What 494 Accountancy Websites Told Us About Why SEO Keeps Failing Your Practice

Most accountancy practices are not losing search visibility because of a technical fault. They are losing it because their website never mentions the work they most want to sell. We measured this across 494 UK practices in September 2026, and the single commonest finding was not a broken tag or a missing sitemap. It was that 344 firms, 69.6 per cent of the sample, had no page at all for their highest value service.

That is a commercial problem wearing a technical costume, and it explains something practice owners tell us constantly: the SEO was done, the report looked fine, and the phone still rang with self assessment enquiries at £250 a time.

I ran this study for a specific reason. I run an SEO consultancy and I am a qualified accountant who no longer practises, so I work for accountants instead of being one. That combination is unusual, and it is the whole point. To a web developer, a CT600 and a Fractional FD are two items in the same list, set in the same font, priced the same in their heads. One is a £400 form. The other is a £3,000 a month relationship. Nobody who has never sat in a practice can see that difference on a page, which is why the pages get built the way they do.

This case study sets out what we found, what each finding costs a practice in search and in AI citation, and what a practice owner should do about it. Every finding carries the number, the chart and a plain statement of what it means for your firm.

Key takeaways for a practice owner

The study, in one section

We audited the public websites of 494 UK accountancy practices with a single instrument over two days. The sample was drawn from the published ACCA and ICAEW practice registers with probability proportional to size across twelve UK regions, which matters because almost every other commentary on professional services websites comes from agency casework. Agency casework is a sample of firms that had already decided their website mattered enough to pay somebody about it, so it tells you nothing about the market.

Thirty six website faults were recorded per firm alongside platform, page count and region. Two interpretive check families were validated blind against 213 passages from 42 firms held out of all earlier development. Prevalence carries Wilson intervals. Pairwise association was tested by Fisher’s exact test and re-tested under Cochran Mantel Haenszel stratification by site size quartile, with Benjamini Hochberg correction throughout.

Getting to 494 usable firms took 926 domains. The attrition rate was 46.7 per cent, mostly single page sites and sites that would not render.

Three limits are stated up front because they govern how every number below should be read. We measured what a website publishes, never whether a firm complies with anything. We recorded no rankings, no traffic, no enquiries and no revenue, so any figure converting a prevalence rate into money would be invented. And the paper publishes its own instrument error rate beside every finding, which is why no firm is named anywhere in it.

The full paper is open access and carries a DOI, so nothing here has to be taken on trust.

Finding one: the service you most want to sell has no page

Sixty nine point six per cent of firms, 344 of 494, have no dedicated page for their highest value service. They have pages for the compliance work instead. Advisory, virtual finance director work, R&D claims, employee ownership trusts and exit planning are mentioned in passing or not at all.

This is the finding that costs the most, and it is invisible on every SEO report ever produced, because a report measures the rankings of pages that exist.

What it costs you in search. A page cannot rank for a query it does not address. A practice with no virtual FD page cannot appear for “do I need a finance director”, “part time FD for a small business” or “fractional CFO versus accountant”. Those are the searches a real virtual FD buyer makes, and note what they have in common: the buyer is not searching for an accountant at all. He is searching for a problem. Your “accountants in Manchester” page will never intercept him, however well it ranks.

What it costs you in AI citation. An AI assistant answering “should I hire a fractional finance director” builds its answer from sources that discuss the decision. A practice with no page on the subject has published nothing for the model to retrieve, nothing to quote and nothing to attribute. You are not ranked low. You are absent from the retrieval set. Recovering from absent is a different job from recovering from position eleven, and it takes content, not tweaks.

The related finding sharpens it. Forty five point one per cent of firms never name a client type they serve. Another 59 per cent mention client types only in passing body text, which we labelled weak and refused to count as specialism. A search engine and a language model both work by resolving entities and relationships. A site that never states “we act for pharmacies” or “we act for construction subcontractors under CIS” gives neither of them a relationship to store.

We wrote about where those pages should point in our guide to the practice niches with real search demand and almost no competition, which is the natural next step once you accept this finding.

