AI for accountants: the 2026 UK guide

What does AI actually change for a practice?

Three things, in order of how quickly they hit you.

It removes time from routine work. Bookkeeping categorisation, document extraction, bank reconciliation queries, first drafts of client letters, summarising a set of management accounts into three bullet points a client will actually read. This is the layer most of the profession is already working on, and it is the least interesting one, because everyone is doing it and it eventually shows up as fee pressure rather than margin.

It changes what clients expect. Your client has already asked an AI assistant whether they should incorporate, and arrived at your meeting with a view. Sometimes it is a good view. Often it is a confident half-answer that ignores their actual circumstances. The job has shifted from supplying information to correcting a plausible-sounding wrong answer, which takes more skill, not less.

It changes how clients choose you. This is the one almost nobody in practice is watching, and it is the reason this guide exists. AI answers now sit between the client and the search results, and they name a shortlist. If your firm is not in that shortlist, you were never in the running, and no amount of good work on the other two layers fixes it.

Will AI replace accountants?

No, but it changes what a practice can charge for.

AI is genuinely good at extraction, categorisation, summarising and drafting. It is poor at judgement, poor at knowing which question the client should have asked instead of the one they did, and structurally incapable of carrying professional responsibility, which under the ICAEW Code of Ethics stays with the member regardless of what tool produced the output.

The practical read: a practice whose fee income rests mainly on compliance processing is exposed, because the cost of that processing is falling for everyone including your clients. A practice selling judgement, sector knowledge and the ability to be phoned at short notice is not. That is not a new observation, but AI has moved the deadline forward.

The uncomfortable corollary is that positioning now matters more than it did. A firm that says "we do accounts, tax and payroll" is describing the part that is getting cheaper. A firm that says "we are the accountants for construction contractors in south London" is describing something a model cannot supply.

Which AI tools are UK practices actually using?

Most practices are not buying a dedicated AI product. They are getting AI through software they already pay for. That is worth knowing before anyone signs up for something new.

UK monthly search volumes from DataforSEO, September 2026. Volume indicates market awareness, not quality.

Two observations from those numbers. First, the tools with the deepest UK adoption are the boring ones that do extraction, not the ones marketed as AI. Second, "AI for accountants" itself attracts 1,600 UK searches a month, which tells you the profession is actively looking, and that whoever answers the question well will be found.

A practical filter before buying anything. Ask what the tool removes from your week, in hours, and who in the practice will own it. If nobody can answer both, the licence will go unused by March. This is the same test you would apply to any other software, and AI does not earn an exemption from it.

How is AI changing the way clients find a firm?

Directly, and faster than most practices realise. Here is what we found when we ran the UK search results in September 2026.

Search "SEO for accountants" in the UK and Google returns an AI Overview at position one, above every organic result. The first actual website appears below it. Anyone who reads the AI answer and acts on it never sees the ranked list at all.

Search "accountants website design" and the AI Overview appears at position two, and it does something more pointed: it names providers. When we checked, it named exactly two. Not ten, not a page of options. Two. Every other firm competing for that market was, for that user, invisible.

Now apply that to your own market. A business owner asks ChatGPT for an accountant for their contracting business in south London. The model returns three or four names. Being the fifth-best firm in the borough used to mean you got some of the traffic. It now means you get none of that conversation.

Who is AI actually citing?

This is the useful part. When we looked at which sources the AI Overview drew on for accountancy queries, the pattern was consistent and it is not what you would expect.

• Software vendors such as Xero and TaxDome, which publish long, well-structured guides.
• Training and compliance bodies such as Mercia Group and 2020 Innovation.
• Roundup and list articles in formats like "15 tips" and "17+ examples".
• Occasionally Reddit, which tells you these models value a direct answer over a polished one.

Accountancy firms themselves barely appear. Not because their expertise is worse, but because their websites are written as brochures. A model cannot quote "we pride ourselves on a personal service". It can quote "MTD for Income Tax applies from April 2026 for sole traders and landlords with qualifying income above the first threshold".

What is AI visibility and how is it different from SEO?

AI visibility, sometimes called generative engine optimisation or GEO, is the work of getting your firm named inside AI-generated answers. It shares foundations with SEO but the objective is different: SEO gets you ranked, AI visibility gets you quoted.

They are not alternatives. The same page should do both, and a page built properly for AI citation tends to rank well anyway, because clear answer-first writing is what users wanted all along.

How do you make a practice citable by AI?

Seven things, in the order they are worth doing.

1. Answer the question in the first two sentences

Most practice websites open with context. "In today's fast-moving business environment, choosing the right accountant is more important than ever." A model cannot use that. Open instead with the answer: "A limited company must file accounts within nine months of its year end. Late filing triggers an automatic penalty starting at £150." Then give the context underneath.

2. Phrase headings as the questions people ask

"Our Services" is not a question anybody types. "How much does an accountant cost for a small limited company?" is. Headings that match real queries give a model an obvious place to lift from.

3. Use specific numbers, and date them

"We save clients money" is unciteable, and it also breaches the ICAEW rules on unsubstantiated claims. "The VAT registration threshold is £90,000 as at September 2026" is citeable and compliant. Specificity is doing double duty here.

4. One idea per paragraph

Models extract passages. A paragraph covering three ideas gets skipped in favour of one covering a single idea cleanly. This also makes your site easier to read, which was worth doing regardless.

5. Add schema a model can parse

FAQPage, Service, LocalBusiness and Article schema make the structure of your content machine-readable rather than something to be inferred. It is invisible to visitors and it is the single most neglected item on most practice websites.

6. Publish the guide nobody else in your niche has written

The citations go to whoever answered the question thoroughly. In accountancy there are still obvious gaps: what the ICAEW marketing rules actually permit on a website, how MTD affects a landlord with four properties, what a dental associate should consider before incorporating. Pick the one your clients ask most.

7. Check your own visibility monthly

Open ChatGPT, Google and Perplexity. Type the question a prospective client would type. Write down which firms are named and which sources are cited. It takes ten minutes, and it is the only way to know whether any of the above is working.

What do the ICAEW and ACCA rules say about AI?

There is no separate AI rulebook. The existing ethics framework applies in full, which is stricter than many practitioners assume.

The practical minimum for a practice: a one-page AI use policy naming the approved tools, a rule about what never goes into a prompt, and a line in your engagement letter. Most firms have none of these, and it is a straightforward gap to close.

Where should a practice start?

If you do one thing this month, do the visibility check. It costs nothing and it usually changes the conversation.

The window here is genuinely open. The commercial accountancy search terms we screened for this brand returned keyword difficulty scores of zero on nine separate phrases carrying costs per click between £18 and £68. That combination, valuable traffic with almost no competition, is not a permanent condition. It is what an emerging channel looks like before the profession notices.