OpenAI said on Monday it has begun testing ads in ChatGPT, one day after being lampooned for its chatbot ad plans in rival Anthropic’s Super Bowl commercial.
The test is occurring in the US for logged-in adult users with Free accounts or the new ad-supported “Go” subscription tier; OpenAI has spared customers who pay for Plus, Pro, Business, Enterprise, and Edu accounts from seeing ads.
When word surfaced about Anthropic’s comedic take on ad-supported chatbots, OpenAI CEO Sam Altman said the ads were dishonest, and tried to paint Anthropic – which has disavowed advertising – as elitist.
“Anthropic serves an expensive product to rich people,” said the aggrieved billionaire in a social media post last week. “We are glad they do that and we are doing that too, but we also feel strongly that we need to bring AI to billions of people who can’t pay for subscriptions.”
OpenAI must pay staff and find a way to deliver returns for investors who have sunk more than $60 billion into the money-losing venture. So with positive cash flow not expected until 2030 – assuming the company survives that long – ad revenue looks like life support
It’s not the only AI company eyeing ad revenue opportunities. Google is reportedly planning to bring ads to its Gemini services later this year. But advertising and AI remain an unproven flavor combination: Perplexity, the AI search startup, paused its advertising operation last October following the departure of ads chief Taz Patel.
OpenAI aims to demonstrate that it can serve ads while maintaining customer trust. Echoing privacy commitments from the likes of Google and Meta, the company says, “we keep your conversations with ChatGPT private from advertisers.”
Specifically, the AI outfit insists that chats, chat history, memories, name, email, precise location, IP address, and sensitive information (e.g. health, politics) are never shared with advertisers.
Nonetheless, the biz has enabled ad personalization as a default in the settings available to ChatGPT users. So those who leave this setting enabled can expect their recent past and present conversations will be fed to systems that choose and serve ads based on what’s discussed.
“Starting in February, if ads personalization is turned on, ads will be personalized based on your chats and any context ChatGPT uses to respond to you,” the support document explains. “If memory is on, ChatGPT may save and use memories and reference recent chats when selecting an ad.”
Disabling ad personalization means only the current conversation will inform the ad targeting – past chatter will be ignored.
This data will not affect the answers ChatGPT provides, OpenAI insists: “Ads do not influence the answers ChatGPT gives you. Answers are optimized based on what’s most helpful to you.”
OpenAI explains that during the test period, the AI company will decide what ads its customers see by matching the topics discussed in current and past conversations with the relevant ads. “For example, if you’re researching recipes, you may see ads for meal kits or grocery delivery,” the company said.
Other signals may also be used for ad selection, like a person’s general location, language, and ad interaction history.
OpenAI insists that its ads, which will appear at the end of ChatGPT’s response to a prompt, are clearly labelled and visually separated from the chatbot content.
The biz similarly separates itself from the marketing messages it presents, stating that it does not endorse or recommend advertisers, or their products and/or services. That’s a level of hesitancy matched by its disclaimer: “ChatGPT can make mistakes. Check important info.”
OpenAI won’t show ads in temporary chats, when users are logged out, after generating an image, and in the ChatGPT Atlas browser. The ad exemption in Atlas may represent a bid to encourage more usage of OpenAI’s browser. It also appears to be an acknowledgement that ad blocking extensions could be used to remove ads if they were allowed in Atlas.
ChatGPT users discussing sensitive or regulated topics, including health, mental health, and politics, will also not see ads.
Advertisers will face restrictions that mean they cannot show spots related to dating, health, financial services, or politics.
But things may change. OpenAI says that its initial advertising test is focused on learning and that its advertising program will evolve over time – which may not be long if current spending fails to generate the necessary returns.
Caveat user. There’s plenty of industry precedent for starting with a product that customers love, then slowly increasing the ad load until it resembles something completely different – and much worse. For a refresher, check out what Google looked like when it started.®