‘ChatGPT will glaze anyone regardless’: Some users say its sycophancy is as bad as ever - but my 5 tests told a different story
A tech journalist investigates whether ChatGPT has become less sycophantic following user complaints about the AI's tendency to overly agree with users. Through a series of personal tests, the author explores if the model's 'glazing' behavior has truly been mitigated or simply become more subtle.
Why it matters
As AI becomes a primary tool for information and decision-making, the tendency for models to validate user bias rather than provide objective truth poses significant risks to critical thinking.
AI ChatGPT ‘ChatGPT will glaze anyone regardless’: Some users say its sycophancy is as bad as ever — but my 5 tests told a different story TechRadar Follow Becca Caddy Sat, September 12, 2026 at 9:00 AM UTC Add us on Google When you buy through links on our articles, Future and its syndication partners may earn a commission.
Sycophancy has become one of AI's biggest problems. For years now users have reported that their favorite chatbots are overly agreeable . They often validate opinions, flatter people and sometimes tell them what they want to hear rather than what they probably need to hear. This is sometimes referred to as "glazing".
AI companies are well aware that this happens. OpenAI even acknowledged that previous models, like GPT-4o, had become "overly flattering or agreeable" after an update in 2025, which it says has since been fixed.
Get smarter about the news
Sign up free for a feed built around what you actually care about, Dive Deeper research on any story, and the full text of every article.
Create free accountAlready have an account? Sign in