On Friday OpenAI announced that it was going to test ads in ChatGPT in a blog post that struck the tone of a plea in mitigation.
For now, the ads are only coming to its free product and low-cost subscription tier in the US, it says. And the answers that ChatGPT provides will never be influenced by ads, which will always be discrete and clearly labelled — and quite large, if the mock-up image (see above) is accurate.
More interesting is whether AI answer engine ads will be contextually relevant (ie, for products that pertain to the user’s query) or just tailored towards the user’s behavioural profile, like they are on social media.
Most people assume it will be the former because that’s how advertising works within regular search engines.
Eric Seufert, however, has some good arguments for why personalised ads would be the better option — not least because it allows OpenAI to monetise a lot more of its searches on ChatGPT.
In 2020, Google said that 80% of its search engine responses don’t carry ads at the top of the results page. This is because the queries demonstrate no commercial intent, and including irrelevant links diminishes the utility of traditional search engine responses.
But AI answer agents may not suffer the same handicap because they provide information, not just links to relevant information elsewhere. As a result, ChatGPT’s responses may be viewed by users more like content than traditional search results.
With personalised — as opposed to contextual — ads, OpenAI could also monetise queries about personal or sensitive issues without seeming predatory. Marc Zao-Sanders wrote in 2025 that his research showed therapy and companionship had become the number-one use case for LLMs*, so it’s no small thing for OpenAI to opt out of selling ads against these kinds of queries.
Of course, personalised ads can come across as invasive for different reasons, as people feel like they’re being surveilled around the internet. And, for now at least, it appears as though OpenAI is putting its faith in contextually relevant ads.
‘To start, we plan to test ads at the bottom of answers in ChatGPT when there’s a relevant sponsored product or service based on your current conversation,’ it wrote in its blog piece.
* While therapy and companionship emerged as the number one use case in Zao-Sanders’ research, it still only accounted for 4% of overall use. Also, Zao-Sanders got his data by monitoring people’s conversations about AI online, mostly within Reddit forums, which might skew his findings.
