ChatGPT Pages and Slides Bring Real-Time Human-AI Co-Editing

OpenAI moved into collaborative documents at DevDay 2026 with Pages and a preview of Slides. Pages is a document format built for human-agent collaboration — writing, research, charts, images, and interactive elements, with comments and real-time editing where connected tools can push live updates into the document. Slides applies the same model to presentations: teammates and agents editing together with comments, launching on Pro, Business, and Enterprise in the coming weeks with export to PowerPoint and Google Slides. Spreadsheets are marked coming soon on ChatGPT Space.

This piece covers what makes Pages structurally different from a generated document, how the launch fits OpenAI's collaboration suite, what it means for the office-software incumbents whose formats it exports to, and how teams should pilot it. It is written for team leads and operators who live in documents and decks.

Pages: what OpenAI shipped

The distinction OpenAI is drawing is between a document an AI produced and a document an AI participates in. A page is persistent and shared — it lives where teammates and agents both reach it, edits happen in real time from either side, and comments thread through the collaboration. Connected tools providing updates is the live-data hook: a page can reflect current state from the systems it references rather than freezing a snapshot at generation time.

Interactive elements and charts push Pages past prose toward the living-document category — content assembled from research, data, and visuals that stays current. The launch demo framing at DevDay showed collaborative editing with visible edits from teammates, comments, and agent contributions in one surface.

Slides and the export decision

Slides brings the same co-editing to presentations, in preview now with the wider launch on Pro, Business, and Enterprise planned within weeks. The consequential detail is export: PowerPoint and Google Slides support means OpenAI is not asking organizations to abandon their presentation infrastructure — the deck can be drafted collaboratively with agents in ChatGPT and delivered in the format the meeting expects.

Export-first is the classic wedge strategy. The moment of creation moves into ChatGPT while compatibility neutralizes the switching objection; if the pattern holds, the incumbent formats become output targets rather than working environments. Spreadsheets — the deepest moat in office software — being explicitly on the roadmap signals how far OpenAI intends to push this.

Why agent participation changes the document

Real-time co-editing between humans is a solved problem; the new mechanic is agents as genuine collaborators. An agent that can edit a shared page directly — rather than pasting suggestions into a chat for a human to transcribe — closes the loop that has made AI document work feel like courier duty. Combined with dots, the always-on agents launched the same day, a page becomes somewhere an agent can leave completed work: research summarized overnight, charts updated when data changes, drafts advanced between meetings.

The comment thread is equally important for governance: agent edits arriving through the same visible, attributable mechanics as human edits keeps the collaboration auditable. Who changed what — person or model — stays inspectable in a way ambient AI rewriting does not.

Who is affected: the office suite triangle

Microsoft and Google now face a competitor that starts from the AI and adds the documents, while their own strategies bolt AI onto established suites. OpenAI's structural advantages are the 1.2 billion weekly users already in ChatGPT and the agent architecture underneath; the incumbents' advantages are decades of format depth, enterprise contracts, and the fact that OpenAI's export feature concedes where delivery still happens. Notion and the modern workspace tools face the sharper threat — Space plus Pages overlaps their core surface directly.

For teams, the near-term question is simpler: whether drafting collaboratively with an agent in ChatGPT then exporting beats drafting in the incumbent tool with its assistant. That is an empirical question about quality and friction, not philosophy.

What to pilot and what to watch

A low-risk pilot: move one recurring document — a weekly status page or a standing research brief — into Pages, connect the relevant tools, and let a dot or ChatGPT hold standing responsibility for keeping it current; measure whether the live-update and agent-edit mechanics actually reduce assembly work. For Slides, test export fidelity early — round-tripping into PowerPoint and Google Slides is where collaborative drafts historically lose formatting, and fidelity determines whether the wedge holds.

Watch the rollout cadence: Slides reaching general availability on schedule, the spreadsheet surface arriving, and mobile editing following. Watch, too, how the incumbents respond on price and bundling — the co-editing pattern is now the competitive frontier of office software.