ChatGPT is adding agents designed to keep making progress between conversations. Called dots, they receive a goal, connected apps and boundaries for autonomous work, then bring results back for review. For SEO and AI Search teams, the development raises a practical possibility: ongoing research can become a delegated workflow rather than a sequence of separate prompts.
What OpenAI has announced
OpenAI’s September 29, 2026 release notes describe dots as always-on agents powered by GPT-6 Astra, each with its own cloud computer. Users specify a goal, connect the required apps and define what the agent may do independently.
The rollout is gradual for eligible Pro and Business Premium users aged 18 or older in supported markets. Pro access initially excludes the European Economic Area, Switzerland and the UK. Enterprise access is a beta, disabled by default. Dots can be created in the desktop app or on desktop web.
OpenAI also announces an introductory month during which dot usage does not count toward eligible users’ plan allowances, with later usage terms to follow. This is a temporary arrangement, not a promise of permanently unmetered operation.
The monitoring opportunity for SEO teams
Our editorial interpretation is that an ongoing agent could reduce the need to restart context for each research pass. A proposed brand-monitoring workflow might collect relevant observations, compare them with previous findings and prepare a report when something meaningful changes.
That is a possible application, not a prebuilt SEO feature established by the release notes. The announcement does not guarantee coverage of every search surface, accurate citation tracking or a particular monitoring frequency. A useful workflow still depends on available tools, source access and a clearly defined objective.
Define evidence before delegating research
A brand mention, a supporting citation and a referred visit answer different questions. An agent instructed simply to “monitor visibility” may produce a report that combines them. A more useful objective specifies what counts as an observation and which comparisons matter.
For example, a team could ask for dated records of a defined query set, the answer wording and accessible supporting sources. It should also ask the agent to identify collection gaps. Missing data should remain distinguishable from a genuine absence of the brand.
Persistent work needs useful reporting boundaries
An agent that keeps working does not need to turn every repeated observation into a notification. Our recommendation is to define the changes worth reviewing: a new factual error, a recurring source change or a documented movement in a comparable query set.
Specify the desired deliverable and the actions permitted. Preparing an evidence summary is different from editing a website or contacting another person. Those choices should reflect the team’s actual workflow rather than follow automatically from the word “autonomous.”
Freshness and comparability remain important
Repeated research is only useful when the observations can be compared. Keep the query, date, location and device context where relevant, and preserve source links. A change in collection conditions can resemble a change in visibility.
The same discipline applies to competitor research. An apparent increase in mentions does not establish a ranking improvement, a causal effect from content changes or a business benefit. The agent’s continuity can support investigation, but it does not replace the evidence needed for those conclusions.
What changes for publishers
The broader implication is that some information gathering may continue while the user is occupied elsewhere. Publishers can consider whether their important information is clear, current and useful for recurring research tasks. There is no evidence in this announcement of a special optimization that guarantees selection by a dot.
Dots introduce an official capability for ongoing delegated work. For search teams, the next step is to evaluate a bounded research workflow against the quality of its findings, while respecting the gradual rollout and the distinction between potential use cases and demonstrated results.