ChatGPT can now set reminders and scheduled tasks

For over two years, using ChatGPT has been a strictly one-way street: you show up, ask a question, get an answer, and close the tab. If you don't initiate, the AI stays completely silent.

That dynamic is finally shifting. OpenAI has begun rolling out a new beta feature called Tasks to ChatGPT Plus, Team, and Pro subscribers globally, giving the chatbot the ability to handle timed reminders and recurring requests.

Moving past reactive chat

To access the feature, users need to pick a dedicated model from the dropdown menu called "4o with scheduled tasks." From there, you set up actions using everyday conversational prompts.

You can set simple one-off reminders, like telling the bot to notify you six months before your passport expires. When the time comes, ChatGPT fires off a push notification across mobile, desktop, or web.

Where it outpaces traditional assistants like Siri or Alexa is recurring synthesis. You can ask it to generate a personalized weekend itinerary every Friday based on your location and weather forecasts, or compile an executive news briefing every morning at 7 a.m. During normal chats, ChatGPT can also suggest relevant tasks on its own, though it requires your explicit sign-off before scheduling anything.

OpenAI added a dedicated tasks manager tab to the web app, allowing users to track, edit, or delete active automations, with a limit of 10 concurrent tasks during the beta.

Guardrails, limits, and the road to Operator

Despite the agentic vibe, OpenAI is keeping this beta on a very short leash.

ChatGPT can browse the web on a schedule, such as checking once a month whether a concert tour announced local dates. However, it cannot run continuous background scraping to catch the exact second tickets drop, nor can it pull out your credit card to buy them.

There are several other limits in place. You cannot set tasks using Advanced Voice Mode, and the creation workflow is restricted to paid subscribers, with free tier rollout and broader mobile support planned for later.

According to Bloomberg, OpenAI is separately prepping an advanced agent codenamed Operator, capable of writing code and booking travel, which could land in the coming weeks. Viewed through that lens, Tasks looks like the introductory training wheels for full autonomy.

Why it matters

In our view, Tasks may look like a basic calendar add-on, but it marks an essential behavioral pivot for ChatGPT.

For daily users, the practical payoff is immediate: it eliminates the chore of manually running identical prompts every day. If you already rely on ChatGPT for routine morning briefings or recurring research digests, setting up two or three scheduled prompts saves genuine friction right now. The 10-task cap is restrictive, but it is plenty of room to test whether an AI morning briefing actually sticks in your morning routine.

Across the industry, this is OpenAI quietly laying the plumbing for the agent era. Sam Altman has publicly claimed that 2025 will see AI agents join the workforce. But an agent cannot function without autonomy over time; it needs an internal clock to wake up, execute a workflow without human prompting, and deliver results.

OpenAI's approach here is sensible. Letting an LLM run wild with background web calls or autonomous payments is a recipe for server strain and catastrophic hallucination errors. Limiting users to fixed time triggers and pre-approved web browses lets OpenAI study real-world usage patterns while keeping safety risks contained. It also puts legacy voice assistants on notice: once ChatGPT masters reminders and contextual briefings, basic assistants have very little unique territory left to defend.