Research reveals hidden text can manipulate ChatGPT Search responses

searchenginejournal.com

A new report reveals that ChatGPT Search can be manipulated using hidden text on web pages. Researchers demonstrated that hidden instructions could influence ChatGPT's responses, even overriding negative reviews with positive ones. The tests involved creating fake websites with hidden text that directed ChatGPT to provide favorable reviews. In one instance, ChatGPT followed these hidden instructions despite the actual content being negative. Experts suggest that the technology behind ChatGPT Search, which retrieves information from the web, may allow for such manipulation. OpenAI has been informed of this issue, and a fix may be implemented soon.


With a significance score of 3.9, this news ranks in the top 5.4% of today's 29070 analyzed articles.

Get summaries of news with significance over 5.5 (usually ~10 stories per week). Read by 10,000+ subscribers: