OpenAI Dots: dot.com Redirects to Grok, and What Dots Actually Does

On 29 September 2026, OpenAI launched an always-on agent called Dots, and the domain dot.com was already sending visitors to the Grok Bot page of its rival xAI.

The redirect is the joke, and the joke has drawn most of the coverage. The product is the story that costs you money. OpenAI Dots is a persistent agent with its own cloud computer, and it is rolling out to a narrow set of plans and regions, with one dot included and more promised for later. If you pay for ChatGPT, or you approve software for a team of five to fifty people, the rollout rules matter more than the domain trivia.

What changed on 29 September: a launch, a redirect and a date

Two things happened within hours of each other, and only one of them was planned.

OpenAI announced Dots at DevDay and published its safety write-up the same day, according to the OpenAI safety post dated 29 September. TechCrunch then reported that dot.com belongs to xAI and forwards to the download page for Grok.

Here is the sequence, with the date each claim was checked.

We ran the last two checks ourselves. A request for dot.com at 23:50 UTC on 30 September (07:50 on 1 October in Singapore) returned a permanent redirect to x.ai/bot, served through CloudFront. The registry lookup for dot.com shows a registration date of 22 June 1994, an expiry of 21 June 2035, and a last-changed stamp of 28 July 2026.

What the dot.com evidence shows, and what it does not

The redirect is real and current. Everything past that is inference.

The registry record does not name an owner, so we cannot confirm who holds dot.com from public data. "Last changed" is also not the same word as "transferred": it covers any edit to the record, and TechCrunch's "transferred in July" comes from a Whois lookup we could not see. The record is consistent with a July change. It does not prove a sale.

TechCrunch gave two readings. The ordinary one is that xAI bought a short name and pointed it at its bot product, since "dot" could be a typo of "bot". TechCrunch also noted that xAI does not own bot.com, and that vot.com is listed for sale. The internet's theory, in TechCrunch's telling, is that someone at xAI learned the product name early and bought the domain as a jab.

TechCrunch said it had asked xAI to explain. We found no published answer from either company when we checked. The second domain tells you little. Dots.com, the closer match to the product name, forwards to www.dots.com, whose record last changed in October 2022. On 1 October the page was a "Welcome Back" relaunch notice for a clothing brand called Dots, with no mention of OpenAI or AI agents.

What dots.com showed when we loaded it, captured 1 Oct 2026. It is unrelated to the agent.

So the honest summary is short. A domain moved, a redirect points at a competitor, and nobody with knowledge has said why. Treat any stronger claim as a guess.

What dots does all day

Strip away the avatar and a dot is three things: a goal list, a computer, and a schedule.

You give it responsibilities, not prompts. OpenAI's own examples include a finance lead who wants an investor deck refreshed as revenue numbers change, an engineer tracking calls to an API until they reach zero, and a developer whose dot watches customer feedback, then builds and tests a small fix. The dots product page says a dot "keeps making progress between conversations" and brings back work for review.

Each dot runs on its own cloud computer, powered by GPT-6 Astra. Other models are listed in the models directory. It can browse, create files and run tools there, and it only reaches your machine if you switch that on. The OpenAI help article, updated on the day of launch, lists what else it can do:

• Run scheduled tasks and recurring checks, reviewed under In progress, Scheduled and Completed.
• Research in the background and form memories from connected apps, even when you have not asked.
• Start Work or Codex tasks on your behalf, including Codex cloud environments you created first.
• Work inside Slack and Microsoft Teams, and use your personal email account once connected.

It also has limits that marketing skips. A dot cannot call you at launch. It has no standalone email address of its own. You cannot create one on mobile, and mobile web is not supported. Texting is a limited beta for Pro users in the US only, through a third-party provider.

The closest sibling inside OpenAI is ChatGPT Work. Work is a task you start and watch. A dot owns a standing job. The two share limits, because tasks a dot starts in Work or Codex count against your usage.

Who this affects: the plans, regions and workspaces that get a dot

Access is where the launch gets narrow. The rules below come from OpenAI's help article and its ChatGPT pricing page, both read on 1 October.

The pricing page lists "Dot, your always-on agent" on the Pro card and leaves it off Plus. So the everyday ChatGPT subscriber at the middle tier is out. Standard Business seats are out too: only the Premium seat qualifies.

OpenAI also says rollout is gradual and "may take several days" to reach your account. Enterprise is off by default until an admin flips it.

Who this hurts: the buyers outside the launch

Every launch has people it skips, and this one has four groups.

Pro subscribers in the EEA, Switzerland and the UK pay for the plan and get no dot, at least for now. OpenAI gives no date for their markets.

Plus subscribers see the feature on the pricing table and cannot use it. Moving to Pro means choosing one of three usage tiers, and OpenAI's page shows Pro only as a "from" price that did not render when we read it, so we cannot tell you the number.

Business admins on standard seats face a real step, because only Premium seats qualify and they cost five times as much. Enterprise admins carry the most work: the beta arrives switched off, and someone has to write the policy before anyone turns it on.

