPT or Perplexity "who's a good IR35 accountant for my limited company?" rather than typing it into a Google search box. The questions they ask ("am I inside or outside IR35?", "umbrella vs limited?", "best accountant for an IT contractor outside IR35?") are open-ended, advice-shaped questions, precisely the kind people now put to an AI assistant.
When a contractor asks an AI that question, the assistant doesn't browse the way a human does. It leans on three things: being able to read your site at all (many crawlers fetch the raw page; a JavaScript-only homepage can return an empty shell); structured data (schema.org / JSON-LD that says "this is an AccountingService, here's the name, address and phone"); and consistent name/address/phone plus supporting signals like a sitemap and FAQ markup it can lift ready-made answers from.
"AI visibility" in this report means foundations readiness, not a live ranking. We are measuring whether the plumbing AI assistants rely on is in place, not claiming a position in any AI engine's answer.
The plumbing is not magic. But in a niche this competitive, where a dozen national firms fight over the same "contractor accountant" intent, the firms whose plumbing is in place give an AI a clearer reason to put their name forward.
Method, and its limitations (read this first: it's the honest part)
We lead with the limitations rather than burying them, because the credibility of this report rests on transparency.
// How we ran it, and what it does not do
- What we ran: our own Foundations Check tool (tools/foundations_check.py), which makes a small number of polite, public-page requests: homepage, robots.txt, llms.txt, sitemap.xml, and /contact / /about. It inspects only publicly available information, uses a normal browser user-agent and short timeouts, reads only a handful of pages per site, and leaves a short delay between firms.
- What it checks: whether the homepage can be read; whether schema.org / JSON-LD is present and of what type; whether there is FAQ markup; whether there is an llms.txt; whether there is a sitemap; whether robots.txt blocks AI crawlers; whether a WAF / bot-challenge interstitial is present; and whether NAP signals are present and internally consistent.
- What it does NOT do, stated plainly: it does not test whether any AI assistant actually names the firm. This is a measure of foundations / readiness, not a live AI ranking; no finding here should be read as "AI ignores firm X." It does not execute JavaScript, so sites that render entirely client-side read as "unreadable" to our tool and to similarly simple AI crawlers; we count those separately and never as "no schema." It is point-in-time: all checks were run on 3 June 2026, and websites change.
- The sample: 28 firms identified via public web searches for contractor/IR35-specialist accountants ("contractor accountants UK", "IR35 specialist accountant", "accountant for IT contractor outside IR35", and similar). We screened for firms that genuinely position on contractor/IR35 work, rejecting directory sites, listicles and generic high-street firms. This is a convenience sample, not a random or representative one; it cannot be generalised to all UK contractor accountants. With n=28, each firm is worth roughly 3.6 percentage points, so small differences are noise.
- A caveat that cuts against us: this niche skews toward larger national firms with dedicated web teams. That biases the llms.txt and sitemap numbers upward versus a small-independent sample, so the identity-schema and FAQ gaps below are arguably understated for the smaller end of the niche.
- A US-weighting caveat on the proxy note only: the one soft cross-check we ran of live search results (see the proxy note at the end) is US-weighted and is not a verbatim engine read. It is offered as colour, not as a measured finding.
- The honesty rule: no number here is rounded up or invented. Where the tool could not read a page, we count that separately and do not record it as "no schema"; claiming a firm lacks schema when we never saw its page would be false.
The sample, in three honest buckets
| Bucket | Count | What it means |
|---|---|---|
| Homepage readable | 21 / 28 | We retrieved real page content. Schema and NAP findings are based on these firms only. |
| Homepage unreadable (blank shell) | 4 / 28 | Server returned a "success"-type response (HTTP 202) but zero readable text and no title: a blank / JavaScript-only shell. Not scored for schema/NAP. |
| Could not fetch | 3 / 28 | Two firms refused the connection on two attempts; one returned a redirect loop (HTTP 307). Not scored, but reported as a fetch outcome, not a specific fault. |
The split itself is a finding: combined, 7 of 28 firms (25%) served either nothing readable or nothing at all to a simple automated fetch, exactly the kind of fetch many AI crawlers perform.
Denominator rule for everything below. Schema and NAP percentages are calculated only over the 21 firms whose homepages we could actually read (denominator = 21), so they describe firms we genuinely assessed. The llms.txt, sitemap, robots and fetch-outcome percentages are over the full sample of 28.
// Convenience sample of 28 firms, checked 3 June 2026. Aggregate and anonymised. Foundations / readiness only, not live AI rankings.
The findings
Every figure below carries its denominator and a plain-English "so what." Schema and NAP percentages are over the 21 readable firms; llms.txt, sitemap, robots and fetch-outcome figures are over the full 28.
// Finding 01 · entity schema
Only 4 of 21 readable firms (19%) tell AI "I am an accounting business"
Of the 21 firms with a readable homepage, just 4 (19%) carried a LocalBusiness, AccountingService, ProfessionalService or FinancialService schema type, the structured signal that lets an AI confidently classify a firm as a contractor-accounting provider. The other 17 readable firms relied on generic schema (Organization, WebSite, WebPage, BreadcrumbList) that is useful for ordinary search but does not say "accountant."
