Last reviewed
May 16, 2026
Sources
13 citations
Review status
Source-backed
Revision
v3 ยท 5,021 words
Add missing citations, update stale details, or suggest a clearer explanation.
ChatGPT Pulse is a proactive personalized briefing feature within ChatGPT, an artificial intelligence chatbot developed by OpenAI. Launched in preview on September 25, 2025, Pulse enables ChatGPT to perform asynchronous research overnight and deliver personalized daily updates to users in the form of visual cards.[1] These updates are based on the user's chat history, saved memories, feedback, and optionally connected applications such as Gmail and Google Calendar.[2] The feature represents a shift from reactive to proactive artificial intelligence assistance, as described by OpenAI CEO Sam Altman, who called it his "favorite feature of ChatGPT so far."[3]
Initially available exclusively to ChatGPT Pro subscribers on mobile devices, Pulse aims to evolve into a core component of ChatGPT, with plans for expansion to Plus users and broader availability.[1] It is designed to provide timely, relevant information without requiring user prompts, such as follow-ups on discussed topics, event suggestions, or progress toward personal goals.[4] Pulse is part of a broader strategic pivot at OpenAI under the leadership of Fidji Simo, the company's CEO of Applications, who has championed asynchronous and agentic experiences as the next phase of consumer AI.[10]
ChatGPT Pulse performs asynchronous research on behalf of users, synthesizing information from multiple sources including chat history, saved memories, user feedback, and optionally connected applications.[2] The feature delivers five to ten personalized summaries each morning in the form of visual cards that users can quickly scan or expand for detailed information.[4] Each card typically includes a brief synthesized summary, an AI-generated image, suggested follow-up prompts, and links to original sources for transparency.[12]
The feature marks a significant evolution in OpenAI's consumer strategy, moving beyond traditional prompt-based interactions toward more autonomous agent-like functionality.[5] OpenAI positions Pulse as "the first step toward a new paradigm" where ChatGPT proactively researches, plans, and eventually takes helpful actions based on user direction.[1] In the launch communications, Altman framed Pulse as evidence of a longer arc in which ChatGPT moves from being "all reactive to being significantly proactive, and extremely personalized," comparing the desired experience to that of a "super-competent personal assistant."[3]
The Pulse interface intentionally departs from the design conventions of social feeds. Rather than presenting an unbounded list of items, each daily briefing ends after a finite number of cards, often with a message such as "Great, that's it for today."[6] OpenAI product lead Adam Fry has described this as a conscious choice to differentiate the experience from engagement-maximized social media products that rely on infinite scroll and algorithmic feeds.[6] The combination of overnight processing, finite morning delivery, and integration with personal data is intended to position Pulse as a daily ritual rather than a continuous distraction.
ChatGPT Pulse was developed as part of OpenAI's efforts to advance AI from a question-answering tool to a proactive assistant capable of independent research and action.[1] The feature builds on existing ChatGPT capabilities like Memory, chat history, and connectors, integrating them with explicit user feedback and external data sources.[2] Pulse sits within a broader portfolio of asynchronous products that OpenAI introduced during 2025, including ChatGPT Agent and the coding-focused Codex, which similarly perform work for users in the background rather than in a synchronous dialogue.[10]
Fidji Simo, OpenAI's CEO of Applications, led the strategic positioning of the feature within OpenAI's consumer roadmap. Simo, who joined OpenAI in August 2025 after serving as CEO of Instacart and a long tenure at Facebook, has publicly framed personal AI assistants as a means of democratizing services previously reserved for the wealthy.[10][7] Adam Fry, OpenAI's product lead for proactive experiences, oversaw the user interface design that intentionally limits content per day in order to avoid the engagement patterns of social media.[4][6] Christina Wadsworth Kaplan, OpenAI's head of personalization, led the development of the personalization stack that draws on chat history, Memory, and connectors to shape each user's daily briefing.[4]
OpenAI partnered with college students through the ChatGPT Lab program to gather early feedback and refine the feature. Key insights from this collaboration included the importance of user guidance in making Pulse valuable, leading to the implementation of enhanced feedback mechanisms and a dedicated "curate" affordance for steering future briefings.[1] Early testing participants provided feedback on its utility for tasks like calendar management, study planning, and trip preparation. For instance, student Isaac Seiler reported receiving helpful updates on train and commute information for a trip to Taiwan based on prior conversations about grant planning.[1]
