When Ms Myra Donohue’s father, who is a financial planner, offered to help her with her finances, she politely declined.

With a background in accounting, she did not want anyone else’s help managing her US$5,000 (S$6,400) of credit card debt, not to mention car payments and other bills. But once Ms Donohue, 28, sat down with her finances, she realised how overwhelming the process would be, especially with two young sons and her partner recently laid off from his electrician job.

She found a way to approach it in July in a Facebook post about how to tackle your finances in seven steps. Step one was simple: List your income and fixed expenses and create a zero-based budget, a method that assigns a specific job to every dollar earned. Remembering how quick ChatGPT had been when she tried using it for other tasks, she decided to plug her numbers into the artificial intelligence (AI) chatbot and ask it to create a personalised financial plan.

Ms Donohue said she was not surprised by the advice the chatbot offered, but she was pleased by how quickly – within seconds – it generated a tailored budget for her.

Myra Donohue said using a chatbot to help set her household budget has given her more confidence handling her finances.

PHOTO: ANDRI TAMBUNAN/NYTIMES

AI chatbots have taken on many roles, including therapist, career coach, even romantic partner or friend. As more than half of Americans manage their finances on their own, many are turning to chatbots like OpenAI’s ChatGPT and Google’s Gemini for help tackling debt, finding better ways to save or figuring out how to invest in the stock market.

Two-thirds of adults who have used generative AI (Gen AI) said they had used it for financial advice, and around 80 per cent of those who acted on that advice said it had improved their financial situation, according to a recent survey of more than 1,000 people by Intuit Credit Karma.

But financial advice served up by chatbots should be handled with caution. The Credit Karma survey found that more than half of the Americans who acted on the financial advice offered by Gen AI said they had made a poor financial decision or a mistake in trying to follow the guidance.

Advisers recommend that people double-check AI’s suggestions and any links it provides with a professional before acting on them. Chatbots tend to take questions at face value and fail to challenge underlying assumptions or gather important context, which could be risky when seeking money advice, said Mr Tyler Gilley, an associate wealth adviser at investment firm Halbert Hargrove.

Still, Ms Donohue said, she has gained more confidence about managing her bills and finances.

“It was really about me getting back on the horse,” said Ms Donohue, a human resources administrator in Grass Valley, California. “I wanted a professional kind of service, but I also wanted to get it done without spending.”

Ms Jennifer Allan, 35, did not realise how much credit card debt she was racking up. As a real estate agent in Clayton, Delaware, with a newborn daughter and no paid maternity leave, she relied on credit cards for everyday expenses like diapers and groceries.

“I woke up one day, and I was like, ‘If I use ChatGPT for everything else, why wouldn’t I just use it to help me pay down this debt?’” Ms Allan said.

Her first prompt: “I’m in credit card debt. I have no idea how much credit card debt I’m in. I don’t know what to do.”

She owed US$23,000 and decided to undertake a 30-day challenge to help her stay focused and drive down her debt. Each day, within the same chat, she asked ChatGPT for a new task to help her raise money, and she did each one.

She sold a fresh watermelon with her debt total tattooed on it for US$51, recovered US$700 from her state’s unclaimed-property registry and donated her plasma for US$80, despite her fear of needles. She also saved nearly US$600 by turning leftover pantry items into meals, she said.

Ms Allan, who documented the daily challenge on TikTok, said she succeeded in paying off nearly half her credit card debt.

ChatGPT also suggested that she contact her credit card companies in an attempt to lower her balance. The companies offered to close the cards, but she declined.

People looking for financial advice feel most comfortable asking chatbots about budgeting and tracking expenses, according to a survey by Empower, a financial services company.

For Ms Kathryn Aguilo, a 30-year-old kindergarten teacher, that is exactly how she has used it.

She first used ChatGPT in 2024 to help her and her fiance save for their 180-person wedding on Long Island, New York. The chatbot suggested they cut back on eating out, set a US$40 limit when they did, and stop opening bar tabs. It also suggested that they bring cash instead of using cards.

After the wedding, Ms Aguilo turned to ChatGPT to build an expense tracker and budget so she and her husband could start saving for a down payment on a home. They closed on a two-bedroom house in January, and Ms Aguilo again went to ChatGPT for advice on how to pay down her 30-year mortgage faster.

“I’m not, like, asking it to create my future children and what they’re going to look like,” Ms Aguilo said. “I do really just use it intentionally for saving, budgeting and to help us.”

While many use AI just to break even, some use it to get ahead. Fresh off a break-up and in the same accounting job for six years, Mr Alexander Stuart, 32, wanted a change.

In late June, Mr Stuart asked ChatGPT to act as a “free college” to teach him about stock investing and how to “become one of the greatest traders”. He had US$400 to put into the market, and the chatbot helped him plan trading strategies, like how to manage risk and choose when to buy and sell.

Then he put the guidance to the test. His first trade was on chipmaker AMD, after the chatbot said it was the best investment out of 500 companies based on mergers, analysts’ notes, trading activity and more. He took the advice, he said, and his investment doubled the same day.

Now Mr Stuart’s account on Robinhood, a popular trading app, stands at around US$1,600 – a big return on investment, though he realises the gains could disappear just as quickly, and he said he is investing for entertainment and educational purposes, not to pay his bills.

Despite the success some people have had after asking AI to fix their finances, others say they have been led astray.

Take Mr Stuart. About a month into investing, he lost nearly US$60 on an Nvidia trade after realising that one of the numbers ChatGPT cited as new was days old, something financial experts warn about when relying on AI chatbots. Since then, Mr Stuart has started manually uploading and verifying the data on his own before letting the chatbots make his investing decisions. He is also comparing ChatGPT’s advice with suggestions offered by Grok, another chatbot.

Uploading information poses some risks, too, though. Ms Allan, who completed the 30-day challenge, uploaded her bank statements to get personalised AI advice, but some advisers recommend that users keep personal details and sensitive information, like a Social Security number, away from the chatbot. NYTIMES