OpenAI said on Monday, Oct. 5, that it will start testing a new kind of ad in ChatGPT: pictures of a product or service that appear while the chatbot is making an image for someone. Testing begins later this month in the US with a first group of advertisers. OpenAI says the ads will be clearly labeled and kept separate from the image being made, and that advertising does not influence the answers ChatGPT gives.
OpenAI's help pages say it does not show any ads to people on its paid Plus, Pro or Business plans, or to accounts it identifies as belonging to someone under 18. OpenAI says ChatGPT reaches 1.2 billion people each week and that ads let it serve a much larger market than subscriptions alone.
The rest of the announcement is about a question every advertiser asks: did the ad actually lead to a sale? Tracing a sale back to an ad the shopper saw or clicked is called attribution, and OpenAI now works with more companies that do it, including AppsFlyer, Triple Whale, Adjust, Northbeam and Branch. It is also working with Haus, Measured and WorkMagic on experiments that compare sales in areas where the ads ran with areas where they did not, to see how much extra business the ads created. OpenAI says that work is still in its early stages.
OpenAI also published a few early results that measurement companies found for individual brands. By DV Rockerbox's count, WeightWatchers paid 15.3% less to win each customer through ChatGPT ads than it paid on average through search ads. WorkMagic found that 67% of the extra purchases the ads brought the wellness brand Dose came from new customers. WorkMagic said in August that advertisers could already track their ChatGPT ads alongside their other ads in its tools.
OpenAI also says its safeguards are meant to keep ads out of emotionally vulnerable or sensitive conversations, and qualifying advertisers can now name phrases to keep their ads away from. DoubleVerify and Integral Ad Science, two companies that check where ads appear on behalf of advertisers, are setting up trials to judge how well those safeguards work without seeing anyone's private conversations.
OpenAI has not said how many advertisers are in the first test, which free users will be part of it, how often people will see an ad while waiting for an image, or when the format could reach other countries. It is also not yet clear whether the results of the safety trials will be made public.