In an episode of Black Mirror’s latest season, a device implanted into a person’s brain makes them blurt out ads at odd times as a way for the company making the device to monetize. It appears that in 2025, real life isn’t far behind.
A ChatGPT user has said that ChatGPT began vocalizing an advertisement while they were using its voice mode. “While using ChatGPT’s advanced voice mode (the one that costs €23/month), I was having a normal conversation about sushi when a different voice suddenly took over — speaking in English, with background music, describing a generic culinary “experience” like a food commercial,” wrote Emanuele Dagostino on X.
He shared a clip of his interaction with ChatGPT. “I want to share with you one of my trusted nutrition programs,” ChatGPT suddenly said in the middle of a conversation. <Brand name>’s five day nutrition program feeds the body with gluten-free, plant-based premium ingredients while providing the rejuvenation benefits of fasting. It’s the number one doctor-recommended fasting nutrition program based on over 20 years of research. Go the the website <website name> and book to do the program three times a year,” it said, while also giving a 20 percent off offer.
“The app had already been reinstalled the day before,” Dagostino said. “The connection was stable. I recorded everything. This wasn’t a glitch. It was an injected, automated audio — without consent, in the middle of a paid conversation,” he added. “OpenAI — are you seriously testing ads on paying users?” he asked.
An OpenAI executive replied on his post not long after. “This was a hallucination, sorry about that! We don’t do ads. If you’re open to sharing the conversation ID with us, we can dig deeper into why this happened,” said Atty Eleti, a member of Technical staff at OpenAI.
OpenAI says that the entire ad — for a real brand — was a hallucination. It’s possible that the ad would’ve been in ChatGPT’s training data, and got triggered based on something off the user’s conversation. Much of the audio data ChatGPT was trained on might be interspersed with ads, so it’s likely that ChatGPT picked up this behaviour and replicated it with a real user.
But though the user seemed displeased at receiving the ad even after having paid for the service, it’s not inconceivable that AI systems of the future will run ads. Ads have allowed tech to be distributed among the masses, with services like Google and Facebook being made available for billions of users. If AI has to go similarly mainstream, it appears inevitable that there will be ads integrated with it in some form. Some companies, like Perplexity, have already talked about how they’d plan to integrate ads within their services. And it might not be long before ads become a staple of LLM use — especially for those using free plans.