Short answer: Pulse is ChatGPT working while you sleep. Instead of waiting for you to ask something, it does research overnight and hands you a small set of personalised update cards the next morning, built from your memory, your chat history, your feedback, and optionally your calendar and email. It launched as a preview for Pro users on mobile, with a stated plan to widen access. It is the clearest example so far of assistants moving from answering questions to starting them, and whether it earns a place in your morning depends almost entirely on how much effort you put into curating it.
How Pulse Actually Works
The mechanic is simple and worth understanding, because it explains both the appeal and the failure modes.
Each night, ChatGPT synthesises what it knows about you from three sources: your memory, your chat history, and the explicit feedback you have given it. From that, it decides what would be useful to you tomorrow and does the research asynchronously. In the morning you get a handful of cards, one topic each, quick to skim and expandable if something looks worth reading properly.
Two design decisions shape the experience more than the technology does.
It is once a day, not a feed. A card lasts the day and then goes, unless you keep it by saving it as a chat or by asking a follow-up, which files it with your normal conversations. There is no infinite scroll, no backlog to clear, and nothing accumulating guilt while you ignore it.
It is curated by you, not inferred at you. You can tell it explicitly what to research, in the same natural language you use for anything else. Something like “give me a Friday roundup of local events” or “focus on professional tennis this week” changes what appears the next day.
The Connected-Apps Question
The part that materially changes what Pulse can do, and the part that deserves the most thought, is connecting Gmail and Google Calendar.
With calendar context, the updates change character. Instead of general interest topics, you get a draft agenda for a meeting you have tomorrow, a reminder about a birthday, or restaurant options near an appointment in another city. That is genuinely a different product from a news digest.
Both connections start switched off, and settings let you change that at any point, which is the correct design. The trade-off is straightforward and worth naming plainly: the feature gets more useful in direct proportion to how much of your life it can see, and everyone draws that line in a different place.
A reasonable middle position is calendar only. Calendar entries are mostly logistics, which is where the assistant adds the most, while email contains the material most people are least comfortable making available to a background process.
Pulse Compared With the Alternatives
Proactive briefing is no longer one product’s idea, so it helps to see where Pulse sits.
| Approach | What it does | Best for |
|---|---|---|
| ChatGPT Pulse | Overnight research from your memory, chats and connected apps, delivered as daily cards | People with an established history in one assistant |
| Gemini Daily Brief | A briefing feature listed among paid Gemini plan inclusions | People already inside Google Workspace |
| Scheduled tasks | A specific query you define, run on a schedule you choose | People who know exactly what they want tracked |
| A newsletter | Editorially curated, identical for everyone | Breadth and editorial judgement |
| Nothing | You ask when you need something | Most people, honestly |
The comparison that matters most is Pulse against scheduled tasks, because they solve overlapping problems in opposite ways. A scheduled task is precise and dumb: it runs the query you wrote, every time, and you get exactly what you asked for. Pulse is imprecise and adaptive: it decides what is relevant, which is more useful when you do not know what to ask and less useful when you do.
If there is one specific thing you want tracked, a scheduled task will beat Pulse at it every time and takes two minutes to set up. Pulse earns its place on the things you would never have thought to ask about, which is a real category and also a harder one to evaluate.
What to Know Before You Rely on It
Availability has moved repeatedly. It launched as a preview for Pro users on mobile, with an explicit plan to learn from early use before rolling out more widely. Anyone reading this later should confirm current availability and plan requirements with OpenAI directly rather than trusting a date in an article.
It is a preview, and it says so. OpenAI’s own description is candid that it will not always be right, and gives the example of advice about a project you have already wrapped up.
It only knows what you have told ChatGPT. If your memory is sparse or your chat history is narrow, the output will be generic. This is the most common reason people try it, find it underwhelming, and switch off.
Curation is the whole product. OpenAI reported that in early testing with students, many only started finding it useful once they told ChatGPT what they wanted to see. That is the single most important practical fact about this feature.
Competitors have similar features. Google lists a Daily Brief among the features its paid Gemini plans unlock, so proactive briefing is becoming a category rather than one product’s differentiator. If you already pay for one assistant, check what it offers before subscribing to another for this.
Getting Something Useful Out of It
The gap between people who find Pulse valuable and people who find it noise is almost entirely down to the first week.
- Spend the first three days curating aggressively. Thumbs down anything irrelevant and say why. Ask directly for what you actually want. The feedback compounds, and skipping this is the main reason first impressions are poor.
- Ask for specifics, not categories. “Industry news” produces a generic digest. “Changes to procurement rules affecting UK public sector software contracts” produces something you cannot get elsewhere.
- Connect calendar before email. It is the higher-value, lower-exposure integration by a wide margin.
- Save the good ones. Cards expire nightly. Keeping one means saving it as a chat or asking a follow-up, so anything you might want next week needs an action from you today.
- Set a review point. After two weeks, decide honestly whether you are reading it or scrolling past it. Features like this get tolerated for months without ever being useful.
- Do not use it as a news source. It is a personalised research digest built on your own context, not an edited publication with editorial standards.
A Decision Framework: Is This For You
Work through these four questions rather than trying it and hoping.
- Do you already use ChatGPT enough for it to know you? The output quality is a direct function of your history and memory. Light users get generic cards and conclude the feature is weak, when the real problem is thin input.
- Do you have recurring, trackable interests? Ongoing projects, a subject you follow, a goal with steps. Pulse rewards continuity and has little to offer someone whose usage is a series of unrelated one-off questions.
- Are you comfortable connecting a calendar? Without it, most of the genuinely differentiated use cases disappear. If the answer is a firm no, expect a fraction of the value.
- Will you actually curate it? If the honest answer is that you will look at it twice and never give feedback, it will not work for you, and that is a reasonable thing to know in advance.
Three or four yes answers means it is worth two weeks of real use. Fewer means you will get more from a scheduled task that runs one specific query you actually care about.
Common mistakes right now
- Judging it on day one, before any feedback has shaped it.
- Requesting broad topics and receiving broad, useless summaries.
- Expecting breaking news rather than personalised research.
- Losing a useful update because it was not saved before the day ended.
- Connecting everything immediately without deciding what you are comfortable exposing.
Where Proactive Assistants Are Heading
Pulse matters less as a feature than as a signal. Every major assistant is moving from a model where you ask and it answers to one where it works in the background and presents results.
That shift changes the skill involved. With a chat interface, the skill is asking well. With a proactive system, the skill is specifying standing interests precisely, giving feedback that actually redirects behaviour, and judging output you did not request, which means you have no expectation to check it against.
That last part is the underrated risk. An answer you asked for gets scrutinised. An update that appears unbidden tends to be accepted, and that is exactly when a plausible but wrong claim slips through. Building the habit of checking anything you would act on is worth more than any configuration setting.
Getting good at directing these systems, and at knowing where they fail quietly, is a learnable skill rather than a product feature, and it transfers across every assistant rather than being specific to one. Learning it in a structured sequence beats accumulating habits by trial and error. If you want a structured route in, explore Coursiv AI lessons and check current plan details on the official site.
FAQ
What is ChatGPT Pulse?
Who can use it?
Does it read my email?
Can I keep an update I liked?
Why are my updates generic?
Your Next Step
If you have access, spend one week treating curation as the actual task rather than as an optional extra. Every morning, thumbs down what missed and say in one sentence what you wanted instead, and once a week ask directly for a specific, narrow thing you genuinely care about. The people who report this feature as useful and the people who report it as noise are mostly the same people separated by that habit.