If playing a game could lower your risk of cognitive decline, would you do it?

That’s the claim that Lumosity, an online “brain training” company based in San Francisco, has used to persuade more than 70 million customers—a figure advertised on the company’s website—to engage with its supposedly neuro-enhancing puzzles. But the evidence for that assertion has been lacking, and now the company has agreed to pay up for peddling what the Federal Trade Commission has described as unfounded claims and deceptive advertising.

Lumos Labs, Lumosity’s parent company, has struck a $2 million settlement with the FTC. The company has also become subject to a more severe penalty of $50 million, though that sanction has been suspended due to Lumos’s “financial condition.” The terms of the deal further require the company to notify users of this development as well as to offer them a simple way to unsubscribe from the service.

SIGN UP: Get Data Sheet, Fortune’s daily newsletter about the business of technology.

Lumosity’s marketing campaigns—which have appeared everywhere from CNN to Sirius XM siri radio and have employed Google goog AdWords alongside keywords like “Alzheimer’s disease” and “dementia”—have advertised that players of its games benefit from improved performance on everyday tasks and protect themselves from mental impairment.

“Lumosity preyed on consumers’ fears about age-related cognitive decline, suggesting their games could stave off memory loss, dementia, and even Alzheimer’s disease,” said Jessica Rich, director of the FTC’s Bureau of Consumer Protection, in a statement. “But Lumosity simply did not have the science to back up its ads.”

WATCH: For more on games see our Fortune video.

Lumosity acknowledged the settlement in an email to Fortune. “Neither the action nor the settlement pertains to the rigor of our research or the quality of the products—it is a reflection of marketing language that has been discontinued,” a spokesperson said, citing a Lumosity-funded study in the open access scientific journal PLOS ONE purporting to show that users experienced memory and reasoning improvements. (Other studies, such as this 2010 paper published in the scientific journal Nature, have reached different conclusions.)

“Going forward, a key focus of our ongoing research is to build on these studies to better understand how training-driven improvements on tests of cognition translate to performance in participants’ everyday lives,” the company said.