These collaborations also surfaced limitations early in development. Participants noted that Pulse occasionally produced irrelevant suggestions when it failed to detect that a project was complete, and that the feature was most valuable when users actively rated cards and provided written curation. These findings shaped the launch design, which prominently features thumbs up and thumbs down controls alongside the curate input.[1][12]
OpenAI publicly introduced ChatGPT Pulse on September 25, 2025, with a dedicated blog post titled "Introducing ChatGPT Pulse" and a coordinated launch through major technology press.[1] Altman amplified the announcement on X with a post in which he called Pulse his favorite ChatGPT feature so far, describing how it "works for you overnight, and keeps thinking about your interests, your connected data, your recent chats, and more."[3] In a launch briefing reported by TechCrunch and Platformer, Fry demonstrated Pulse using examples such as a daily Arsenal soccer news brief, Halloween costume suggestions, and a family travel itinerary.[4][6] Wadsworth Kaplan demonstrated how Pulse had generated running routes and dietary-aware restaurant ideas for an upcoming London trip based on her saved preferences.[4]
| Date | Event |
|---|---|
| August 2025 | Fidji Simo begins role as OpenAI CEO of Applications, overseeing consumer products including Pulse[10] |
| September 25, 2025 | Official announcement and launch of ChatGPT Pulse preview for Pro subscribers on mobile[1] |
| September 25, 2025 | Sam Altman publishes X post calling Pulse his favorite ChatGPT feature so far[3] |
| September 26, 2025 | Wider tech press coverage from TechCrunch, Platformer, The AI Insider, and others[4][6] |
| Late 2025 / early 2026 | Web client access for Pulse rolled out to Pro subscribers[8] |
| 2026 (planned) | Gradual expansion to ChatGPT Plus, with Team and Enterprise integration also on roadmap[8] |
| 2026 (planned) | Long-term goal of broader rollout to all ChatGPT users as efficiency improves[1] |
ChatGPT Pulse uses overnight asynchronous research to assemble each user's morning briefing. Once a user enables Pulse and Memory, the system runs a research pass during off-peak hours that draws on the user's recent chats, long-term Memory entries, thumbs up and thumbs down feedback, curated topic requests, and content from any opted-in connectors.[2][12] The output is a small set of self-contained cards, typically between five and ten, that synthesize the information gathered into focused topics.[4]
Because Pulse is compute-intensive, OpenAI has indicated that this overnight processing is one of the main reasons for limiting initial availability to the ChatGPT Pro tier. According to Altman, more compute-heavy features will continue to be gated to higher tiers while the company expands data center capacity, including through partnerships with Oracle and SoftBank.[4] Submitting a curation request before approximately 10 PM local time gives Pulse a window in which to incorporate that guidance into the next morning's briefing.[12]
The Pulse interface is built around a stack of visual cards rather than a continuous text stream. Each card consists of a short headline, a synthesized summary, an AI-generated image, and optional follow-up actions such as opening a related conversation, saving the card to chat history, or sending feedback. Users can expand a card to read a longer treatment, ask follow-up questions, or pivot into a regular chat thread.[4][12] Cards that are not engaged with are automatically removed after 24 hours, and only cards that are explicitly saved or that spawn a conversation persist in the user's chat history.[8]
This ephemeral design is paired with a deliberately finite daily quota. Once a user has scrolled through the day's cards, Pulse explicitly indicates that the briefing is complete instead of generating additional content on demand.[6] OpenAI executives have framed this as part of an effort to avoid replicating the doom-scrolling patterns of social media and short-form video products.
The Pulse personalization engine combines several inputs:[2][12]
Wadsworth Kaplan has highlighted that thumbs up and thumbs down feedback only personalizes the experience for the individual user and does not propagate preferences across other users' Pulse outputs.[4]
Pulse's personalization quality is closely tied to how much context the user is willing to share. The feature is designed to function with minimal data through Memory and chat history alone, but its most distinctive use cases tend to involve connectors that bring in calendar events, emails, and contacts.