Finding two: six firms in ten never say who regulates them

Three hundred and one firms, 60.9 per cent, never state who regulates the practice anywhere on the site. Seven point five per cent publish professional indemnity insurance details. Two firms in 494 name an anti money laundering supervisor.

That last number looks like a broken detector. It is not. The validation pack held 34 candidate passages from 22 firms and two blind raters found no qualifying statement in any of them. Where an instrument finds almost nothing and independent readers confirm nothing is there, the absence is the finding.

Read this carefully, because the honest version is narrower than the alarming version. Every practising firm is supervised. The Provision of Services Regulations allow insurance details to be shown at the office rather than online. The finding is that a prospective client cannot find it on the website. It is never that the firm has no supervision or no insurance.

What it costs you in search. Regulatory status is the clearest trust and expertise signal a professional services site can publish, and it is exactly the kind of corroborating detail that quality raters and quality systems look for. A page that claims expertise without a single verifiable credential is a page asserting authority rather than demonstrating it.

What it costs you in AI citation. A language model deciding which of five local practices to name in an answer needs a reason to prefer one. Regulated status, a named supervisor, a membership number and a named human being are all facts it can lift and attribute. A site with none of them offers nothing to prefer it by. Twenty one point seven per cent of firms have no about or team page at all, and eight per cent have no identifiable human anywhere on the site. An assistant asked “who would I actually be speaking to” finds nobody.

Finding three: five faults that are really one failure

Missing local business markup, an untappable telephone number, conflicting telephone numbers, an absent footer postcode and a missing contact page all associate strongly with one another once site size is held constant, with common odds ratios running from 4.3 to 56.5.

These are not five faults on a checklist. They are one failure, in which nobody owned the firm’s contact details, expressing itself in five places. Seventeen point four per cent of firms carry no local business markup at all. Eight point five per cent publish a telephone number that is never a tel: link, which on a phone means a client has to copy it by hand.

What it costs you in search. Local visibility depends on a search engine resolving your practice to a real place with a consistent name, address and telephone number. Conflicting numbers and an absent postcode break that resolution. The practical effect is that the firm is findable as a business and not findable as a local one, which is the difference between appearing in the map pack and not.

What it costs you in AI citation. An assistant asked “find me an accountant in Cardiff who handles CIS” needs to return something a person can act on. A firm with no structured contact data supplies no address to state and no number to dial. It gets left out of the answer in favour of a firm that supplied both, regardless of how good the practice is.

The fix here is cheap, and this is the group of findings we normally clear first inside our technical and on page SEO work, because it moves local visibility faster than anything else on the list.

Finding four: a bigger website is not a healthier one

Almost every finding correlates with site size, and the direction of that correlation splits the faults into two families that need opposite remedies.

Omission faults become rarer as sites grow. The firm never built the thing. Naming a client type is absent on 84 per cent of the smallest quartile and 11 per cent of the largest. A page for a high value service is missing from 98 in every 100 of the smallest quartile against 39 in the largest. Analytics, structured data, an about page and a regulatory statement all follow the same pattern.

Neglect faults become commoner as sites grow. The firm built something and stopped tending it. Superseded tax figures appear on 4 per cent of the smallest quartile and 41 per cent of the largest. Broken links rise from 12 to 51 across the same range. Duplicate titles rise from 30 to 67.

What it costs you in search. A remedy aimed at the wrong family does nothing. A small practice does not need a technical audit. It needs pages that do not exist yet. A large practice does not need more pages. It needs the ones it has to stop contradicting each other and stop publishing a VAT threshold that changed years ago.

What it costs you in AI citation. Stale figures are the sharpest version of this. Eighteen point four per cent of the panel publish a tax figure HMRC has replaced, and it rises to 41 per cent among the largest sites. A language model that retrieves your page and finds a superseded threshold either repeats your error, which is a liability, or learns that your site is not a reliable source on tax, which is worse and lasts longer. Currency of information is one of the few things a machine reader can check cheaply, and accountancy is the sector where it is checked hardest.

Practices that publish regularly are the ones exposed here, which is why we set out a maintenance rhythm rather than a publishing sprint in our note on how an accountancy blog should be structured so it keeps earning after it is written.

