OpenAI Pages turns documents into shared spaces for AI agents

OpenAI used DevDay in San Francisco on 29 September to introduce Pages, a word processor designed for people and AI agents to work together. TechCrunch reported that Pages can support writing, research, charts, images and visualisation. OpenAI also introduced Space, where colleagues can work with ChatGPT and their own Dots, plus collaborative slides made through conversation in ChatGPT. The announcements move AI assistance closer to the document itself rather than keeping it in a separate chat window.

That change matters because the familiar line between drafting with AI and editing AI output becomes harder to describe when an agent can work inside the document over time. A page can be given instructions to check a team channel and update itself with what it finds, according to TechCrunch. Instead of a simple sequence where you ask for text, receive it and revise it, the document can become an active workspace where human and automated contributions may arrive at different stages.

Space extends the same idea beyond a single page. TechCrunch reported that pages and files can live together there while colleagues collaborate with ChatGPT and Dots. OpenAI describes Dots as always on agents that pursue goals continuously in the background with minimal oversight. They are available to ChatGPT Pro and Business Premium subscribers in eligible markets. For ordinary writers, this means some AI contributions may come from continuing tasks rather than from one clearly defined prompt.

That creates a practical authorship question for anyone who must declare AI use to a university, journal, editor or employer. Saying that AI was used may still be accurate, but it can leave out important context if an agent researched material, refreshed a page or contributed repeatedly during revision. A useful disclosure policy needs to account for the role AI played in the work, not only whether a chatbot produced an early draft. The relevant distinction is increasingly about contribution and responsibility.

One useful habit is to keep a simple record of what you asked an AI system to do and what it returned or changed. That record does not need to become a complicated audit trail for every sentence. It can instead capture the significant steps, such as research requests, generated passages, chart creation, image creation or instructions that allowed an agent to update a page. If someone later asks how the work was produced, you have a clearer basis for explaining the process.

The scale of ChatGPT makes this more than a niche workflow question. The Verge reported that OpenAI said ChatGPT now has 1.2 billion users each week. As tools such as Pages, Space, Dots and collaborative slides bring generation and ongoing agent activity into everyday documents, writers may need language that describes mixed authorship more precisely. You might have written some sections directly, revised generated material elsewhere and allowed an agent to maintain another part of the same project.

The practical answer is not to make AI assisted work look as though no AI was involved. It is to follow a clear disclosure policy that fits the university, journal, editor or employer receiving the work. When AI is part of the process, keep enough information to explain what it did and what you did. Pages and Dots make that habit more useful because the contribution may continue inside the workspace rather than arrive once as a block of text. Clear process records make honest disclosure easier.