OpenAI Dots is ChatGPT's new always-on AI agent. Each Dot gets its own cloud computer, can connect to thousands of apps and keeps working while you're away. I tested it the day it launched, and my verdict is blunt: it isn't worth switching to yet. It's slow, parts of it don't work, and Claude, Grok Bot and Hermes already do the job better.
📺 Watch: Is OpenAI Dots Worth It?
🆓 Start free: FREE AI Course + Community + 1,000 AI Agents — join 75,000+ members learning AI at no cost. → Join free
That's not me hating on OpenAI. I paid for the plan, set up my own Dot, called it "Big J", and pushed it through real tasks: research briefs, a website build, image generation, a restaurant booking and a phone call. Some of it impressed me. Some of it just sat there. Below you'll find exactly what happened, so you can decide whether ChatGPT Dots deserves your money right now.
If you'd rather skip the trial and error and just use the agents that already work, that's what we do inside the AI Profit Boardroom. More than 3,000 members get full courses on Claude Opus 5.5, Hermes Agent and Grok Bot, plus the exact workflows I use every day.
What OpenAI Dots is: what OpenAI actually launched
OpenAI announced Dots at DevDay on 29 September 2026. OpenAI says Dots are personal, always-on agents powered by GPT-6 Astra. The big difference from a normal ChatGPT chat is that a Dot keeps working on its own after you close the tab. You hand it a job, walk away, and it carries on.
Here's what OpenAI and launch coverage say you get:
- Its own cloud computer. Every Dot runs on a dedicated cloud machine, so it can browse, click and work in apps without touching your laptop.
- 4,000+ app connections. OpenAI says Dots can connect to more than 4,000 apps.
- Learning over time. Dots learn from your feedback, according to OpenAI.
- Calls and messages. Per launch coverage, a Dot can call and message people for you.
Availability is where it gets narrow. OpenAI says Dots are rolling out to ChatGPT Pro and Business Premium subscribers in eligible markets. The Pro launch excludes the European Economic Area, Switzerland and the UK. You get one Dot per user at launch, included in the subscription. Multiple Dots and a Microsoft Agent 365 integration are on the roadmap, and Enterprise, Edu and Healthcare workspaces get beta access through their admins.
OpenAI's launch pitch is that Dots can do simple things like book a table, or take on your most ambitious work. Keep that line in mind. I tested both ends of it.
Setting up ChatGPT Dots (and the access traps)
Getting into ChatGPT Dots was harder than it should be. You need ChatGPT Pro. My personal account worked fine. My business workspace, though, showed a message saying dots aren't available in the workspace yet. So being a paying business customer doesn't automatically get you in.
On another account I upgraded the plan first, and only then was told the plan wasn't available in my region. That's the trap. Check your region and account type before you pay, not after. If you're in the UK, Switzerland or the EEA, Pro access to Dots isn't there at launch.
Once you're in, setup itself is simple. You choose whether your Dot runs on a cloud computer or on your local computer through the desktop app. Then you customise it: a name, a colour and a little pet. I named mine Big J. It's a fun touch, but it doesn't make the agent any faster.
One more gotcha: Dots isn't available inside the Codex app. The browser version works, so use that.
📺 Watch: GPT-6 + Doug: OpenAI's Secret AI
What worked when I tested Dots
Credit where it's due. A few parts of the Dots agent genuinely worked, and some felt properly useful.
- Research briefs and PDFs. I asked for research and it produced briefs and PDFs, saved straight into the workspace.
- Website and quiz builder. This was the standout. The design was genuinely nice, better than I expected.
- Parallel tasks. You can run more than one job at the same time instead of waiting in a queue.
- Take-over mode. When the agent needs you to log in or click something, you grab the screen, do it, then hand control back. That's the right way to handle logins.
- Cloud computer controls. You can pause, rename, reboot or delete the cloud computer.
- Slack connection. Connecting Slack worked.
- Image generation. It generated images, although one of them wouldn't open (more on that below).
- The table booking. It did book a restaurant table through my connected Google account.
So the foundation is there. A Dot with its own computer, take-over mode and parallel work is the right idea for an agent. The problem is how it feels when you actually use it.
📺 Watch: New OpenAI Codex Mobile Update is INSANE!
What didn't work with the Dots agent
This is where the Dots agent fell over for me, and these aren't small things.
- The call button didn't work. Calling is one of the headline features. It would not answer a call.
