Please preserve regular Chat with live webpage context after Atlas

ChatGPT Feature requests
1

Atlas is being discontinued, but none of the currently proposed replacements preserve one of its most important and distinctive capabilities:

the ability to browse the web manually while having an ordinary ChatGPT conversation that can understand the active webpage.

I am not asking for an agent to operate the browser on my behalf.

I am not asking for a Codex workflow or an isolated Work task.

I am asking to preserve a different interaction model:

  1. The user navigates the web normally.

  2. ChatGPT can read the active page when permission is granted.

  3. The user can discuss that page in a regular Chat conversation.

  4. The conversation can continue naturally across follow-up questions and, ideally, inside an existing Project.

Why the current replacements are not equivalent

The new desktop application and the Chrome sidebar route browser interaction primarily into Work or Codex-style workflows.

Those workflows may be useful for delegating tasks, but they are not equivalent to an ordinary ChatGPT dialogue.

In Atlas, I can remain in control of the browsing process while using ChatGPT as a thinking partner beside the page. I can read, question, compare, challenge, interpret, and gradually develop an understanding of the material.

This is fundamentally different from asking an agent to complete a predefined task.

Pasting a URL into ChatGPT is also not an adequate replacement because many pages are:

  • authenticated;

  • dynamically generated;

  • dependent on their current UI state;

  • accessible only within a personal account;

  • meaningful in relation to what is currently visible or selected.

Uploading screenshots or copying text manually breaks the flow of research and loses much of the live page context.

Why this interaction model matters

Atlas supports a valuable human-in-the-loop mode of working:

  • the human retains agency and decides where to navigate;

  • ChatGPT provides interpretation and reasoning;

  • the webpage becomes shared conversational context;

  • understanding develops through an ongoing dialogue rather than an isolated task.

This is especially important for research, education, technical documentation, complex interfaces, long-form reading, professional analysis, and any workflow where the goal is not merely to complete an action but to understand something deeply.

It may not be the most common use case, but it is a unique capability that ordinary browsers and ordinary ChatGPT do not provide separately.

Request

Please consider at least one of the following:

  1. Continue supporting Atlas, even if only for existing users or in a reduced maintenance mode.

  2. Add live active-page context to regular Chat mode in the new ChatGPT desktop application.

  3. Allow the Chrome sidebar to use ordinary Chat conversations, not only Work or Codex tasks.

  4. Allow users to attach the current browser tab to any existing Chat or Project.

  5. Preserve explicit permission controls so that webpage access remains transparent and user-controlled.

The essential requirement is simple:

Let users browse manually and talk with ordinary ChatGPT about the page they are currently viewing.

Please do not replace this collaborative thinking experience entirely with agentic task execution.

Atlas demonstrated that this interaction model is possible. Losing it would remove an important and genuinely innovative way of communicating with ChatGPT.

I hope OpenAI will clarify whether this capability is planned before Atlas becomes unavailable on August 9.

Related community requests

Other users have raised closely related concerns in topics titled:

  • “Please let us keep using Atlas — ChatGPT is not the same experience”

  • “Exploring alternatives to Atlas for continuous web browsing context”

  • “Please preserve the Atlas Project workflow in the ChatGPT Chrome extension”

These requests describe different parts of the same underlying need: preserving ordinary ChatGPT conversations with live webpage context, rather than limiting browser access to isolated Work or Codex tasks.

The topics can be found through the community search using the titles above. I will add direct links when my account permissions allow it.

2

Additional findings after testing the suggested alternatives

Following the reply from OpenAI Support, I tested both suggested replacements myself:

  • the new ChatGPT desktop application;
  • the ChatGPT Chrome extension.

Unfortunately, neither reproduces the Atlas workflow described in this request.

The desktop application contains both Chat and Work modes.

Interestingly, the application already includes a built-in browser panel even while using regular Chat.

However, when Chat is active, the application explicitly states that browser context is only available in Work mode.

If I choose to remain in Chat, ChatGPT cannot see the currently open page.

So the limitation is not that browser context is technically unavailable.

It appears to be intentionally restricted to Work mode.

This is an important distinction because the request is not about browser access in general.

It is specifically about preserving browser context inside ordinary Chat conversations.

From my perspective these are two fundamentally different interaction models:

  • Work → delegate a task to an agent.
  • Chat → think together while the human remains in control of browsing.

Atlas supported both.

The current implementation supports only the first.

I have confirmed this behaviour by testing both the desktop application and the Chrome extension after receiving the support response.

3

Totally support this. Have you heard anything else from OpenAI in this respect? Today is the last day for Atlas, and things remain the same unfortunately.

4

Unfortunately, no.

I contacted OpenAI Support, escalated the request to a human specialist, and they confirmed that my feedback had been reviewed and shared internally. However, they couldn’t provide any roadmap information or confirm whether this capability is planned.

Since then, I’ve also tested the latest desktop app and Chrome extension again.

The situation remains the same: browser context is available only in Work/Codex workflows. Regular Chat still cannot access the live active webpage, which means existing conversations and Projects cannot continue seamlessly while browsing.

I still hope OpenAI will reconsider this distinction. In my opinion, Work and Chat solve different problems, and Atlas uniquely supported a collaborative browsing workflow that is currently missing.

5

Update: regular Chat now appears to have live webpage context

Good news — the situation has changed.

I tested the latest ChatGPT desktop app again, and regular Chat can now see and discuss the webpage open in the built-in browser without requiring a switch to Work mode.

This is exactly the capability that was missing when I originally created this thread.

Even better, based on my testing so far, this appears to behave as ordinary Chat rather than as a Codex/Work session, so it does not seem to consume the separate Codex limits.

If this is indeed the intended behavior, then OpenAI has effectively preserved the most important Atlas workflow:

  • the user navigates the web manually;

  • ChatGPT shares the live webpage context;

  • the conversation remains a normal Chat rather than an agentic Work task;

  • the user can continue reasoning and discussing the page without handing control of browsing to an agent.

That is a very welcome change, and I want to explicitly thank the team for implementing it — especially right around the time Atlas is being retired.

There is still one important gap:

the Chrome extension/sidebar still appears to route browser-context conversations through Work/Codex rather than regular Chat.

So while the desktop app now seems to preserve the core Atlas interaction model, the Chrome extension does not yet provide the same continuity with ordinary ChatGPT conversations and Projects.

Still, this is a major improvement, and it addresses the central concern raised in this thread.

Thank you, OpenAI, for listening.

6

Thanks for sharing. Not working for me yet, though. Even after updating the app to the latest version. Hopefully, that’s only because they’re still rolling it out.

The chrome extension/sidebar issue is still as you described it unfortunately.

Cheers

7

Correction to my previous update

Unfortunately, I need to correct my previous post.

After additional testing, I confirmed that regular Chat in the new desktop app does not actually have access to the content of the webpage open in the built-in browser.

It can see that a browser tab is open and has access to its URL, but it cannot read or reason over the page itself.

I initially interpreted the new behavior as live webpage context being available in regular Chat, and I apologize for the confusion — especially to @seba.d, since my previous reply suggested that the main issue had been resolved.

So the original problem remains:

Regular Chat still cannot access the live content of the page the user is manually browsing.

To obtain actual webpage context, the app still requires switching to Work mode. The Chrome extension likewise continues to route browser-context interaction through Work/Codex.

In other words, the core Atlas workflow is still missing:

manual browsing + ordinary ChatGPT conversation + live webpage content as shared context.

My original feature request therefore still stands.