Updated
Updated What changed on
Fact-checked Only Google's guidance on structured data and AI search, Google's FAQ rich result change, what Google-Extended is, and which AI crawlers control training and which control search were re-checked against the 14 October 2026 source checks. The law firm study's date was checked against that page. Nothing else on the page was re-checked.
- Reworded the 32% figure. It counts firms with no schema, or no schema saying they are a business. It no longer calls them invisible to AI search, because Google says structured data is not required for its AI features, though it still helps pages qualify for rich results.
- Clarified the silent block finding. GPTBot and ClaudeBot are training crawlers, so a 403 to them blocks training. It does not prove a firm is missing from ChatGPT or Claude search, which use OAI-SearchBot and Claude-SearchBot.
- Corrected the Google-Extended test. Google-Extended is a robots.txt token with no user agent of its own, so a request sent under that name does not reflect any real Google crawler.
- Changed the self-check to test OAI-SearchBot, Claude-SearchBot and PerplexityBot, the crawlers ChatGPT, Claude and Perplexity search use.
- Removed the claims that schema is how AI engines confirm what a site is and which sources they prefer. Google says no special schema.org markup is needed for its AI features.
- Noted that Google stopped showing FAQ rich results on 7 May 2026, and dropped FAQPage markup from the recommended fixes.
- Replaced "two weeks ago" with the law firm study's actual date, 22 May 2026, and linked to it.
UK accountants in AI search: the data, the gaps, the silent crawler blocks.
The 30-second answer
I audited 56 UK accountancy firm websites in May 2026. More than a quarter silently block AI training crawlers at the hosting layer, including two of the largest firms. That is worth checking against the search crawlers ChatGPT and Claude use (OAI-SearchBot, Claude-SearchBot). Almost none carry the structured data that states they are an accountancy practice. This matters because 69.5% of UK Google searches now end without a click, the highest rate of the six countries measured, so being quoted inside the answer is increasingly the only visibility on offer.
I checked 56 UK accountancy firm websites - the Big 4, the top 25 mid-tier, and a sample of regional practices and small independents - against two signals: the structured data they carry, and whether AI crawlers can reach them.
The findings are worse than the same study I ran on UK law firms on 22 May 2026. The headline numbers:
The published research this audit sits against
of UK Google searches end without a click on anything, the highest rate of the six countries measured and above the United States at 68.0%.
How it was measured: Similarweb mobile and desktop clickstream panel, January to April 2026. Zero-click means no click to anywhere, including to Google's own properties and to paid results. Germany was lowest at 62.1%.
more organic clicks went to a brand cited inside an AI Overview than to one that was not, in Seer's 2025 study (data to September 2025). Treat it as dated: Seer's April 2026 update says cited brands saw lower click-through as impressions more than doubled while clicks stayed flat.
How it was measured: 3,119 search terms across 42 client organisations, 25.1M organic and 1.1M paid impressions, June 2024 to September 2025, so the data is now more than 12 months old. Informational and educational queries only, which are the ones most exposed to AI Overviews, so this is not a whole-market figure. In absolute terms it is 0.70% against 0.52%. Paid click-through was 91% higher when cited.
of URLs cited in Google AI Overviews appeared in the first 10 results blocks for that query. Ahrefs counts blocks, not organic results, so this is not the same as ranking in the top 10.
How it was measured: 863,000 keyword SERPs and 4 million AI Overview URLs. Ahrefs measures the top 10 as blocks, counting ads, featured snippets and video packs as separate blocks.
