A hands-on review published Oct. 2, 2026, reported that OpenAI Dots completed a website update and assembled a video from local files after receiving access to the reviewer’s computer. Several online errands, by contrast, ran into security checks or needed human help. The early results point to a practical split: Dots handled work on personal files and systems more smoothly than some tasks on third-party websites.
Website work and video assembly
The reviewer reported that Dots redesigned and deployed an update to a personal website, then assembled local video files into a clip sized for social media. The work followed access to the reviewer’s computer and extensive permissions. For the website, the reviewer spent 10 minutes describing changes by voice before receiving a new version to review.
Those are results from individual tasks, not a general performance score. Still, they show the kind of work that went well in this test: tasks involving files and a website the user could grant Dots access to.
Online errands needed human intervention
The same hands-on account described several snags on websites. During an internet-provider checkout, a press-and-hold human check stopped the workflow until the reviewer completed it. The site then asked for bank-account details in exchange for a $5 monthly bill discount; the reviewer chose to finish the purchase herself.
Dots also offered a $35 coworking day pass instead of finding the free tour and entry option the reviewer wanted. An IKEA account security check kept looping, while a local restaurant initially blocked its ordering page. Dots eventually placed a teriyaki order through Uber Eats after the reviewer helped it log in.
The results varied by task. In one case, a person handled a verification step; in another, login help let the order proceed. OpenAI says Dots can make mistakes and advises users to review consequential work.
What OpenAI says about Dots, plans and usage
OpenAI announced Dots on Sept. 29, 2026, as a persistent AI agent accessed through ChatGPT. OpenAI says it uses GPT-6 Astra, has its own cloud computer and browser, and can connect to more than 4,000 apps through plugins. Users can also grant it access to their own computer. OpenAI lists ChatGPT, voice, Slack and Microsoft Teams as ways to interact with Dots.
OpenAI says the first Dot is included with eligible Pro and Business Premium subscriptions. Enterprise, Edu and Healthcare workspaces can try a beta when an administrator enables it. In the reported test, the Pro subscription cost $100 per month; that was the account price, not a separate fee for Dots.
OpenAI also distinguishes between conversations with Dots and work it delegates to other products: Dot conversations do not count toward ChatGPT usage limits, while tasks started in Codex or ChatGPT Work count toward those products’ limits. OpenAI says its background proactive-research tools are read-only and cannot send messages, change connected-app content, or control a browser or computer.
Who might find Dots useful?
Based on the reported results, Dots may be a better fit for people who want help with their own files or computer-based work and can review the output than for anyone expecting every online purchase or booking to run without assistance. That is a practical reading of these individual tasks, not a guarantee of how another user’s workflows will go.
Meta Muse is another agent to consider. In a separate early-use comparison, its interface was judged easier for beginners, while Dots fit an existing Codex and Slack workflow more closely. Those were individual impressions, not a controlled comparison.