Finding five: your platform decided this, not your practice

Platform predicts technical condition more strongly than any deliberate choice we measured. Sitemap absence runs at 0 per cent on Wix, 15 per cent on WordPress and 55 per cent on bespoke builds. Missing structured data runs 5 per cent, 15 per cent and 45 per cent across the same three.

Two readings follow. Technical hygiene is very largely a platform default rather than an achievement. And platforms trade one failure mode for another: the platform with no missing sitemaps and almost no missing structured data also has the highest rate of firms with no page for the work they most want to sell. It is technically tidy and commercially silent.

What this means for you as a buyer. A supplier who shows you a clean technical scorecard may be showing you your platform’s defaults with their name on the report. Ask which of the green ticks they caused. The findings that actually needed somebody to decide something, the service pages, the client types, the regulatory statement, are the ones that stay red across every platform.

Finding six: the file that speaks to AI assistants, and who chose to publish it

Eighty eight firms, 17.8 per cent of the panel, publish an llms.txt file, the proposed convention for supplying site information to language model agents. That rate is far above the general web, and it is not deliberate.

Adoption is total on two platforms and absent on three others. Every Wix site in the sample has one. Every Duda site has one. Eleven per cent of WordPress sites have one, and there it tracks the presence of particular SEO plugins, which name themselves inside the generated file. Two of the 88 files introduce the firm by a platform placeholder name rather than its trading name, in a document written specifically for machine readers.

Meanwhile, no firm in the panel blocks a language model crawler in robots.txt. Whatever position a practice might hold on the use of its published material for model training, not one of the 494 has expressed one.

What it costs you. Both facts describe the same condition. This is an area of consequential configuration in which the platform decides and the practice does not know a decision has been taken. The decision is not costless in either direction either: research reported by Grossman and colleagues in 2026 found that sites blocking a major AI crawler receive reduced visibility in that engine’s generated summaries, so a practice choosing to block would be trading one kind of exposure for another. That is a decision worth taking deliberately. None of these firms has taken it at all.

Getting named inside an AI answer is now a separate discipline from ranking, and it is what our AI and LLM search visibility service exists to do: restructure a site so that assistants can retrieve it, attribute it and quote it by name. The groundwork for it is covered in our guide to preparing a site for AI and voice search before the traffic shifts.

Finding seven: 38 of our own 53 findings were never real

Our first pass tested 630 pairs of findings and produced 53 significant associations after correction. Re-testing with each firm compared only against firms of a similar size produced 17. Only 15 of those 17 were among the original 53. Thirty eight collapsed, and two associations appeared that the first test had missed entirely.

The 38 were not relationships. They were pairs of attributes that both happen to be commoner on small sites. One example: an apparent negative association between broken links and missing analytics vanished once size was held constant, because larger sites have more links to break and are likelier to be measured. The two attributes move in opposite directions for reasons that have nothing to do with each other.

We report this prominently, and second in the paper, before any finding, because the same trap is available to anyone auditing websites at scale. An uncontrolled correlation between two website attributes is very easy to produce and very hard to tell apart from a real finding.

This is the most useful thing in the study for a practice owner. A practice acting on that first pass, or an adviser billing against it, would have directed remediation at 38 relationships that do not exist while missing two that do. It is the one cost this research actually measured, and it is a cost of analysis rather than of websites.

Why the usual supplier cannot fix this

The gap is industry knowledge, and it is not a criticism of developers or marketers. It is a description of what they were hired to know.

A web developer builds what the brief describes. A marketer optimises for the terms the brief names. Neither was ever given the information that would let them see the problem, which is that the services on a practice’s list have wildly different economics. Set out as menu items in the same font, a CT600 and a virtual FD engagement look like siblings. One is a £400 form that a client shops on price. The other is a recurring relationship worth thousands a month that a client shops on trust and specialism. A site built without that knowledge will treat them identically, and it will optimise for the cheap one because the cheap one has more search volume.

That is exactly what the data shows. Seventy per cent of firms have no page for the expensive work. Ninety five per cent of Wix firms have none. The pattern is too consistent to be a run of bad suppliers. It is a knowledge gap, and it sits between two professions that do not overlap.

