ChatGPT for accounting: what it can do, and what to keep out of it

Most accountants who use AI started with ChatGPT. It is genuinely useful for drafting and explaining, weak at anything that has to tie to a ledger, and governed by data terms that depend on which plan you’re on.

Updated . By AI Accounting Vendors. Not affiliated with OpenAI.

Where ChatGPT helps accountants

  • Drafting. Client emails, engagement-letter language, process documentation and first drafts of memos, written in your voice and then edited by you.
  • Explaining. A plain-language walk through a concept for a client or a new staff member, as long as someone who knows the subject checks it.
  • Spreadsheets. Writing and debugging Excel or Google Sheets formulas, lookups and pivot setups.
  • Summarizing. Pulling the points out of a long document you’re allowed to share with it.

Where it goes wrong

A general chatbot isn’t connected to your ledger, so any number it gives you about a client’s books is either something you pasted in or something it made up. It can state a rule, a threshold or a citation that doesn’t exist in the same confident tone as one that does. And it keeps no audit trail a reviewer could follow.

That is the gap purpose-built tools try to close. The tools in this directory connect to accounting systems, cite their sources, or keep a record of what the AI proposed and who approved it. 64 of the 72 listed describe a review step. See the directory or start with AI tax research tools if research is your use case.

What OpenAI says about your data

The answer depends on the kind of account, according to OpenAI’s own documentation:

Personal ChatGPT accounts
OpenAI’s Data Controls FAQ says that turning off “Improve the model for everyone” in Settings, under Data controls, stops new conversations from being used to train its models. Check the setting rather than assuming it. Temporary chats aren’t used to improve models, though OpenAI may keep them for up to 30 days for safety.
ChatGPT Business, Enterprise and the API
OpenAI’s enterprise privacy page says it doesn’t train its models on your data by default, and describes SOC 2 Type 2 audits for ChatGPT Business, Enterprise and the API Platform.

For a firm, the practical consequence is simple: decide which accounts staff may use for client work, and make sure nobody pastes client data into a personal account without checking its data controls.

Client and tax data: the extra rules

Tax return information carries federal rules on disclosure and use. Treasury regulations at 26 CFR 301.7216-2 set out what a preparer may disclose without the taxpayer’s consent, and the FTC’s Safeguards Rule lists tax preparation firms among the businesses that must keep a written security program and oversee their service providers. Neither mentions ChatGPT, and this isn’t legal advice, but both are reasons to treat a chatbot like any other vendor before client data goes in.

A one-page AI policy for a firm

Most firms don’t need a long document. They need four decisions, written down:

  1. Which AI tools and account types staff may use for client work, and which are off limits.
  2. Which data may never go into them: for many firms, tax return information and anything identifying a client.
  3. That AI output is reviewed by a person before it reaches a client, a return or the books.
  4. Who approves new AI tools, using the same questions you’d ask any vendor.

For the fourth, the AI vendor questionnaire builds the question list for you.

ChatGPT and accounting: questions

Can ChatGPT do accounting?
It can help with parts of the work, such as drafting client emails, explaining a concept, writing spreadsheet formulas and summarizing documents. It isn't connected to your ledger, it can state wrong figures or rules with confidence, and it doesn't keep an audit trail. Treat its output as a first draft for a professional to check.
Does ChatGPT train on what I type?
For personal accounts, OpenAI's Data Controls FAQ says that turning off "Improve the model for everyone" in Settings, under Data controls, stops new conversations from being used to train its models. Temporary chats aren't used to improve models, though OpenAI may keep them for up to 30 days for safety. For ChatGPT Business, Enterprise and the API Platform, OpenAI says it doesn't train its models on your data by default.
Is ChatGPT safe for client data?
That depends on the plan and your obligations. Business plans carry different data terms from personal accounts. Tax preparers also face federal rules on disclosing and using tax return information (26 CFR 301.7216-2), and firms covered by the FTC Safeguards Rule must oversee service providers. Many firms bar client data from personal accounts entirely. This is not legal advice.
Can accountants use ChatGPT for tax research?
As a starting point only. A general chatbot can misstate tax law fluently and may not cite primary sources you can check. Tools built for tax research ground answers in the Code, regulations and rulings and cite them; see the AI tax research category.

Sources

Official documents read on October 15, 2026. Vendor facts are sourced on each listing.

  1. OpenAI: Enterprise privacy at OpenAI (ChatGPT Business, Enterprise and API Platform)
  2. OpenAI Help Center: Data Controls FAQ
  3. 26 CFR 301.7216-2, Permissible disclosures or uses without consent of the taxpayer (eCFR)
  4. FTC Safeguards Rule: What Your Business Needs to Know (Federal Trade Commission)