So what: in a niche where every firm claims contractor expertise in prose, the four firms that also say it in machine-readable form stand out to an AI. The other eight in ten read to an assistant as a generic website.
// Finding 02 · blank to a crawler
A quarter of firms served a homepage that read as blank, or nothing at all
4 of 28 firms (14%) returned an HTTP 202 response with no readable text and no title: a JavaScript-only or placeholder shell. A further 3 of 28 (11%) could not be fetched at all (connection refused twice, or a redirect loop). Combined: 7 of 28 (25%) gave a simple crawler nothing usable.
So what: this is potentially the highest-impact issue of all. Human visitors with a full browser see normal sites; a simple AI crawler reading the raw response may see nothing. You cannot be recommended on the strength of content a crawler never receives. We flag this conservatively as a readability / reachability risk, not in every case a confirmed hard block.
// Finding 03 · mis-targeted schema
Every readable firm had some JSON-LD, but mostly the wrong kind
0 of 21 readable firms (0%) had no structured data at all, a better baseline than our general-accountancy benchmark, where a quarter had none. The problem here is not the absence of schema, it's mis-targeted schema: 17 of 21 readable firms carry only generic markup.
So what: these firms have already invested in structured data. They simply have not pointed it at the specific types (AccountingService / FAQPage) that do the work in this niche. That makes the fix cheap: redirect existing effort, don't start from scratch.
// Finding 04 · FAQ markup
FAQ markup is more common here than in general accountancy, but still a minority
5 of 21 readable firms (24%) carried FAQPage schema, visibly higher than the 0% we found in our general UK-accountancy benchmark. That makes sense: IR35 content is naturally Q&A-shaped. But it still means roughly three in four readable specialist firms, 16 of 21 (76%), are not handing AI assistants ready-made, liftable question-and-answer content, despite their entire topic being built on repeatable contractor questions.
So what: this is the cheapest available edge in the niche. The whole specialism is built on FAQ-shaped questions; only a quarter of firms are formatting them the way an AI prefers to quote.
// Finding 05 · NAP
NAP basics are usually present, but only ~1 in 3 firms put them where AI reads
| NAP signal | Count | % of readable (n=21) |
|---|---|---|
| Strong NAP on page (phone and UK postcode detectable) | 16 | 76% |
| Partial NAP (one but not both) | 4 | 19% |
| No detectable NAP signal | 1 | 5% |
| NAP inside JSON-LD (machine-readable) | 6 | 29% |
| No structured NAP for an AI to read reliably | 15 | 71% |
So what: the address is on the page for humans, but for 15 of 21 firms (71%) it is not in the machine-readable layer AI prefers. Only 6 (29%) carry NAP inside their structured data, where an AI can read it reliably.
// Finding 06 · llms.txt, sitemaps, robots
llms.txt adoption is above average; sitemaps and robots are patchy
llms.txt present: 10 / 28 (36%), markedly higher than the 12% in our general-accountancy sample, because this niche skews toward larger, more digitally sophisticated national firms. Still, nearly two-thirds (18/28) have not adopted the emerging AI-crawler-guidance standard. Sitemap present: 20 / 28 (71%), so roughly a third have no discoverable sitemap. robots.txt found: 22 / 28 (79%).
So what: a short llms.txt and a discoverable sitemap are low-cost, low-adoption edges, present for fewer than two in five and seven in ten firms respectively.
// Finding 07 · crawler blocking
Outright AI-crawler blocking is rare, but a few firms self-sabotage
1 / 28 (4%) explicitly disallows one or more major AI crawlers (GPTBot, ClaudeBot, Google-Extended, CCBot, Bytespider, Applebot-Extended) in robots.txt, telling AI tools to stay away. Confirmed WAF / bot-challenge hard blocks: 0 / 28 (0%); 1 / 28 (4%) had a managed challenge present while still serving content (not a hard block). Separately, the 3 "could not fetch" firms are, in practice, invisible to the kind of plain fetch an AI crawler uses, whatever the intent behind it.
So what: almost no firm here is deliberately blocking AI, so the readability problems above are mostly accidental, not chosen, which means they're fixable.
What good looks like (free value, whether or not you ever buy from us)
These are the foundations a contractor-specialist firm should have. None is expensive, and most are a one-off change you can hand to your web person. They are useful to any firm reading this, customer or not.
