I’m trying to determine whether anyone else has seen a similar issue with older, very long archived ChatGPT conversations.
I have a number of archived conversations that had previously reached the conversation limit. Around February/March 2026, many of them became partially inaccessible.
The behavior differs by client. On Web/Android, affected conversations show "The conversation is too long, please start a new one.”
On iOS, some of the same conversations can still appear through Search and may open there, but unarchiving fails every single time.
The affected conversations still appear in my data export, so they do not appear to be fully deleted. The issue looks more like an inconsistent conversation/archive state, metadata, or loading problem affecting older long conversations.
My goal is not to continue these conversations. I’m just trying to access the existing history so I can copy/save needed information and then clean up/delete the archived chats.
The export only helps partially, because it does not fully replace the in-product conversation view, especially where edited/regenerated paths inside the same conversation matter.
Has anyone else run into older long archived conversations becoming inaccessible or impossible to unarchive like this? If so, were you able to recover access, or was export the only usable workaround?
Thanks for sharing the detailed write-up, @Mattaro. Also appreciate @EricGT pointing out the related topics below this post. They're worth a look since a few describe very similar behavior.
From what you've described, this doesn't sound like the conversations are necessarily gone, especially if they're still present in your data export. It seems more like an issue with how some older, very long archived conversations are being loaded or how their archived state is being handled across different clients.
We'll keep an eye on this to see if others can confirm the same pattern or share any workarounds beyond using the export. If anyone has recovered access to these older archived chats, it'd be helpful to hear what worked.
The short version: let AI help convert them as needed.
I have not tried to recover older archived/exporeted ChatGPT conversations for use again inside ChatGPT. However, for several months I have been archiving the .codex directory for certain projects and then converting those files to text so they can be reviewed with other tools. Also doing the same for .claude directory.
I cannot say whether I have seen changes specific to OpenAI’s archived chat format, but I do recall that the formats are not always satable across tools. What helped was using AI to first create a recognizer that could identify the format, and then having AI create a parser or converter for that format.
So, if the archived data is available but not directly usable, one practical approach may be:
Preserve the original files unchanged.
Ask AI to inspect a small sample and identify the structure.
Have AI generate a recognizer for the different variants.
Then have it generate a parser or converter to extract the conversation text.
That approach worked well enough for my Codex archives, though I cannot confirm that it will work the same way for older archived ChatGPT conversations.