How this compares to UK law firms
I ran the same methodology on 73 UK law firm websites in May 2026. The accountancy results are sharper on almost every measure. Accountants are further behind on schema, far further behind on industry-specific markup, and three times more likely to be silently blocking AI bots at the hosting layer.
| Measure | UK law firms (73 firms) | UK accountants (56 firms) | Worse? |
|---|---|---|---|
| Reachable for audit | 65 firms | 41 firms | comparable |
| No structured data at all | 22% | 22% | same |
| Industry-specific schema (LegalService / AccountingService) | 25% | 7% | 3.5x worse |
| Has FAQ schema | 3% | 2% | same (both terrible) |
| Silent AI bot block (hosting-layer 403) | 9% (6 of 65) | 27% (11 of 41) | 3x worse |
| Intentional robots.txt AI bot block | 1 firm | 0 firms | accountants chose nothing |
| No schema, or no schema saying they are a business | 29% | 32% | slightly worse |
The pattern is the same in both sectors - professional services firms that sell expertise and credibility send weak or blocked signals to the AI surfaces their clients increasingly use to evaluate them. Accountants are further along the gap because AccountingService schema is barely used, and because more accountancy firms sit on hosting stacks that block AI bots without configuration.
The silent AI bot block - including the Big 4
The most striking finding is that 11 of 41 reachable firms - 27% - silently block at least one major AI crawler at the server layer. None of them chose to. Every single block is invisible to the firm because:
- The robots.txt file allows all AI bots (zero intentional disallow rules across the entire sample)
- The website loads normally in a browser - Chrome gets a 200 response
- The block only shows up when you make the request with a GPTBot or ClaudeBot user agent header
- The hosting layer (Cloudflare, Akamai, similar WAF or bot-management product) intercepts the AI bot request and returns 403 before the firm's website ever sees it
Affected firms in this sample include PwC, BDO, Crowe UK, FRP Advisory and TaxAssist, all blocking GPTBot and ClaudeBot simultaneously. They also returned 403 to a request sent as "Google-Extended", but Google-Extended is a robots.txt token with no user agent of its own, so that request does not reflect any real Google crawler. Six additional firms block ClaudeBot specifically while leaving GPTBot through, which usually points to a Cloudflare bot management ruleset that classifies ClaudeBot more aggressively than the OpenAI crawler.
None of these firms have published a position on AI training data. None of them have a robots.txt rule expressing a stance. The block is happening at infrastructure they probably do not know is there, blocking AI training crawlers. Whether it also blocks the search crawlers ChatGPT and Claude use (OAI-SearchBot, Claude-SearchBot) needs its own check.
The honest version
A firm can legitimately decide it does not want AI engines training on its content. That is a defensible commercial position. The problem is when the decision is made by default by a hosting provider's WAF ruleset, with the firm having no idea it has been made, and no measurement of what business it is costing them. Across this sample of 41 reachable accountancy firms, every single silent block looks accidental.
The schema gap - 93% have no AccountingService markup
Structured data is a clear way for a website to state that it belongs to an accounting practice rather than, say, an accountancy training provider, a payroll software vendor, or a generic professional services firm. The relevant schema.org type is AccountingService, available since 2014. Google says structured data is not required for its AI features, but it still helps pages qualify for rich results.
Among the 41 reachable firms in this sample, only 3 (7%) carry AccountingService schema on the homepage. The breakdown:
- 27 firms (66%) have generic
OrganizationorCorporationschema - the website is identified as a business, but not as an accountancy practice specifically - 9 firms (22%) have no JSON-LD schema at all on the homepage
- 3 firms (7%) use AccountingService - the markup that actually says "this is an accounting firm"
- 1 firm (2%) has LocalBusiness schema
- 2 firms (5%) have Person schema for named partners
The gap shows up on queries like "best chartered accountants near me" or "how much does an accountant cost for a limited company". For AI answers to those, Google says no special schema.org markup is needed. AccountingService is still the clearest way to state what the business is, and structured data helps pages qualify for rich results.
Almost nobody uses FAQ schema
Just one firm in 41 (2%) carries FAQPage schema on the homepage. Yet the queries AI Overviews now answer for prospective accountancy clients are almost all FAQ-shaped:
- "how much does an accountant cost UK"
- "do I need an accountant for a limited company"
- "what does a chartered accountant do"
- "can an accountant save me money on tax"
- "how do I switch accountants"
An accountancy site that publishes those questions as content and attributes the answers to a named ICAEW, ACCA or AAT qualified accountant gives AI engines a clear answer to quote. FAQPage markup is still valid Schema.org, but Google stopped showing FAQ rich results on 7 May 2026.