| Data source | Description | Default state | Notes |
|---|---|---|---|
| Chat history | Previous conversations with ChatGPT, including ongoing projects and questions | Required (Memory must be enabled) | Forms the baseline of Pulse personalization[2] |
| ChatGPT Memory | Long-term saved facts and preferences across sessions | Required | Pulse cannot run without Memory enabled[2] |
| User feedback | Thumbs up and thumbs down ratings, plus written curation prompts | Always active when Pulse is on | Influences only the individual user's future briefings[4] |
| Gmail connector | Read access to email content for context, summaries, and suggestions | Off by default | Content from Gmail is not used to train OpenAI models when connected via the connector[12] |
| Google Calendar connector | Read access to events, attendees, and times for meeting prep and reminders | Off by default | Used for agenda generation, conflict-aware suggestions, and birthday reminders[12] |
| Google Contacts connector | Read access to contact information | Off by default | Supports personalized references in emails, gifts, and event suggestions[12] |
| Box connector | Access to documents in the user's Box workspace | Off by default | Listed among supported connectors at launch[4] |
| Timely news and trends | OpenAI-supplied news context, weather, and similar feeds | Always active when Pulse is on | Used to balance personalization with broader discovery[12] |
Connectors are configured under Settings, in a dedicated Connectors panel, where each integration can be toggled on or off and where users can grant or revoke access at any time.[12] OpenAI states that connector content is not used to train its models and is automatically deleted from Pulse's daily output after 24 hours unless explicitly saved.[1][12]
When enabled, Pulse appears as a dedicated tab inside the ChatGPT mobile app for Pro subscribers and, since late 2025, as a corresponding section in the ChatGPT web app for Pro accounts.[8] Within this surface, users can:[1][12]
These controls reflect a broader pattern in OpenAI's recent consumer products of separating Memory, connectors, and proactive behavior into distinct toggles, allowing users to combine them or restrict them as they see fit.
At launch, ChatGPT Pulse was available exclusively to subscribers of the ChatGPT Pro plan, priced at US$200 per month.[4] ChatGPT Pro was introduced by OpenAI in December 2024 as the top tier of ChatGPT, originally designed to provide unlimited access to OpenAI's most capable models, including reasoning models such as o1 pro and later GPT-5 Pro.[7] Pulse extended this pattern of gating the most compute-intensive features behind the Pro tier, joining other Pro-only offerings such as advanced agentic browsing and long-running tasks.
OpenAI executives have repeatedly cited compute capacity, rather than business strategy, as the rationale for the Pro-first rollout. Altman noted at launch that the company would need to expand infrastructure before opening Pulse to lower-priced tiers, referencing partnerships with Oracle and SoftBank for new data center capacity.[4] Fry similarly framed Pulse as a preview that would benefit from being refined with a smaller, more committed user base before broader release.[4]
OpenAI has publicly committed to expanding Pulse beyond the Pro tier, but has not announced specific dates. The official launch blog post and accompanying interviews indicated that ChatGPT Plus subscribers, who pay roughly US$20 per month, would be the next group to receive access once Pulse's efficiency improved.[1][4] Reporting in late 2025 and early 2026 noted that Pulse had reached the web for Pro users but that Plus and free-tier rollouts remained unconfirmed.[8] Coverage of OpenAI's business roadmap also suggested that ChatGPT Team and ChatGPT Enterprise customers would eventually receive Pulse as part of their plans, though detailed timing had not been published.[8]
| Plan | Approximate price | Pulse access at launch | Planned access |
|---|---|---|---|
| ChatGPT Free | US$0 | None | Possible long-term once efficiency improves[1] |
| ChatGPT Plus | ~US$20 / month | None | Confirmed planned rollout, timing unannounced[1][4] |
| ChatGPT Pro | US$200 / month | Available on mobile from September 25, 2025; web added later | Continues with new capabilities first[1][8] |
| ChatGPT Team | ~US$25 / user / month | None at launch | Expected as part of broader Team rollout[8] |
| ChatGPT Enterprise | Custom enterprise pricing | None at launch | Expected as part of broader Enterprise rollout[8] |
OpenAI has positioned Pulse as an opt-in proactive experience layered on top of existing ChatGPT privacy controls. The feature is unavailable unless ChatGPT Memory and "reference chat history" are both enabled, and connector data is fully opt-in. Within the Pulse and Connectors settings, users can:[1][12]
OpenAI has stated that content from Gmail and Google Calendar connectors is not used to train its models, and that Pulse cards are deleted automatically after 24 hours unless saved or engaged with.[1][12] Interactions with Pulse may still contribute to model improvement if the user has left the global "Improve the model for everyone" setting enabled, which is a separate control from the Pulse-specific options.[12]