1. Add accountant-identifying schema (fixes Finding 1)
Add AccountingService (or ProfessionalService / LocalBusiness) JSON-LD to your homepage, with your real name, address, telephone, url and areaServed. This is the single most-missed signal: only 19% of readable firms had it. Putting your address inside that block also fixes the structured-NAP gap (Finding 5): 71% of readable firms had no machine-readable NAP at all. A minimal example you can copy and adapt:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "AccountingService",
"name": "Your Firm Ltd",
"url": "https://yourfirm.co.uk/",
"telephone": "+44 20 1234 5678",
"areaServed": "United Kingdom",
"address": {
"@type": "PostalAddress",
"streetAddress": "1 Example Street",
"addressLocality": "Leeds",
"postalCode": "LS1 1AA",
"addressCountry": "GB"
}
}
</script>
2. Mark up your IR35 FAQs (fixes Finding 4)
You already answer "Am I inside or outside IR35?", "umbrella vs limited?", "what counts as outside IR35?" on your site. Wrap those Q&As in FAQPage schema so an AI can lift them directly. Only 24% of firms here do this, despite the whole specialism being FAQ-shaped; it is the cheapest edge available. Keep the answer text identical to what's visible on the page:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [{
"@type": "Question",
"name": "Am I inside or outside IR35 as a limited-company contractor?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Your plain-English answer here, matching the on-page content exactly."
}
}]
}
</script>
3. Make sure a non-JavaScript crawler can read your homepage (fixes Finding 2)
This is the big one: a quarter of firms here failed it. If your site renders entirely client-side, a simple AI crawler may receive a blank shell. Ask your web person whether your homepage is server-rendered or pre-rendered; if it isn't, the fix (SSR / static pre-render of the key content) is the single highest-leverage change you can make. Identity schema and FAQ markup are worth nothing to an AI that never receives any readable content in the first place.
4. Add a basic llms.txt and a sitemap (fixes Finding 6)
A short llms.txt at your domain root that points AI crawlers to your key pages is a low-cost, low-adoption edge (only 36% here have one). Make sure sitemap.xml exists and is discoverable; roughly a third of firms here had none.
5. Don't accidentally block the crawlers you want (fixes Finding 7)
Check your robots.txt is not disallowing GPTBot, ClaudeBot, Google-Extended or similar unless you genuinely intend to. One firm here was telling AI tools to stay away.
The order that matters most: be readable first (3), then identify yourself (1), then hand over your answers (2). A firm that gets all three right, in this specialism, today, is in a genuinely small group.
What this means for contractor accountants in 2026
Four honest implications, no hype, no guarantees.
- Your buyers are AI-native; your foundations aren't. Contractors research in ChatGPT and Perplexity before they ever fill in a contact form, yet a quarter of specialist firms in this sample served a simple crawler nothing readable, and only ~1 in 5 readable firms identifies itself to AI as an accounting business in machine-readable form. The audience moved to the AI front door faster than the sites did.
- The gap isn't "no schema"; it's the wrong schema. Every readable firm had structured data; almost none had the accountant-identifying types, or (in three-quarters of cases) FAQ markup. You have already paid for structured data. Pointing it at the right types is a small change with outsized return in a crowded niche.
- IR35 content is FAQ-shaped, so use it. Only 24% of readable firms use FAQ markup, despite the entire specialism being built on repeatable contractor questions. That is the cheapest available edge: hand the AI the exact Q&A it wants to quote, in the format it prefers.
- First-mover advantage here is real and narrow. This is a tight national niche: a dozen firms compete for the same "contractor accountant" and "IR35 accountant" intent. The handful that get readable + identity schema + FAQ + structured NAP right give AI assistants the clearest reason to name them. Today that handful is genuinely small.
None of this is a promise that doing these things will get you recommended by any AI engine. It is a description of the foundations the best-prepared firms in this sample have in place, and most do not.
About Renownly, and how this data was produced
Renownly runs AI-visibility audits for independent UK accountancy firms. The tagline, "Be the name AI knows", is the outcome we work toward: helping a firm become legible to the AI assistants their clients increasingly ask for recommendations.
This report is first-party research. We did not buy a dataset or scrape an AI consumer interface. We built our own foundations_check.py tool and ran it, the same way we run a paid audit, across 28 public websites on a single day. The value isn't a secret algorithm: it's that we did the work transparently, kept every denominator, separated "couldn't read" from "had nothing," and refused to round a single number up. The raw per-firm data is retained internally; this published report is aggregate and anonymised by design, for fairness and data-minimisation, so no individual firm is named with a shortcoming.
The no-guarantee disclosure (stated in full)
// Our standing no-guarantee statement
No one can guarantee that an AI assistant will recommend your firm. The tools are run by third parties and change without warning, and this report measures foundations rather than outcomes. We report on the current state of third-party AI tools and provide best-practice recommendations; we do not control those tools and do not guarantee rankings, recommendations, traffic or revenue. AI results vary and are dated at the time of testing. Get the machine-readable foundations right, measure properly, and improve your odds.
Proxy note (clearly labelled, not a core finding)
The solid core of this report is the foundations data above. We did not perform verbatim AI-engine reads. As a soft cross-check we ran two prospect-style queries ("best accountant for IR35 contractors", "who is the best contractor accountant UK 2026"); the results were dominated by third-party directory and listicle pages ("top contractor accountants 2026", ContractorUK directory, comparison sites) rather than firms' own pages. That is a useful signal that in this niche the intermediaries currently own the high-intent answer space, and individual firms are mostly surfaced through directories rather than directly. This is a search-results proxy (US-weighted, not a verbatim engine read), offered only as colour, not as a measured finding.