Budgeting Apps, AI Labels and the 17 Million Dollar Cleo Settlement

How CodoraTech is funded: CodoraTech is supported by advertising and, in some articles, by affiliate links. Where an article contains affiliate links, we say so at the top of that article.

The personal budgeting market advertises artificial intelligence heavily, but the function underneath is usually narrower than the marketing suggests: automatic transaction categorisation and pattern matching. The clearest documented harm in the category so far came from neither a bad forecast nor a bad recommendation. It came from subscription mechanics. On 27 March 2025 the Federal Trade Commission filed a complaint against Cleo AI, Inc. in the U.S. District Court for the Southern District of New York, and a settlement order worth 17 million dollars for consumer refunds was executed the same day.

Prices and penalties on the record

• YNAB: 14.99 dollars a month or 109 dollars a year, with a 34-day trial that needs no card when signing up direct
• YNAB’s pricing page describes no AI features; the product is a rules-based zero-based budgeting method
• Copilot Money: 13 dollars a month or 95 dollars a year, with a one-month free trial
• FTC v. Cleo AI, Inc., filed and settled 27 March 2025: 17 million dollars for consumer refunds, plus a 10-year consent order
• CFPB, June 2023: roughly 37 percent of the U.S. population, about 98 million people, used a bank chatbot in 2022

What these apps mean when they say AI

Copilot Money states its AI function in one sentence on its own marketing page: “Our AI learns your spending patterns and tags every transaction automatically. The more you use it, the smarter it gets.” That is classification. It is genuinely valuable, because miscategorised transactions are among the fastest ways to abandon a budget, but it is not financial advice, and the distance between those two things is where most of this category’s marketing sits. Copilot also promotes a conversational money assistant and personalised recommendations spanning spending, budgets, investments and net worth.

Copilot names no underlying model and no AI vendor anywhere on that page. For a product whose core function requires a complete transaction history, that omission is worth registering before signing up. Pricing is 13 dollars a month or 95 dollars a year, which works out at 7.92 dollars monthly, after a one-month trial.

YNAB is the useful counter-example. It is frequently filed under AI budgeting tools by people writing about the sector, yet its pricing page describes no AI features whatsoever. The method is rules-based and zero-based: every dollar receives an assignment. Calling it an AI product misdescribes what a buyer is paying 14.99 dollars a month, or 109 dollars a year, to receive.

Cleo AI, the FTC and the Restore Online Shoppers’ Confidence Act

The Commission brought three charges, in its own framing: that Cleo deceived consumers “about how much money they could get”; that it deceived them about “how fast that money could be available”; and that it “made it difficult for consumers to cancel Cleo’s subscription service.” The settlement figure was 17 million dollars, directed to refunds.

The specifics, as set out in coverage by the National Law Review, are concrete. Cleo advertised cash advances of up to 250 dollars, while consumers discovered their far lower actual eligible amount only after they had subscribed. It promoted money as available “today” or “instantly” while charging an undisclosed 3.99 dollar express fee for same-day delivery that frequently did not arrive until the following day. Consumers carrying an outstanding advance were prevented from cancelling their subscription inside the app.

The statutes engaged were Section 5 of the FTC Act and the Restore Online Shoppers’ Confidence Act, the latter for failure to “clearly and conspicuously disclose all material terms” before collecting billing information. The resulting order runs ten years, bars misrepresentation of advance amounts, fees and cancellation terms, and requires a cancellation mechanism matching the method used to sign up. The Commission voted 2-0 to approve it. Nothing in the action turns on the quality of the underlying model, which reframes the risk for anyone shopping in this category: the plausible failure mode is billing, not arithmetic.

600 dollars in the first month, according to YNAB

YNAB states that new users save 600 dollars in their first month and more than 6,000 dollars after one year, and that 91 percent of users say the product positively changed their financial perspective. These are the company’s own marketing claims, published without a methodology. There is no disclosed sample, no control group and no definition of what counts as a saving against what baseline.

That distinction is easy to see when the two kinds of number sit side by side. The prices are verifiable: 109 dollars annually works out at 9.08 dollars a month, and YNAB says the annual plan saves 70 dollars against paying monthly. The trial terms are verifiable too, including a free 365-day trial for college students who provide proof of enrolment and a 10 percent lifetime annual discount for legacy YNAB 4 users. The outcome claim is not verifiable by a reader in the same way, and should be read as advertising rather than as evidence.

Chatbots reached every large U.S. bank before the guidance did

The Consumer Financial Protection Bureau published an issue spotlight titled “Chatbots in Consumer Finance” in June 2023, and its scale figures explain why the regulator moved. Roughly 37 percent of the U.S. population, around 98 million people, engaged with a bank chatbot during 2022, a figure the report projected would reach 110.9 million users by 2026. All ten of the largest U.S. commercial banks deploy them. The bureau put annual cost savings at approximately 8 billion dollars, around 0.70 dollars per interaction, and noted that Bank of America’s Erica had passed one billion interactions by October 2022.

The bureau set out three risks in its own words: a “Risk of noncompliance with federal consumer financial laws”; a “Risk of diminished customer service and trust when chatbots reduce access to individualized human support agents”; and a “Risk of harming people” that matters most where “a person’s financial stability is at risk.” Legal commentators read the spotlight as a signal that the bureau intended to police chatbot deployment in banking actively.

One caveat belongs with that document. It also carries figures stating that 80 percent of consumers who interacted with chatbots were left frustrated and that 78 percent needed to reach a human afterwards. Those derive from third-party survey work cited inside the report, not from data the bureau collected, and they should be described as survey data cited in the spotlight rather than as regulatory findings.

Before a free trial converts

1. Find the price on the company’s own pricing page rather than in a roundup, and note whether monthly and annual figures are both shown.
2. Check that the cancellation route matches the signup route, and confirm no product state, such as an outstanding advance, can block it.
3. Read what the product says its AI does. Transaction tagging, conversational chat and recommendations are three different things carrying three different risks.
4. For any savings claim, look for who measured it, over what period and against what baseline. Absent all three, treat it as marketing.
5. Diary the trial end date on the day you sign up, since the trial length is the one term the seller controls entirely.

Sources: Federal Trade Commission · National Law Review · Consumer Financial Protection Bureau · YNAB · Copilot Money