ChatGPT Prompts for Accountants to Explain Tax to Clients (Without Sounding Like a Textbook)

The short version: the right ChatGPT prompts turn dense tax rules into two or three sentences a client can repeat back to you, and that single skill saves accountants hours of repeated phone calls every quarter. Below are the exact prompts I use with accountancy clients, organised by the tax situations that come up most, plus the one uncomfortable truth about why so many practices still explain tax badly on purpose.

The real reason clients keep asking the same question three times

I sat in on a client review meeting a few years back with a nine-partner accountancy firm outside Manchester. The senior partner spent eleven minutes explaining the difference between the annual investment allowance and writing down allowances to a client who runs a small logistics company. Good explanation, technically correct, full of the right terms. The client nodded the whole way through, then turned to me afterwards and said, "So do I get the money back this year or not?"

That is the entire problem in one sentence. Accountants explain tax the way they were taught to think about it, not the way clients need to hear it. Clients do not want mechanisms, they want outcomes: how much, when, and what happens if they get it wrong.

ChatGPT is good at this specific translation job because it has no professional pride wrapped up in sounding technical. Give it the right instruction and it will strip a rule down to what matters to a non-accountant in about four seconds, which is a lot cheaper than the eleven minutes I watched happen in Manchester.

Set the rules before you write a single prompt

Three non-negotiables, and the middle one is the one accountants skip and then regret:

• Never paste real client figures, names, National Insurance numbers, or company details into ChatGPT. Use placeholders like "a sole trader earning £45,000" instead of your actual client's actual numbers.
• Always tell it who the audience is. "Explain this to a client with no financial background" produces completely different output from no instruction at all.
• Always ask for a version you can check against current HMRC guidance before you send it. ChatGPT is a drafting tool for tone and clarity, not a source of tax law, and tax bands and thresholds change every April.

If you want a longer walk through prompt structure and the data rules specifically for accounting work, I put together a separate guide: chatgpt prompts for accountants: 21 prompts that save hours, without ever pasting client data. This post is narrower, it is just the client-facing explanation prompts.

Prompts for explaining personal tax situations

These are the ones I hand to sole practitioners and small firms who spend January and February fielding the same five questions on repeat.

Self-employed income tax basics

"A client is newly self-employed, earning around £38,000 a year. Explain in plain English, using no jargon, how they'll be taxed, when payments are due, and what a payment on account is. Write it as if explaining to a friend over coffee, three short paragraphs maximum."

Dividend versus salary for a limited company director

"Explain to a first-time limited company director why accountants often recommend a mix of low salary and dividends, rather than a high salary. Use a concrete example with round numbers, avoid the phrase 'tax efficient', and keep it under 150 words."

I flagged "avoid the phrase tax efficient" deliberately. Clients hear that phrase and switch off, because it sounds like something they are meant to trust rather than understand.

Capital gains tax on selling a second property

"Write a short explanation for a client selling a rental property, covering what capital gains tax is, roughly what triggers it, and the 60-day reporting deadline to HMRC after completion. Assume they have never sold a rental property before. No headings, conversational tone, five sentences maximum."

Inheritance tax for a client who just inherited money

"Explain to someone who has just inherited £80,000 from a parent why they personally don't pay tax on the inheritance itself, but why the estate might have paid inheritance tax already. Keep it to two short paragraphs and reassure them without oversimplifying."

Prompts for explaining business tax to owners and directors

These get reused constantly because most small business owners do not know the difference between what the company owes and what they personally owe.

VAT registration threshold

"Explain to a sole trader approaching £90,000 in turnover what happens when they cross the VAT registration threshold, what it means for their pricing, and one practical decision they need to make in the next month. Plain English, no acronyms without explaining them first."

Corporation tax rates in plain terms

"A client's limited company made £65,000 profit this year. Explain how corporation tax applies, why the rate can differ depending on profit level, and roughly what proportion of profit they should expect to set aside. Round numbers only, no exact percentages that might date the answer."

That last instruction matters more than it looks. Rates and thresholds move, and a client who screenshots your explanation and quotes it back to you eighteen months later at the wrong rate is not a fun conversation to have.

IR35 for a contractor client

"Explain IR35 to a contractor who has just been told by their agency that a new contract is 'inside IR35'. Cover what that phrase changes for their take-home pay in practical terms, without using the words 'deemed employment' more than once."

Making Tax Digital for a small landlord

"Explain to a landlord with two rental properties and rental income of £32,000 a year what Making Tax Digital for Income Tax will mean for how they keep records, using a simple before-and-after comparison. Keep it under 120 words."

Prompts for the emotional side of tax conversations

Half of what clients need explained is not the tax rule, it is the reassurance underneath it. This is the part most accountants under-use ChatGPT for, and it is where it earns its place.

