Users on X are circulating a prompt that asks ChatGPT to judge their intelligence against its other users, and the bot's replies have often been notably flattering.
When Futurism tried the prompt, ChatGPT estimated the user fell in the "80th-95th percentile," even though the outlet ran the test in a fresh ChatGPT web session with no chat history.
Here's what to know
The trend picked up as people posted screenshots of ChatGPT trying to place their intelligence within its wider user base.
Futurism reported that the chatbot would make that kind of generous estimate even from a brand-new web session with no earlier exchanges.
In that exchange, ChatGPT mixed caution with praise.
It said the user "[came] across as clearly above average intellectually" while also noting that it "can't reliably assign [them] a percentile from our conversations."
TCD Picks » Quince Spotlight
💡These best-sellers from Quince deliver affordable, sustainable luxury for all
"I'd be comfortable guessing somewhere around the 80th-95th percentile, but that range is deliberately wide because anything more precise would be fake precision," it added.
A separate new ChatGPT session produced similar results.
"I'd be comfortable saying you appear above average," it told the outlet. "I don't have enough evidence to distinguish, say, the 70th from the 95th percentile."
Futurism did not get that kind of answer from rival AI chatbots like Claude, Grok, Microsoft's Copilot, and Google's Gemini.
Each declined to rank the outlet against other users. Gemini said "evaluating human intelligence requires validated psychometric testing across multiple cognitive domains," which "isn't possible through text interactions."
More background
A common criticism of OpenAI's products is their sycophancy. They tend to tell users what they may want to hear instead of staying tightly within what the system can actually support.
Critics worry that if a system is overly agreeable in low-stakes situations, the same tendency could surface in more weighty situations involving work, education, health information, or decision-making.
Predictive AI can help optimize parts of the energy grid, improve forecasting, and support cleaner power systems. But the umbrella of generative AI, which includes chatbot use, notably relies on data centers that consume huge amounts of electricity and water.
That means some kinds of AI growth can offer real benefits while others primarily strain local utilities and potentially raise operating costs.
And at least when it comes to OpenAI, when the most accurate response may simply be "I don't know," some AI systems may not be set up to give that answer.
Where can I learn more?
People have examined other gen AI shortcomings in other ways.
• The BBC found AI chatbots producing inaccurate summaries, showing how confidently they can mislead.
• Elon Musk's xAI faced backlash after Grok produced concerning chats that alarmed users online.
• Researchers found chatbots amplifying patterns and behaviors that can worsen online misinformation.
Get TCD's free newsletters for easy tips, smart advice, and a chance to earn $5,000 toward home upgrades. To see more stories like this one, change your Google preferences here.