Pulse content goes through OpenAI's standard safety filters before being delivered to users, with the goal of preventing harmful content from appearing in the morning briefing. Topics that would violate OpenAI's usage policies are excluded from the candidate set during overnight generation.[1] Coverage of the launch noted that OpenAI's policy and safety teams were continuing to review Pulse for potential issues related to echo chambers, mental health, and overpersonalization, with adjustments to be made over time as the feature scales.[4]
Because Pulse can reach into private email and calendar data when connectors are enabled, OpenAI has emphasized read-only access patterns and minimal-permission scopes for connected services. Independent guides on enabling Pulse have advised users to grant the narrowest permissions consistent with their use cases, particularly for Gmail and Calendar.[9]
As a preview feature, Pulse has acknowledged limitations:[1]
| Aspect | ChatGPT Pulse | Traditional ChatGPT |
|---|---|---|
| Initiative | Proactive; ChatGPT starts the conversation with a daily briefing[1] | Reactive; responds when prompted by the user |
| Cadence | Once daily; research runs overnight and is delivered in the morning[2] | On-demand within a conversational session |
| Output format | Scannable visual cards, expandable for detail and follow-up[1] | Standard conversational text and images |
| Data sources (opt-in) | Chats, Memory, feedback, Gmail, Google Calendar, Contacts, Box, news[2][12] | Primarily in-session context unless tools or connectors are explicitly used |
| Controls | Curate prompts, thumbs up and down, delete feedback history, disable Pulse or connectors in settings[2] | Standard chat controls and data settings |
| Persistence | Cards expire after 24 hours unless saved or extended into a chat[8] | Conversations persist by default in chat history |
| Tier availability at launch | Pro only, mobile then web[1][8] | Available across Free, Plus, Pro, Team, and Enterprise plans |
ChatGPT Pulse enters a crowded market for daily information products, ranging from algorithmic news feeds to enterprise productivity briefings. Compared with these tools, Pulse differs primarily in its use of personal AI memory and its conversational follow-up surface.
| Product | Approach | Personalization basis | Output format | Source citations |
|---|---|---|---|---|
| ChatGPT Pulse | Proactive daily briefing inside ChatGPT mobile and web for Pro[1] | ChatGPT Memory, chats, feedback, Gmail and Calendar connectors[2][12] | Visual cards with follow-up chat | Yes, via links on cards[12] |
| Google Discover | Algorithmic news and content feed in the Google app and Chrome | Search history and Google account signals | Continuous feed of articles | Links to publishers |
| Apple News | Curated and algorithmic news feed in the Apple News app | Apple ID interests and explicit topic follows | Articles, audio briefings, and stories | Links to publishers |
| Perplexity Daily and Discover | Curated topic stories and scheduled tasks in Perplexity | Topic follows, scheduled queries, and search history | Cited summaries and threads | Yes, with strong source emphasis |
| Arc Browser Catalyst and Boosts | Browser-side personalization and one-off remixes from The Browser Company | URL-based interventions and user-installed boosts | In-page UI changes, not a feed | Not applicable |
| Microsoft Copilot Daily and Viva Briefing | Audio and email briefings inside Microsoft 365 environments | Microsoft Graph (mail, calendar, files) | Audio digests and email summaries | Limited |
| Feedly Leo | AI-curated newsletter and dashboards from RSS feeds | RSS subscriptions, prioritization rules | Newsletter and dashboards | Yes, with article links |
Pulse's distinguishing factors include its reliance on conversational memory accumulated across all prior ChatGPT use, its ability to pivot from a card into a free-form chat, and the deliberate finiteness of its daily briefing. By contrast, Google Discover and Apple News are designed around continuous feeds, while Perplexity Daily and Viva Briefing focus on specific verticals such as cited research or workplace productivity. The Browser Company's Arc Catalyst and Boosts represent a different paradigm, modifying web pages rather than producing a feed of summaries.
Reporting and OpenAI's own demonstrations have highlighted a number of categories where Pulse is intended to add value:
In enterprise-style usage hinted at by the OpenAI roadmap, Pulse is expected to support knowledge workers by combining email, calendar, and document context into morning briefings tailored to current projects, in a manner analogous to Microsoft's Viva Briefing but built on top of conversational AI.[8]
Early reviews praised Pulse's card-based interface and its conversational follow-up loop. Coverage from outlets such as TechCrunch, Platformer, The AI Insider, and various practitioner blogs highlighted the design choice to limit each day's briefing to a finite set of cards and to encourage completion rather than continuous scrolling.[4][6][11] Demonstrations of personalized travel itineraries, sports updates, and project follow-ups were frequently cited as differentiators from generic news feeds.