Independent agent vendors feel it last. An always-on agent bundled into a plan that teams already pay for is a hard price to beat, and our roundup of the best AI agents is now worth rereading with that bundle in mind.

What a dot costs, and what the bundle hides

OpenAI says the first dot comes with Pro or Business Premium at no extra cost. Your plan also includes an allowance for deeper work, with extended limits for the first month after launch. That phrase is a date: the generous limits are a launch offer, not a steady state.

The numbers we can source are these.

• Business Premium seats cost $125 per user per month, or $100 on annual billing, according to The Decoder on 11 August 2026.
• Standard Business seats stay at $25 per month, or $20 on annual billing, per the same report.
• Premium seats carry five times the usage of standard seats, with weekly resets.
• Conversations with a dot do not count toward ChatGPT limits, but work it hands to Codex or ChatGPT Work does.

Do the arithmetic for a team of ten on monthly billing:

The gap is the real price of "a dot for everyone". It buys five times the usage, not ten times the output.

OpenAI says you will later be able to add more dots and raise a dot's speed or monthly output. It has not published those prices. Expect the included dot to become the free sample.

Data and privacy: what a dot can reach

A dot holds more of your information than a chat does, so the data rules deserve a close read. They come from OpenAI's safety and privacy post.

Training depends on your plan. For Business, Enterprise and Edu workspaces, content is not used to train models by default. For personal plans, your "Improve the model for everyone" setting decides, and when it is on, training can include actions a dot takes and automations you set up, after personal identifiers are removed.

Background research has its own wrinkle. OpenAI says it does not train directly on those research threads. But a note from that research can enter a later conversation, and that conversation may be used for training, depending on your settings. OpenAI's own example is a Lisbon trip note.

Disconnecting an app stops new sharing. It does not erase what the dot already learned: to wipe that, you reset the dot, which deletes its conversations, memories and scheduled tasks.

What it can do without asking is narrower than you might fear:

1. Read what you connected, analyze it and draft work.
2. Buy only with cards already saved on a merchant's site, and only with your approval.
3. Ask before permanently deleting data, running unrecognized software or gaining new security-sensitive access.
4. Hand back password changes and transfers between your financial accounts.

A separate system called Auto-review checks planned actions, such as the recipient of an email, against your Custom Rules before they run. OpenAI is frank that "dots can still make mistakes", and a password pasted into a readable message may stay visible to the model, since secure sign-in protects only its own flow.

Grok Bot, Muse and ChatGPT Work: where the field stands

Dots is not alone. Two other always-on agents landed within a month.

Grok Bot is xAI's version, in beta since its launch post. It comes with SuperGrok, SuperGrok Plus, SuperGrok Heavy, Cursor Pro, Pro+, Ultra and Cursor Teams plans, on desktop and iOS, with usage separate from your plan. xAI says Bots can message one another and learn a routine by watching you do it once. The buyer angle is reach: if you already pay for Grok or Cursor, it is bundled.

Muse is Meta's, launched on 8 September for US adults, through its own app and WhatsApp. The AP reports it runs on a dedicated virtual machine and can open a browser, fill in forms and negotiate for you. It skews to errands, not office work.

The short verdict, for three buyer types:

• Pick a dot if your team already lives in ChatGPT, Slack or Teams, and your work is recurring and review-heavy.
• Pick Grok Bot if Cursor or SuperGrok is already on your card and you want Bot-to-Bot handoffs.
• Pick Muse for personal errands in the US, not for company data.

A full three-way comparison is coming as its own guide. If your stack is mixed and the answer is unclear, Smart Match sorts the field by your tools and constraints.

The turn: this may change less than it looks

The best case against this article is that Dots changes little. Task agents already exist. Manus runs jobs in its own hosted environment, and what Manus is for covers where it fits. OpenClaw and Hermes Agent offer the self-hosted route, and Microsoft Copilot already sits inside the office suite. Workflow tools such as Zapier have scheduled work for years.

That argument is partly right. "Always on" is a label, and any of these can run on a timer. What a dot adds is the bundle: one memory, one set of connected apps, one bill, and approval rules you write in plain language.

It is also not proven. OpenAI's examples are vendor demos, and we have not measured how often a dot finishes a job unattended. Our own test of agents handling money, AI agents with real money, is a reminder that autonomy claims need a stopwatch. So the answer to the turn is narrow. Dots matters for the ChatGPT-shaped teams who can get it, and not much for anyone else yet.

What to do this week

You do not need to decide today. You do need a plan.

1. Check your plan and region. Pro outside the EEA, Switzerland and the UK, or Business Premium, are the day-one paths.
2. On Enterprise or Edu, keep the beta off until security has written Custom Rules.
3. Start with one recurring job that has a review step, such as a weekly feedback digest.
4. Set the training preference before you connect anything.
5. Book a date after month one to compare cost against hours saved.

