ChatGPT Pulse is essentially ChatGPT on autopilot: instead of waiting for you to ask something, it quietly works in the background once a day, learns what you care about from your past chats and settings, then drops a personalized briefing into your account the next morning as a set of swipeable visual cards. Think of it as a daily AI digest that remembers your projects, hobbies, and upcoming plans, and nudges you with useful ideas and links before you even think to ask.​

At its core, Pulse is a proactive layer on top of regular ChatGPT. Normally, you open the app, type a question, and get an answer; with Pulse, the flow is flipped. Each night, the system does asynchronous research on your behalf, using your chat history, saved memories, and feedback signals to decide what might actually help you tomorrow, then packages that into short cards you can skim at a glance or expand into full conversations. For example, if you’ve been talking about learning Python, planning a trip to Japan, and trying to eat healthier, your morning Pulse might surface a couple of quick practice exercises, a visa or transit reminder, and a few dinner ideas tailored to your schedule and preferences.​

Right now, Pulse is a Pro-only feature and lives on the web, iOS, and Android apps, not the desktop apps, and it relies heavily on ChatGPT’s Memory system being turned on. Memory here is the long-term layer that stores things you’ve explicitly saved or that the model has inferred over time—your preferences, ongoing projects, favorite topics—so Pulse can use that context to decide what’s relevant without you having to restate it every day. If Memory is off, Pulse can’t really personalize anything meaningful, which is why OpenAI requires it for this feature, but still lets you switch both Memory and Pulse off at any time from settings if you’re not comfortable.​

The “cards” are where the experience feels different from a standard chat log. Each morning, Pulse refreshes with a small set of visual summaries: compact tiles you can scan in seconds, each focused on one idea—maybe a new article to read, a project next step, a set of to‑dos, or a quick explanation of something you’ve been wrestling with. Tap a card and it opens up into more detail, where you can ask follow‑up questions, ask it to rewrite or expand something, or save that card as a full chat thread so you can return to it later. Anything you don’t save or interact with disappears after a day, so Pulse behaves more like a rolling brief than a permanent inbox.​

What shows up in those cards isn’t random. Behind the scenes, Pulse pulls from several different signals: your explicit memories and saved facts, the topics you’ve been chatting about, how you interact with previous cards (thumbs up, thumbs down, opening or ignoring), your “Curate” feedback (where you tell it what you want more or less of), optional connected apps, and a layer of current news and trends that line up with your interests. The idea is to strike a balance between “this is clearly about me” and “this is something I might not have discovered on my own but is still relevant,” rather than just spamming you with generic tech headlines or random lifestyle tips.​

The Curate feature is where you can push Pulse in a more intentional direction. Inside Pulse, you can tap Curate and literally tell it what you want tomorrow’s research to focus on—anything from “find local free events this weekend that are rain‑or‑shine” to “pull five new research papers on LLM evaluation benchmarks” or “give me a weekly training progression for my half‑marathon in April.” If you share that request before around 10 pm local time, OpenAI aims to include it in the next morning’s Pulse, though they’re upfront that timing can vary. Over time, your curate prompts plus your feedback history become another strong signal that shapes what Pulse prioritizes and what it quietly drops.​

Crucially, Pulse also gives you a fair amount of control. You can give quick thumbs up or down on individual cards, open the three‑dot menu to leave detailed feedback, or report anything that looks off or potentially harmful under OpenAI’s Terms of Use. There’s a dedicated feedback history section where you can review and delete the input you’ve given Pulse if it no longer reflects what you care about, which matters because those signals are part of how it personalizes your daily brief. And if the whole thing becomes distracting, you can hide Pulse from the main chat view while still leaving it accessible from the sidebar, or turn off its use of memories in proactive suggestions entirely.

Where things get more interesting is when you connect external apps. Pulse is being built to work with connectors, starting with Gmail and Google Calendar, which are opt‑in and off by default. When you toggle on “Allow proactive activity” for those apps, Pulse can scan relevant emails and calendar entries in the background to surface genuinely useful suggestions—like restaurant ideas near the hotel for an upcoming work trip, reminders to prepare for a recurring weekly meeting, or a quick draft agenda for a call you have later in the day. OpenAI says content from Gmail and Calendar used this way won’t be used to train the model, and you can disconnect those integrations at any point.​

If you zoom out, Pulse is part of a broader shift from reactive chatbots to proactive AI agents. Instead of being a fancy search box, ChatGPT becomes a kind of ambient executive assistant: it remembers that you’re job hunting, studying for an exam, managing a side hustle, or trying to get in shape, and then uses the quiet hours of the night to line up ideas, resources, and reminders so you wake up with a head start. For busy people, that could look like a curated morning brief across work, personal life, and learning—one place where industry news, project nudges, travel details, and personal goals are stitched together by an AI that actually knows your context.​

There are trade‑offs and open questions, of course. A system that proactively reads your chats, optionally dips into your email and calendar, and then proposes actions raises the usual privacy and autonomy concerns, even if the knobs to turn things off are clearly there. At an industry level, though, Pulse is widely framed as a glimpse of where assistants are heading: not just answering your questions, but deciding which questions are worth asking you in the first place, and doing that in a way that feels helpful rather than intrusive.​