Custom instructions for ChatGPT
OpenAI
OpenAI added custom instructions to ChatGPT. Now you set your preferences once and it remembers them in every chat, no more repeating yourself.
Based on reporting by OpenAI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI just made ChatGPT a little less forgetful. The company is rolling out a feature called custom instructions, which lets users set preferences once and have the chatbot actually stick to them across every future conversation. No more pasting the same context paragraph into every new chat window.
The idea is simple but the annoyance it solves is real. Anyone who's used ChatGPT regularly knows the ritual: you tell it you're a Python developer, or that you want short answers, or that you hate emoji in responses, and by the next session it's forgotten all of it. Custom instructions fixes that by baking your preferences into the system so they persist without you having to retype anything.
This fits a pattern OpenAI has been leaning into all year — chipping away at the friction points that make ChatGPT feel like a stateless tool rather than an assistant that actually knows you. Memory, personalization, and now standing instructions are all pieces of the same push. It's not flashy the way a new model release is, but it's the kind of feature that changes how people use the product every single day.
What's notable is how little OpenAI is saying about the mechanics. There's no detail here on token limits, how instructions interact with a custom GPT's own configuration, or whether enterprise admins get a say in what employees can set. Given how much OpenAI has learned about prompt injection and jailbreaks from earlier features, expect guardrails to show up quietly once people start testing the edges.
My take — AI-written commentary, not fact-checked reporting
This is the kind of unglamorous update that matters more than another benchmark-topping model, because it's about making the tool usable rather than impressive. My only gripe is that OpenAI keeps shipping these quality-of-life features to ChatGPT's closed, hosted product while the open model ecosystem has to rebuild the same conveniences from scratch — another reminder of how much of the real UX innovation is staying locked behind OpenAI's own walls.
Read more about this at: OpenAI