Two practices, and what changed

A four partner practice in the south west built a virtual finance director service and priced it at £2,500 a month. Eighteen months on it had two clients, both referrals. The website carried a single line about it inside a services list, and the firm ranked well for “accountants in” its town, which produced around 25 enquiries a month, almost all sole traders wanting bookkeeping at £40 a month. We built four pages addressing the decision the buyer was actually making rather than the service the firm was selling, covering when a business needs a finance director, what a part time FD costs, and how the role differs from an accountant. Enquiry volume fell to 19 a month. Enquiries for the virtual FD service went from none to six, and three converted inside the first quarter. Lower traffic, materially better fees.

A twelve person practice in the midlands had the opposite problem. Ninety pages, a busy blog, strong rankings, and 51 of those 90 pages sharing a title with another page. Its tax rates page still stated the VAT registration threshold as £85,000, and its pension content still described the Lifetime Allowance as currently set at £1,073,100, two years after it was abolished. Nothing needed building. We consolidated the duplicate titles, dated every figure and put a quarterly review on the tax pages. Rankings for the firm’s advisory terms recovered within two months without a single new page, because the site stopped contradicting itself.

Both are illustrative composites drawn from the pattern in the data, with figures kept realistic and internally consistent. The named client results below are real.

Our own client work follows the same shape. Total Books Accountants Ltd in Cardiff grew from roughly 200 to about 1,600 organic visits a month while repositioning towards advisory and virtual finance director work, with an optimised Google Business Profile now driving more than 80 per cent of its local visitors. RX Virtual Finance Ltd, which specialises in pharmacy accountants, reached number one for “pharmacy accountants” and the first page within about 60 days, on a niche nobody else had claimed.

Where else this pattern shows up

The mechanism is not unique to accountancy. It appears in every sector where the cheapest service has the highest search volume and the most valuable service has the highest trust barrier.

Financial advisers publish pages about pensions and nothing about the intergenerational wealth conversation that produces the large engagement. Law firms rank for “solicitors near me” and have no page on the shareholder dispute work that pays for the office, which is the same failure we address in our work with law firms. Dentists rank for check ups and bury implants. Healthcare providers list conditions rather than the pathways patients search for. Hotels compete on room rates against portals rather than on the reason somebody is coming to the city. Letting agents chase tenants when landlords are the fee. Ecommerce brands optimise the cheapest SKU because it converts fastest, and starve the range that carries the margin.

In every one of those, a supplier without sector knowledge will optimise towards volume, because volume is what the tools show them. Naming the high value service and building the page that answers the buyer’s actual question is the same fix each time, and it is what our semantic SEO and topical authority work is built to do.

What a practice owner should do this month

Answer these seven questions about your own site. Each takes under a minute, and the research says most practices will fail at least four.

1. Name your highest fee service. Does it have its own page? Not a line in a list. A page.
2. Search your site for a client type you actually act for. Pharmacies, construction, hospitality, dental. Is there a page, or a passing mention?
3. Look at your footer. Does it state who regulates the practice, and is your postcode there?
4. Tap your phone number on a mobile. Does it dial, or does it just sit there as text?
5. Open your tax rates page. Is every figure on it current today, and is the page dated?
6. Search your site for one of your service page titles. Does more than one page carry it?
7. Ask an AI assistant to recommend an accountant for your specialism in your town. Are you named? Where you are not, ask it what it would need in order to name you.

Question seven is the one most owners have never tried, and it is the one that tends to end the conversation about whether any of this matters.

Frequently asked questions

Read the research, then talk to somebody who has read your fee ledger

The findings in this case study come from published, citable research rather than from agency anecdote, and the paper reports its own error rate beside every number. That is the standard we think a practice owner should hold any supplier to, including us.

True SEO Consultants Ltd, based at Startup Stiwdio, University of South Wales, 86-88 Adam Street, Cardiff, CF24 2FN, works with practices across the UK and onboards clients worldwide through a fully remote digital onboarding and delivery process, so the work described here applies whether your practice is in Cardiff, Manchester, Edinburgh or further afield. We work as a growth and visibility mentor rather than a typical agency selling a fixed number of blog posts, and we bill against agreed KPIs rather than hours.