Method
Sample selection prioritised coverage across firm size tiers rather than statistical representativeness. From the Accountancy Age Top 50+50 ranking and the wider UK chartered accountant directory I selected: the Big 4 (4), top 10 mid-tier (6), top 25 mid-tier (10), top 50 mid-tier (15), regional and 50-100 tier (15), and a sample of independent and small-practice firms (10). Of the 56 firms in the sample, 41 returned HTTP 200 to a real-user Chrome request and form the denominator for all percentage findings.
Per-firm checks
- Reachability
- GET request with Chrome 126 User-Agent, follow up to 3 redirects, 12-second timeout. Status 200 = reachable.
- Schema parse
- Extract every
<script type="application/ld+json">block from the HTML body, JSON-parse each, walk the @graph or array, collect every @type value seen. - AI bot check
- Three additional GET requests using GPTBot/1.2, ClaudeBot/1.0 and Google-Extended user agents. Compare each status against the Chrome status from the same firm. Google-Extended is a robots.txt token with no user agent of its own, so that request does not reflect any real Google crawler. The search crawlers (OAI-SearchBot, Claude-SearchBot) were not tested.
- Robots.txt parse
- Fetch /robots.txt, parse User-agent blocks for GPTBot, ClaudeBot and Google-Extended, check for "Disallow: /" rule under each.
- Silent block flag
- Firm flagged as silently blocking when Chrome status is 200, AI bot status is 403/429/503, AND robots.txt does not intentionally disallow that bot. Filters out cases where the block is a choice the firm made.
- Sample limit
- 15 firms did not respond at all (DNS, timeout or persistent 4xx/5xx) and were excluded from the percentage findings. They are listed in the raw data set, not in the headline numbers.
How to check your own firm in 10 minutes
Three checks. None of them need a paid tool. None of them need a developer.
- Schema - paste the firm homepage URL into Google's Rich Results Test (search.google.com/test/rich-results). Look at the JSON-LD section. If you do not see
AccountingServicein the @type list, the page is not telling Google or any AI engine that this is an accountancy practice. If there is no JSON-LD section at all, the firm is in the 22% with no schema. - AI bot access - open a terminal and run
curl -A "GPTBot/1.2" -I https://yourdomain.com. If the response code is 200, the firm is reachable by OpenAI's training crawler. If it is 403, 429, 503 or 999, the firm is being silently blocked. Repeat the same command with"OAI-SearchBot","Claude-SearchBot"and"PerplexityBot"as the user agent, since those are the crawlers ChatGPT, Claude and Perplexity search use. GPTBot and ClaudeBot only control training, and Google-Extended is a robots.txt token, not a crawler you can test. Compare against a plain curl - if Chrome gets 200 and the bots get 403, the block is at the hosting layer. - AI answer test - open ChatGPT and ask "best chartered accountants in [your city]", then "how much does an accountant cost for a small limited company UK". Open Perplexity and ask the same questions. Check whether your firm is named in any of the answers and which sources are cited. If you never appear, but you know clients are choosing accountants this way, the two checks above are the first places to look.
If a firm fails all three checks, fixing them in this order delivers the fastest result: AI bot access first (a one-day fix once the hosting provider is identified), schema second (a developer task that should take an afternoon for AccountingService markup; FAQ rich results stopped showing in Google on 7 May 2026), AI Overview content third (a longer-term project of building the cited content).
What this means for UK accountancy practices
The accountancy market is shifting in three ways at once. First, AI Overviews are now triggering for the high-intent queries that drive new-client research - cost queries, "do I need an accountant" queries, comparison queries. Second, ChatGPT and Perplexity are answering recommendation queries that used to go to Google. Third, the firms that get cited in those AI answers are the ones with named-accountant E-E-A-T signals and content built as direct answers rather than generic blog posts.
