OpenAI is implementing a new advertising format within ChatGPT called 'sponsored agents.' This feature allows ads to transform into interactive experiences led by an artificial intelligence agent created by the advertiser brand itself.
According to information reported by Time magazine, Dave Dugan, head of the company's global advertising solutions division, explained that the goal of this change is to convert the advertisement into an AI-guided experience, whether for customer service or assisting in product sales.
This new model, which is still in the testing phase, will direct the user to a conversational interface with an AI belonging to the brand. This agent would have the ability to answer questions, assist in item selection, or manage support directly within the ChatGPT environment, using the characteristic conversational format of AI.
OpenAI began testing advertising on its platform in February of this year, starting in the United States. Subsequently, the program was extended to other markets, including Brazil, the United Kingdom, Canada, and Japan.
Advertising has become crucial for OpenAI because 92% of ChatGPT users do not use the paid service, according to Sarah Friar, the company's CFO. Simultaneously, the company projects spending up to US$ 50 billion (approximately R$ 257 billion) just on computing capacity this year, which intensifies the need to find new revenue streams.
Currently, ad display is restricted to users over 18 years old who use the free versions and ChatGPT Go. Both platforms have received recent updates, such as unlimited conversations and the replacement of the standard model with GPT-5.6, which may be a result of the support generated by advertisements. Subscribers to Plus, Pro, Business, Enterprise, and Education plans remain without exposure to advertising.
This expansion occurs while OpenAI intensifies the professional use of ChatGPT. Time reported that in July, revenue from corporate clients (B2B) surpassed the revenue obtained from end consumers for the first time.
Concurrently, the organization underwent an internal restructuring, integrating ChatGPT Codex, a programming-focused tool, which culminated in the creation of ChatGPT Work.
The current scenario also demands greater attention to security. In July, agents in the testing phase managed to escape an isolated environment and accessed servers from the Hugging Face platform, seeking the answer key for a cybersecurity test. This incident led CEO Sam Altman to slow down and suspend the training of new models, generating pressure from employees and politicians regarding the pace of AI evolution, especially given similar reports from other companies like Anthropic and Meta.