For a wider view of the category, best agentic AI in 2026 and the agents directory list the alternatives. The tools directory covers the rest of the stack. The matchup between the underlying models sits in our Astra versus Grok comparison, and the chatbot-level rivalry has its own ChatGPT versus Grok guide. OpenAI's other DevDay agent, Codex Security, is worth a look if you run code.

What to watch next

Three dated things will tell you whether Dots is a product or a demo.

OpenAI said texting is coming next. Today it is a US-only beta for Pro. If it reaches Business by the end of 2026, dots become a phone-first tool, and the EEA gap will matter more.

OpenAI also promised extra dots and higher output. If a price for either appears before 31 December 2026, the included dot is a sample, and our arithmetic above needs redoing.

And watch the domain. Domains are cheap and launches are not. If OpenAI ships a dots URL of its own while dot.com still points at Grok Bot, the quiet winner of the week will be whoever bought the address first.

Frequently asked questions

What does OpenAI Dots do?

A dot is an always-on agent in ChatGPT with its own cloud computer. You give it a standing responsibility, such as refreshing a deck or watching customer feedback, and it keeps working between conversations. It brings back drafts and decisions for your review, and it can start Work or Codex tasks.

Which ChatGPT plans include dots?

Pro in markets outside the EEA, Switzerland and the UK, and Business Premium seats in all supported regions. Enterprise, Edu and Healthcare workspaces get a beta that an admin must switch on. Free, Go and Plus do not include a dot, per OpenAI's pricing page and help article read on 1 October 2026.

How much do dots cost?

The first dot is included in Pro and Business Premium at no extra cost. Business Premium seats are $125 per user per month, or $100 annual, according to The Decoder. OpenAI's Pro price did not render on its pricing page when we read it, and extra dots have no published price yet.

Does OpenAI own dot.com?

We could not confirm that. On 30 September 2026 at 23:50 UTC, dot.com returned a permanent redirect to x.ai/bot, and TechCrunch reports the domain belongs to xAI. The public registry record does not name an owner, and neither company had published an explanation when we checked.

Are dots safe to connect to work accounts?

OpenAI describes sandboxed cloud workspaces, Auto-review checks before actions, and approval rules you can edit. Business, Enterprise and Edu content is not used for training by default. OpenAI also says dots can still make mistakes, so start with low-risk jobs and review the output.

Covered in this guide

• OpenAI Dots: OpenAI's always-on agent in ChatGPT runs on GPT-6 Astra with its own cloud computer, and asks approval before consequential actions.
• Grok Bot: xAI's always-on AI agents that sign into your existing apps and finish multi-step jobs, bundled with eligible SuperGrok and Cursor plans.
• ChatGPT: ChatGPT is OpenAI's AI assistant, with 1.2 billion weekly users by OpenAI's count, the GPT-6 model family and plans that run from a free tier to a premium Pro tier.
• ChatGPT Work: OpenAI's ChatGPT Work agent plans and finishes multi-step projects across connected apps and files, launched July 9, 2026 for Business and Enterprise plans.
• Codex Security: OpenAI's application security agent scans connected GitHub repositories, tests likely bugs in a sandbox, and drafts fixes for review as pull requests.
• Cursor: Cursor is an AI code editor built on VS Code, widely adopted across large enterprise engineering teams, with Agent Mode, Tab completion, and Cloud Agents.
• GPT-6 Astra: OpenAI's flagship model, launched September 2026 as the first ever rated at the Preparedness Framework's Critical cybersecurity level.
• Grok: Truth-seeking AI chatbot with real-time web and X integration powered by advanced reasoning.
• Hermes Agent: Hermes Agent is Nous Research's MIT-licensed self-improving AI agent, past 212,000 GitHub stars by July 2026. Self-hosted, model-agnostic, and free to use.
• Manus: Autonomous AI agent that plans and executes multi-step work from one prompt: research, documents, apps and video. The September 2026 release adds Cascade, Automations and Studio. Free tier available.
• Microsoft Copilot: Your everyday AI companion powered by GPT-5.6 (preferred model July 2026), GPT-4o, and GPT-5 for productivity, creativity, and intelligent assistance
• Muse: Meta's personal AI agent books travel, negotiates bills and fills out forms from your accounts; launched Sept 8, 2026 with free, $20 and $100 tiers.
• OpenAI: OpenAI builds the GPT-5.6 model family (Sol, Terra, Luna), o3, ChatGPT (900M+ weekly users), and the OpenAI API. Closed a $122B round at an $852B valuation in March 2026, the largest private funding round in history.
• OpenClaw: OpenClaw is a local-first, open-source AI agent scoring 48.7% on SWE-bench Verified. Self-hosted and free, with a paid cloud tier available.
• xAI: Elon Musk's AI company (roughly 4,000 to 4,900 employees) builds the Grok models and merged into SpaceX, taking the combined business public on Nasdaq as the largest IPO on record.
• Zapier: The automation layer for agentic AI connecting 8,000+ apps

Sources