Simon Willison, an independent AI researcher, documented new ChatGPT Work features on X in late March 2026. The mobile and web app feature now includes a browser with screenshot capability and can deploy web apps to Cloudflare Workers, branded 'ChatGPT Sites'.
Key facts
- ChatGPT Work adds browser and screenshot tools
- Deploys web apps to Cloudflare Workers
- Branded as 'ChatGPT Sites'
- Documented by Simon Willison on X
- Available in mobile and web apps only
ChatGPT Work — a feature in OpenAI's mobile and web apps, distinct from the same-named mode in the desktop app — has gained two capabilities that blur the line between conversational AI and a full development environment. According to @simonw, it now includes a built-in browser that can take screenshots, and it can deploy web apps to Cloudflare Workers, a capability branded 'ChatGPT Sites'.
The browser integration is notable because it lets the model interact with live web pages, capture visual state, and presumably reason about what it sees. This is a step beyond text-based browsing, which ChatGPT has supported since the Browse with Bing feature launched in 2023. The screenshot capability suggests the model can now perform visual QA or inspect rendered output directly.
The Cloudflare Workers deployment is the more significant feature. It turns ChatGPT from a chat interface into a deployment pipeline. A user can describe a web app, have it generated, and push it to a live URL on Cloudflare's edge network without leaving the app. This is a direct challenge to platforms like Vercel's v0 and Replit's Agent, which already offer AI-driven app generation with one-click deployment. OpenAI's move into this space signals that it sees deployment as a core part of the AI-assisted development workflow.
Willison, who runs the popular Datasette project and is known for documenting AI tooling, noted the confusion between the mobile/web ChatGPT Work and the desktop app's mode with the same name. The desktop mode is a different product, and the naming collision could confuse users.
The company has not disclosed pricing changes or API access for these features. The deployment capability appears to be a consumer-facing feature, not a developer API. This means developers who want to automate deployments programmatically will still need to use Cloudflare's own tools or the OpenAI API directly.
The browser and deployment features are part of a broader trend where AI assistants are becoming agents that can act on the web, not just generate text. OpenAI's move aligns with Anthropic's computer-use feature, which lets Claude control a desktop browser, and Google's Project Mariner, which does the same in Chrome. The difference here is that ChatGPT Work deploys code, not just interacts with pages.
For now, the features are available in the mobile and web apps. It is unclear whether they will come to the desktop app or the API. The naming collision between the two ChatGPT Work modes suggests OpenAI may be consolidating features over time.
Watch for whether OpenAI extends these capabilities to the API or the desktop app. If the browser and deployment tools become available programmatically, they would enable a new class of autonomous agents that can build, test, and ship web apps end-to-end. That would be a significant step beyond current AI coding assistants.
Key Takeaways
- ChatGPT Work adds a browser with screenshots and deploys web apps to Cloudflare Workers, per Simon Willison.
- The deployment capability, branded 'ChatGPT Sites', positions OpenAI against Vercel and Replit.
What to watch
Watch for OpenAI's next developer update to see if the browser and Cloudflare deployment features are exposed via the API or the desktop app. If they are, expect a wave of autonomous web-app-building agents that use ChatGPT as the orchestration layer.