This study confirms that most UK accountancy firms - including the Big 4 - are not set up for any of that. The work to fix it is not large. AccountingService schema is a single JSON-LD block. FAQPage markup is optional now: Google stopped showing FAQ rich results on 7 May 2026. Resolving a silent AI bot block is a configuration change at the hosting layer or in Cloudflare. None of it needs a website rebuild. What it needs is somebody noticing the gap and fixing it before competitors do.
This is the work I do for UK accountancy practices. The full programme is documented at Accountant SEO Consultant. The shorter version: if you want me to run the three-check audit on your own firm and tell you which of the gaps applies, send the URL via the contact form. Free, no follow-up sequence, single-paragraph reply within one working day.
Questions about the study.
How many UK accountancy firms carry structured data that identifies them?
In a May 2026 study of 56 UK accountancy firm websites - including the Big 4, top 25 mid-tier and a sample of regional practices - 32% carried either no structured data at all, or none that identified them as an accounting business. Google says structured data is not required for its AI features, so this is a missed signal rather than invisibility, but it still affects rich-result eligibility. Among the 41 reachable firms, only 7% used the AccountingService schema type. The other 93% do not state in their markup that the site belongs to an accountancy practice.
Are UK accountancy firms blocking ChatGPT and other AI crawlers?
Not on purpose. In the study, zero firms intentionally blocked GPTBot, ClaudeBot or Google-Extended via robots.txt. But 27% silently blocked at least one AI bot at the hosting layer - the server returned 403 to GPTBot or ClaudeBot while serving Chrome users normally. Affected firms include PwC, BDO, Crowe UK, FRP Advisory and TaxAssist. The block is almost certainly accidental WAF behaviour at Cloudflare, Akamai or similar. GPTBot and ClaudeBot are training crawlers, so this blocks training rather than proving the firm is missing from ChatGPT or Claude search, which use OAI-SearchBot and Claude-SearchBot. The same WAF rule is worth checking against those.
Why does AccountingService schema matter for an accountancy website?
Google says structured data is not required for its AI features, but it still helps pages qualify for rich results. AccountingService markup is a clear, low-cost way to state what the business is. Generic Organization or WebPage schema does not identify the practice as accountants. In the study, only 3 of 41 reachable firms (7%) used AccountingService markup.
How can an accountancy firm check if it is visible to AI search?
Three checks take about ten minutes. Run the homepage through Google's Rich Results Test and the Schema.org Validator to see what structured data exists - look specifically for AccountingService or LocalBusiness. Test the live server response against the OAI-SearchBot, Claude-SearchBot and PerplexityBot user agents (curl -A "OAI-SearchBot" yourdomain.com) and check the status code is 200, not 403. Those are the crawlers ChatGPT, Claude and Perplexity search use. GPTBot and ClaudeBot only control training. Then ask ChatGPT and Perplexity questions a client would ask, like "best chartered accountants in Manchester" or "how much does an accountant cost for a limited company UK", and see whether the firm appears. If the firm has no AccountingService schema, blocks AI bots at the server, or never gets cited, those are the three things to fix in order.
Does this matter when accountancy is a referral business?
Referrals still drive most new accountancy clients - but the way potential clients research firms before contacting one is shifting fast. UK searches for "how much does an accountant cost" and "do I need an accountant for a limited company" now trigger Google AI Overviews. Sole traders and small business owners are asking ChatGPT for accountant recommendations. A firm invisible to those answer surfaces loses the consideration step entirely, even when the original lead came from a personal referral. AI search visibility is now part of the credibility check, not a separate marketing channel.
Raw data and replication
The raw per-firm data is captured in the project repository - status codes for each user agent, schema types found, silent-block flag, intentional-block flag. The crawl methodology is reproducible with about 50 lines of Node.js using the standard https module and the curl one-liners above.
If a press, trade publication or research outlet wants the per-firm data table or a tailored cut by tier (Big 4 vs mid-tier vs regional), request it through the contact form.
Run the three checks on your firm.
Send the firm URL via the contact form and I will run the schema check, the AI bot access check and the AI answer test. Single-paragraph reply within one working day with the specific gaps and the fastest fix. Free, no follow-up sequence.
Send the URL