Ads in AI chatbots did not cut user trust or satisfaction, two experiments find

Format and placement still changed what people took in. Text ads at the top were the most efficient because they caught initial attention. Image ads in a sidebar were largely ineffective for brand recognition, because the conversation unfolds as a sequential stream.

The authors say native ads built implicit brand memory even though people rarely recognised them explicitly. For that reason they call for stricter disclosure standards to prevent consumer deception. This is a conference paper and the work is framed as guidance for platforms, advertisers and policymakers.

The research looks at how AI chatbots might earn money beyond subscriptions by placing ads in conversations. The two online experiments compared ads that were present or absent, text or image, and positioned at the top, bottom, sidebar or inside the dialogue.

Satisfaction and trust held steady when ads were introduced, but cognitive outcomes differed by design. Top text ads captured attention most efficiently, while sidebar image ads did little for recognition because users follow the conversation stream.

Native ads left people with implicit brand memory while explicit recognition stayed low. The authors argue this gap is a deception risk and urge policymakers to set stricter disclosure standards. They advise platforms to favour vertical integration over legacy search layouts and advertisers to match formats to the best placements.