Canadian firms lose money after using AI for tax advice

Canadian businesses are losing money after using general-purpose AI tools such as ChatGPT for tax, bookkeeping and financial guidance, according to a survey of accountants and bookkeepers commissioned by Dext.

The survey of 500 accounting and bookkeeping professionals found that 50% said they were aware of businesses that had suffered direct financial losses after acting on incorrect or misleading AI-generated advice. Respondents cited overpayments, missed allowances, penalties, fines and compliance issues as outcomes they had seen.

The survey results also pointed to a rise in AI use among business clients. Some accountants said clients used AI outputs in discussions about professional advice and the value of accounting services.

“The damage is no longer hypothetical,” said Paul Lodder, VP Accounting Product Strategy, Dext. “Businesses are already losing money, and accountants are spending valuable time correcting avoidable mistakes, from tax and payroll errors to misinterpretation of expenses.

“AI has a powerful role to play in finance but there's a fundamental difference between specialist tools built for accounting and bookkeeping, and general-purpose chatbots that don't know a business's true financial context.”

Client behaviour

Respondents said they saw a sustained increase in client use of public AI tools during 2025. Some 76% of accountants and bookkeepers said they had seen more clients using tools such as ChatGPT or other large language models for financial, tax or bookkeeping queries.

Professionals also reported that AI-generated outputs increasingly featured in conversations with clients. The survey found that 70% had seen an increase in clients using AI outputs to question or challenge professional advice. A further 68% said they had seen a rise in clients suggesting AI could replace the need for professional accounting services.

The research linked the growing use of general-purpose AI for finance questions with an increase in errors found in client records. Only 7% of respondents said they had never encountered public AI-driven mistakes in client work.

Common errors

Accountants and bookkeepers said mistakes associated with public AI outputs appeared frequently. The survey found that 11% encountered client mistakes on a daily basis. Another 29% said they saw them weekly. A quarter said they saw mistakes monthly.

The most common errors respondents reported related to the interpretation and treatment of expenses. Some 44% cited incorrect interpretation of business expenses. Another 43% cited incorrect tax claims or charges.

Respondents also reported flawed personal tax planning at 36%. Payroll errors appeared at 35%. Incorrect business tax planning advice also came in at 35%.

The survey results also pointed to time spent correcting issues. Among respondents who encountered public AI-related mistakes, 44% said they spent up to three hours per month fixing errors caused by AI-generated advice. Another 27% said they spent between four and six hours. A further 11% said they spent seven to 10 hours.

Risks in 2026

Respondents said they expected the risk profile to worsen in 2026 if businesses continued to rely on public AI tools without professional oversight. The survey found that 27% warned of a higher risk of insolvency or business failure.

Other concerns centred on inappropriate claims and compliance outcomes. Some 42% expected increased misuse of AI outputs to justify inappropriate or fraudulent claims. Another 40% predicted rising fines and penalties. A further 38% anticipated greater scrutiny from the Canada Revenue Agency due to incorrect or late filings.

Nearly half of respondents, 46%, said they believed it would become more common for businesses to make decisions based on false confidence from incorrect AI outputs.

Regulation debate

The survey found a strong push among respondents for restrictions on how public AI tools present information connected to tax and financial advice. Some 94% said regulation and restriction was needed.

Within that group, two-fifths said public AI tools should face restrictions when providing financial or tax-related advice. Another 62% called for formal regulation.

One area of concern among accounting professionals involved how businesses interpret AI outputs. Some accountants said clients treated responses from chatbots as if they were tailored guidance, even when the tools lacked business-specific records and context.

“If we head into 2026 with more businesses treating AI outputs as trusted tax and financial advice, without professional oversight, the consequences could be severe. The focus now should be on responsible guardrails, clearer restrictions around financial advice, and better education for businesses on what these tools can and cannot safely be used for,” said Lodder.

Dext said the survey covered accountants and bookkeepers across Canada and included a range of firm sizes, regions and industries.