Explaining a tax bill that came as a shock

"Write a short, calm message to a client who has just been told they owe £4,200 more than they expected in tax. Explain that this can happen because of how payments on account work, without sounding defensive, and end with a clear next step."

Explaining a late filing penalty

"Write a client-facing explanation of why HMRC's late filing penalty structure means a fine can appear even if the tax owed is zero. Keep it factual, not apologetic, and end with what the client should do next."

Reframing a refund so it doesn't sound too good to be true

"Explain to a client receiving an unexpectedly large tax refund why this happened, in a way that reassures them it's correct rather than a mistake that will be reversed later."

A step-by-step way to use these day to day

Here is the workflow I recommend to firms, tested with a bookkeeping practice in Leeds that now uses a version of this for every new client onboarding pack:

• Step 1: pick the tax topic that comes up for this client type. Don't write a generic explanation, write for the exact situation.
• Step 2: run the prompt with placeholder numbers, never the client's real figures.
• Step 3: read the output out loud. If you stumble over a phrase, a client will too. Edit it.
• Step 4: swap the placeholder numbers for the client's real ones by hand, after you've closed the ChatGPT conversation.
• Step 5: save the good ones. That Leeds practice now has 14 saved explanation templates covering their most common client questions, and their office manager estimates it cuts about 6 hours a month off partner time spent on repeat phone calls.

Six hours a month does not sound dramatic until you multiply it by an hourly rate of £150 to £250, which is where most UK partner time gets billed when it isn't lost to explaining the same VAT threshold question for the ninth time that quarter.

The uncomfortable part nobody in the profession likes to say out loud

Here is the bit that makes some accountants twitch when I say it in workshops: a proportion of the profession has a quiet financial incentive to keep tax explanations a little bit confusing. Complexity is part of what clients are paying for. If a client fully understood every rule, some of them would start asking why they need you at all, or worse, why the fee is what it is.

I don't think this is usually a conscious decision. It is more that clarity has never been rewarded the way technical accuracy has, so nobody built the habit. But the accountants who use plain-English explanations well are not losing clients over it, they are winning referrals, because a client who understands their own tax position tells their business owner friends "my accountant explains things " and that sentence sells more new work than another technically flawless but unreadable HMRC letter ever will.

The firms doing this best right now are not the biggest ones. They are small, two to six partner practices who realised that the explanation is the product, as much as the filing is.

What to do when a prompt still comes out too technical

This happens more than people expect, usually because ChatGPT defaults to hedging language when tax rules do have exceptions. Two fixes that work almost every time:

• Add "explain this as if the client is intelligent but has zero financial training" rather than just "explain simply." The first gets you respect without jargon, the second sometimes gets you condescending.
• Ask it to rewrite its own answer: "Rewrite that in half the length, removing every word that isn't necessary to understand the outcome." This second pass is usually where the sharpest version appears.

If you're building this into how your whole practice works rather than using it ad hoc, that is a slightly bigger project than a list of prompts, and it is closer to what an AI implementation coach would help set up across a team so everyone is using the same standard rather than each partner improvising their own version.

Prompts organised by client type, if you want a quick reference

• New sole traders: income tax basics, payment on account, allowable expenses in plain terms.
• Limited company directors: salary versus dividends, corporation tax, director's loan account explanations.
• Landlords: Making Tax Digital, capital gains on sale, mortgage interest relief changes.
• Contractors: IR35 status, umbrella versus limited company trade-offs.
• Retirees and pension drawdown clients: tax-free lump sum limits, how drawdown income is taxed.

If you work across multiple client professions rather than just accountancy clients, there's a wider set of role-based prompts worth having in your back pocket in ChatGPT Prompts List by Job: 60 Prompts for 12 Professions, and if you're still working out the basics of how ChatGPT handles limits and longer sessions when you're drafting a lot of these in one sitting, that's covered in ChatGPT Prompts and Limits.

Related: chatgpt (this site).

Frequently asked questions

Is it safe to use ChatGPT to explain tax to clients?

Yes, as long as you never type real client names, figures, or account details into it, and you check every output against current HMRC guidance before sending it, since ChatGPT can be confidently wrong on exact rates and thresholds that change each tax year.

Will clients notice if an explanation came from ChatGPT?

Not if you edit it. Read every output aloud and fix any phrase that sounds unnatural coming from you specifically, because a client who has worked with you for years will notice a sudden change in tone faster than you'd expect.

What's the single best prompt to start with?

Pick your most repeated client question, the one you've answered on the phone at least five times this year, and ask ChatGPT to explain it "as if the client is intelligent but has zero financial training, using a concrete example with round numbers." That one prompt structure covers most tax situations.

Can these prompts replace proper client communication or advice?

No, and they shouldn't try to. They speed up the first draft of an explanation you would have written anyway, they don't replace your professional judgment on what advice a specific client needs.

Official documentation

• OpenAI's own platform documentation