Altman's framing of Pulse as the next step toward a "super-competent personal assistant" was repeated widely in technology press coverage, with several outlets noting that Pulse represented one of OpenAI's most ambitious attempts to date to make ChatGPT feel like a continuously present collaborator rather than a tool used in discrete sessions.[3][11]
Pulse also attracted significant criticism, particularly around privacy, monetization, and the broader implications of proactive AI. A representative Hacker News discussion thread about the launch surfaced concerns that included the following themes:[13]
Independent commentary has also questioned whether Pulse risks creating filter bubbles by reinforcing user preferences through thumbs up and thumbs down feedback, and whether the use of overnight AI research compounds the risk of hallucinated facts being presented as morning fait accompli without prompt review.[8] Privacy-focused commentary noted that Pulse fits a broader pattern in which conversational AI providers ask for increasing access to personal data in exchange for richer assistance, and called for clearer disclosures around how connector content is processed and retained.
The Pro-only pricing of US$200 per month at launch was another common point of criticism, with some commentators arguing that Pulse's debut as an exclusive feature for a small subset of paying users limited its impact and made it difficult to assess as a mass-market product.[8][13]
Industry analysts have framed Pulse as part of a broader shift toward "ambient agents" that operate continuously in the background, anticipating user needs without explicit prompting.[5] Coverage described Pulse as a deliberate move toward an AI agent pattern that proactively summarizes news or personal context and then invites action, while keeping the experience intentionally finite each day.[6][4] Some observers compared the strategic intent of Pulse to other proactive AI initiatives at Anthropic, Google, and Microsoft, all of which have invested heavily in calendar-aware and email-aware assistants.
Fidji Simo's framing of Pulse as part of an effort to democratize access to high-quality personal assistance was widely reported. She stated at launch: "We're building AI that lets us take the level of support that only the wealthiest have been able to afford and make it available to everyone over time."[4] In subsequent interviews during late 2025 and early 2026, Simo positioned Pulse alongside ChatGPT Agent and other asynchronous experiences as steps toward a "true personal super-assistant" capable of helping users manage important parts of their lives and work.[10]
Pulse plays a strategic role in OpenAI's effort to differentiate ChatGPT from competing chat assistants and to deepen daily engagement with paying subscribers. By turning the morning into a structured ritual centered on ChatGPT, OpenAI hopes to establish ChatGPT as the first place users check for information, planning, and follow-ups. This positioning aligns with statements by Simo and Altman about ChatGPT becoming a continuously available assistant rather than a tool used only when explicitly summoned.[10][3]
The Pro-only rollout also reflects how OpenAI has used its top subscription tier as a testbed for compute-intensive features. ChatGPT Pro launched in December 2024 at US$200 per month with unlimited access to OpenAI's most capable models, and has since served as the first home for products such as advanced agentic workflows and Pulse.[7]
Industry analysts note that Pulse positions ChatGPT to compete with multiple categories of incumbent products:[5][8]
The feature represents part of a broader industry trend toward ambient and asynchronous AI agents that operate continuously in the background, anticipating user needs without explicit prompting.[5] Multiple competing products, including Perplexity's scheduled tasks and Daily Discover surfaces, Microsoft Copilot Daily audio briefings, and Google's Discover feed within Gemini and Search, have followed similar trajectories during 2025 and into 2026.[5]
Observers have repeatedly raised the question of how OpenAI might eventually monetize Pulse beyond subscriptions. The visual card format and recurring daily delivery resemble products that monetize through advertising or sponsored placements, and Altman has publicly hinted at exploring advertising-supported experiences within ChatGPT.[13] OpenAI has not committed to introducing ads in Pulse, but commentary at launch noted that the surface area created by Pulse, including curated cards with follow-up calls to action, would be well suited to sponsored content if the company chose to pursue that path in the future.
OpenAI has outlined several planned enhancements:[1][8]
Altman has framed Pulse as one step in a longer trajectory toward significantly more proactive and personalized AI, describing the goal as making ChatGPT a "super-competent personal assistant" that anticipates user needs.[3] Simo has echoed this framing in subsequent communications, positioning ChatGPT as a future "personal super-assistant" capable of managing important parts of users' lives and work.[10]