- A simple avatar took 20 to 30 minutes. The request just sat there before it finally finished.
- A generated image wouldn't open. It said it was done. I couldn't view it.
- Slow start-up. The cloud computer took about two minutes just to start a simple research task.
- Where are the 4,000 apps? OpenAI talks about 4,000+ apps, but I couldn't find an obvious apps section. Dots itself suggested I use the Zapier MCP through the plugins section.
That last one tells you a lot. If the agent has to point you to a third-party connector to reach the apps, the "thousands of apps" claim isn't something you'll feel on day one.
Honestly, the whole thing felt like a version of Hermes Agent for beginners that's still not that smooth. It reminded me of Kimi Claw, which is basically a polished OpenClaw, and of OpenAI's earlier agent that launched big and then faded away. I'll be fair: Dots is in beta, and OpenAI can move fast. But you pay for what it does today, not what it might do in three months.
Announcement vs reality: the "book a table" test
OpenAI's own example was booking a table. So I tested exactly that. It worked, and I want to be clear about that, but it took around five minutes. Stranger still, the Dot opened the restaurant site's terms and conditions first before getting on with the booking.
Five minutes for a table you could book yourself in one. That's the gap between the announcement and the reality. The simple end of the pitch works, slowly. The "most ambitious work" end is harder to believe when a basic avatar takes half an hour and the call feature won't answer.
When I judge an AI agent, I care about three things: does it finish the job, how long does it take, and how often do I have to rescue it. That's the thinking behind my Goldie Bench tests. On all three, OpenAI Dots is behind the tools I already use.
Is OpenAI Dots worth it? The better alternatives
No, not yet. If you're already happy with another AI agent, don't switch to OpenAI Dots. And if you're new to agents, you'll get results faster elsewhere. Claude Sonnet 5.5 or Opus 5.5 with browser use is faster. Grok Bot felt smoother with its own computer. Hermes, now with its own Bot Screen, is more useful. Even Manus was better than this months ago.
| Tool | Why it's better than Dots right now | Full comparison |
|---|---|---|
| Claude (Sonnet 5.5 / Opus 5.5) | Browser use is faster and finishes tasks without the long waits | ChatGPT Dots vs Claude |
| Grok Bot | Its own computer felt smoother and more responsive | ChatGPT Dots vs Grok Bot |
| Hermes Agent | More useful day to day, now with its own Bot Screen | ChatGPT Dots vs Hermes |
| OpenClaw | The open agent that Dots most resembles, with more control | ChatGPT Dots vs OpenClaw |
If Hermes is the one you're curious about, see how its screen works in my Hermes Bot Screen breakdown. For the Claude side, here's what we know about the Sonnet 5.5 release date and what changed. And if you want to wire all of these agents into one system instead of juggling tabs, start with the Agent OS guide.
Not sure which agent fits your business? Book a free strategy session and we'll map out where AI agents can actually save you time, and which ones to skip.
ChatGPT Dots FAQ
What is OpenAI Dots?
OpenAI Dots is a personal, always-on AI agent inside ChatGPT, announced at DevDay on 29 September 2026. OpenAI says each Dot is powered by GPT-6 Astra, runs on its own cloud computer, connects to 4,000+ apps, learns from your feedback and keeps working while you're away. Per launch coverage, it can also call and message people for you.
Is ChatGPT Dots available in the UK?
Not on ChatGPT Pro at launch. OpenAI says the Pro rollout excludes the UK, Switzerland and the European Economic Area. I hit a region block myself after upgrading an account, so check availability before you pay for a plan.
Is OpenAI Dots worth it?
Not yet, in my testing. The website builder, take-over mode and research briefs were good, but the call feature didn't work, simple tasks took minutes, and an avatar took 20 to 30 minutes. Claude with browser use, Grok Bot and Hermes all did the job better. It's beta, so it may improve quickly, and I'll retest when it does.
Skip the beta headaches and use agents that already work
OpenAI Dots has a good idea at its core. An agent with its own computer that keeps working while you sleep is exactly where this is all heading. It's just not there yet, and you don't need to wait for it.
Inside the AI Profit Boardroom, 3,000+ members follow full courses on Claude Opus 5.5, Hermes Agent and Grok Bot, so you can set up an agent that finishes the job today. It's $69 a month locked in, normally $110.
Want help picking the right setup for your business? Grab a free strategy session and we'll build the plan together. Either way, you now know the truth about OpenAI Dots before you spend a penny on it.
