OpenAI Releases ChatGPT Pulse Preview, Shifting ChatGPT From Reactive to Proactive

OpenAI has released a preview of ChatGPT Pulse to Pro users on mobile, a new experience in which ChatGPT proactively conducts research and delivers personalized updates instead of waiting for a prompt. The company announced the release today and described Pulse as a departure from the conversational model that has defined ChatGPT since its launch.

"Today we're releasing a preview of ChatGPT Pulse to Pro users on mobile," OpenAI said in its announcement. "Pulse is a new experience where ChatGPT proactively does research to deliver personalized updates based on your chats, feedback, and connected apps like your calendar."

The sentence is short, but the shift it describes is structural. Every interaction users have had with ChatGPT to date has been initiated by the user: a question goes in, an answer comes out. Pulse inverts that arrangement. The system decides what to look into, performs the research on its own, and pushes the results to the user as an update.

What OpenAI said, precisely

The announcement contains three concrete claims about how Pulse works, and each one marks a design decision worth examining.

First, Pulse draws on "your chats." The system uses conversation history as a signal for what the user cares about. That means the corpus of past exchanges — questions asked, topics returned to, preferences expressed in back-and-forth — becomes an input for deciding what proactive research to run.

Second, Pulse uses "feedback." Users' reactions to prior output, not just their queries, shape what the system investigates next. This establishes a feedback loop: the user responds to a Pulse update, and that response tunes future updates.

Third, Pulse connects to external applications, with OpenAI naming the calendar as an example. This is the component that turns Pulse from a summarization feature into something closer to an assistant that operates with awareness of the user's schedule and commitments. Connected apps give the proactive system context that pure chat history cannot supply.

OpenAI has framed the release as a preview, and it is limited in two visible ways: it is available to Pro subscribers only, and only on mobile. Both limits are consistent with how the company has staged prior feature rollouts — narrow availability first, wider distribution later, if the preview behaves as intended.

Why proactive AI is a consequential step

The stakes here extend beyond a single feature flag. Reactive chatbots put the burden of initiation entirely on the user. The user must know what to ask, remember to ask it, and return to the product to ask again. A proactive system removes each of those friction points — and removes the corresponding moments of user control.

For OpenAI, that has clear commercial logic. A chatbot that waits to be addressed competes for attention every single session. A system that delivers unprompted, personalized updates creates a recurring touchpoint on its own schedule. Pro subscribers pay for the highest tier of ChatGPT access, and Pulse gives that tier a differentiating capability at a moment when premium AI subscriptions face intensifying competition across the industry.

The personalized-update model also changes the economics of engagement. If Pulse reliably surfaces information the user would otherwise have searched for, it captures queries that might have gone to a search engine, a news app, or a rival assistant. OpenAI did not quantify any usage targets in the announcement, but the design points in that direction.

The data dimension

Pulse's three input streams — chats, feedback, and connected apps — concentrate a substantial amount of personal signal in one system. The calendar integration is the most notable of these. Schedule data reveals meetings, deadlines, travel, and the people a user works with. Combined with chat history, it allows the proactive system to infer not just topics of interest but time-sensitive relevance: what matters today, what is coming up, what the user is likely to need soon.

OpenAI's announcement does not detail the privacy controls attached to these inputs, how long Pulse research data is retained, or whether users can exclude specific conversations or apps from the system. Those questions typically get answered in documentation accompanying wider releases. For now, the announcement's language is the authoritative description: personalization based on chats, feedback, and connected apps like the calendar.

Why a preview, and why mobile

The preview framing signals that OpenAI expects the behavior of a proactive system to need tuning. A chatbot that answers a bad question produces one bad answer. A proactive system that misjudges a user's interests produces unwanted interruptions, and interruption is the failure mode users punish fastest. Feedback as an explicit input suggests OpenAI anticipates needing a correction mechanism from day one.

The mobile-first constraint is equally deliberate. The phone is where push updates already live, where users check in briefly and often, and where a proactive assistant fits established behavior patterns. Desktop use skews toward long, task-focused sessions initiated by the user. Mobile is the natural habitat for a product that initiates contact.

Limiting the preview to Pro subscribers also gives OpenAI a smaller, paying, more tolerant population during the period when Pulse's judgment about what deserves attention is least reliable.

The competitive and research context

Proactive assistance has become a shared direction across the AI industry, and for an identifiable reason. Reactive systems hit an engagement ceiling: they can only be as present in a user's life as the user's own prompt frequency allows. Research on agent-style systems — models that plan, act, and report back with limited supervision — treats proactive behavior as a core capability, not an add-on.

OpenAI's move puts a version of that capability into a shipping consumer product for its highest-tier subscribers. The company has not published benchmark results for Pulse; the announcement describes an experience, not a measured capability. What it demonstrates is deployment: proactive research operating on real user data, in a live product, at scale for the Pro subscriber base.

That deployment makes Pulse a data-gathering exercise as much as a product. Every user reaction to a Pulse update — acted on, ignored, or corrected through feedback — becomes evidence for refining the system's judgment about what to surface and when.

What to watch next

The announcement leaves the rollout path unstated beyond today's preview. The observable questions for the coming months: whether OpenAI extends Pulse beyond Pro subscribers, whether it reaches desktop and web, which additional connected apps join the calendar, and what controls users receive over the system's access to their history and schedules.

Each of those decisions will determine whether Pulse remains a premium convenience or becomes the default way ChatGPT relates to its users — an assistant that no longer waits to be asked.