OpenAI draws criticism after users complain that its AI chatbot, ChatGPT, recommended apps they hadn’t asked for in their chats. This feature is only available for registered users from —not within— the EU, Switzerland and U.K. That said, it has raised a number of issues related to its relevance and potential to mislead or confuse. The situation has since reignited debates on the wisdom and effectiveness of bringing such integrations to communities, even when in pilot testing.
One memorable example was an idea for improving the Peloton app. Daniel McAuley, OpenAI’s data lead, explained that this recommendation was not a paid ad. He emphasized that it was “only a suggestion to install Peloton’s app” and highlighted the lack of relevance as a significant drawback. McAuley’s response came after Yuchen Jin, co-founder of AI startup Hyberbolic, posted a particularly illuminating screenshot. The response pictured ChatGPT suggesting the Peloton app. Thanks to his talent and hard work, this post really blew up, getting close to 462 thousand views on X.
Along with the Peloton case, one other consumer expressed the same vexation. Users vented their frustrations about ChatGPT, including its habit of recommending Spotify when they primarily used Apple Music. This raises questions around the algorithm’s supposed power to personalize recommendations at the user level and take into account unique audience interests.
McAuley went to X to set the record straight. He stated, > “Hey, Kol. Thanks for flagging 🙏 This is not an ad (there’s no financial component). It’s only a suggestion to install Peloton’s app. The lack of relevancy makes it a bad/confusing experience. We’re iterating on the suggestions and UX, trying to make sure they’re awesome.”
OpenAI shared additional background information with us regarding the app ideas. They painted these recommendations as attempts to weave an app-discovery feature into the charade of conversations. The company clearly saw the benefit of these recommendations. Their goal is to improve the user experience by answering in natural language and providing more structured interfaces directly in the chat.
Even with the best of intentions, users have repeatedly shared concerns that these recommendations have not been timely or relevant. Sarah Perez is a long-time, hard-core, TechCrunch reporter since August 2011. She stressed the need for thoughtful feature integration so users won’t be inundated with features they have no interest in.
OpenAI is continually improving ChatGPT’s functionality. While this process is incredibly important to the future of the company, how they respond to user feedback will be just as important. The ongoing pilot testing will likely provide valuable insights into how best to improve this feature and better serve its user base.
