ChatGPT, like many chatbots, is presented as a hyper-capable personal assistant. But among the many things it confuses, one is particularly confusing: it can’t tell time.
When I ask ChatGPT what time it is, I’m never sure what I’ll get. Sometimes, it tells me it can’t do it. “I don’t have access to your device’s real-time clock or your location, so I can’t tell you the exact local time,” it wrote to me about a week ago at 4:15 p.m. Eastern Standard Time. “But I to do Know today’s date according to my system: 2025-11-20(Bolded by ChatGPT, I believe, to make sure I wasn’t overlooking things it was doing well.) Sometimes it asked me to specify a city or time zone, only to tell me it couldn’t reliably check the time that way – ,It’s 12:42 AM in New York (Eastern Time, assuming your system clock is correct),, ChatGPT told me in the morning. Wrote at 11:08 pm. and sometimes it does Tells the exact time, until I ask a few minutes later and it goes wrong again.
We are not the first to bring this up. The problem of timing often comes up on Reddit and ChatGPT forums. One user urged OpenAI to “pay attention to this” because it “discredits” AI models “far superior to my cognitive abilities.” Features such as web search have offered some solutions. But even years after its launch, vanilla ChatGPT remains indifferent to the ticking of the clock – and as absurd as the situation may seem, there’s a simple reason for that.
Telling the time is a simple matter for any computer or phone, because there are small chips inside them ticking. But the big language and visual models powering generic AI systems like ChatGPT, Google’s Gemini, Anthropic’s Cloud, and others are built for a very different purpose. By default, they take user questions and predict answers based on their training data. This does not include constant, real-time updates about things like the time, unless they specifically search the Internet for that information.
“A language model works in its own space of language and words. It’s only referencing things that have entered this space,” said AI robotics expert Yervant Kulbashian, who wrote about the concept of time being understood by AI in 2024. The VergeIt is like an exiled man on an island in the middle of the ocean, who has a huge collection of books but no watch,
Why can’t OpenAI build a bridge to that island and give ChatGPT access to the system time clock? The short answer is, it can. As I chatted with Pasquale Minervini, who researches natural language processing at the School of Informatics at the University of Edinburgh, Scotland, his desktop ChatGPT app immediately gave him the correct time in Milan, Italy—where he was located during our interview. He said, “If you give it access to a watch it’s capable of telling time. Otherwise, it’s something that was kind of born in that moment.” He said the information provided at the time was likely “implicit in the context” of the app. He previously enabled the “Search” function on the ChatGPT app, meaning ChatGPT is allowed to tap into his computer’s built-in time tools, in addition to the web, to get the time.
That’s all OpenAI told us. “The models powering ChatGPT do not have built-in access to the current time, so the latest facts require ChatGPT to occasionally call a search to get the latest information,” spokeswoman Taya Christianson wrote. The Verge,
There are some problems with keeping LLMs aware of time, Kulbashian said. ChatGPT’s context window, or the portion of information “remembered” at any time, has limited space. Every time ChatGPT consults a system clock, it adds a piece of information to that context window – to use another metaphor, imagine someone keeping a clock on a desk that stops every second. “If you start adding more things to your desk, you’ll eventually have to start removing things,” Kulbashian said.
When updated frequently enough, clock clutter can amount to mere noise to an AI system. “You can confuse robots,” Kulbashian said. “If we’re having a conversation, and suddenly someone comes up and says, ‘It’s 5:45.’ ”It is now 5:46.” In contrast, it is relatively easy to include something like a date in the system prompt at the beginning of a chat – as shown by an apparent ChatGPT system prompt leak.
ChatGPT users can tell the time without any hassle by specifically asking the chatbot to find it. (Some other chatbots, like Google Gemini, will automatically search for the time.) You can also use an open-source Model Reference Protocol to connect AI applications to your data. That said, sending AI models to search the web or allowing them to access personal data comes with risks, such as malicious signals being fed into bots that are scattered across the Internet, Minervini said.
Minervini, who finds blind spots in consumer AI technology as part of her research, says there’s actually a whole list of time-related tasks it hasn’t mastered. They inspired pioneering AI models with images of analog clocks and found that the models had difficulty reading the position of the clock’s two arms. Calendars are also “weird,” he told me.
Perhaps the bigger issue for the average user is that ChatGPT can’t reliably explain what its limitations are. A human assistant who does not know the time can understand it; one he regularly Lie About knowing it will probably be fired. But, of course, the big language models aren’t lying – they’re just predicting what you want to hear, as always.
“We’re constantly improving how it knows when to do this,” said OpenAI’s Christianson.
<a href