ChatGPT won't let you share your conversation? Error 413, explained

You clicked "Share", ChatGPT thought about it, and threw back Error 413. Nothing you did is wrong — the chat has simply grown too big to package into a link. Here's what that means and exactly what to do next.

What Error 413 actually means

413 is a standard HTTP status code — "Payload Too Large". When you create a share link, ChatGPT has to bundle up the entire conversation and hand it to its server to publish. If that bundle is bigger than what the server will accept in one request, the server refuses it and returns 413.

In plain terms: your conversation is too long to share. This is a limit on ChatGPT's side, not a bug you caused and not something broken on your machine. Retrying, switching browsers, or clearing your cache won't change it, because the size of the conversation is the problem, and that hasn't changed.

The frustrating irony: it's the biggest, most valuable threads — the long research sessions, the deep debugging, the projects you've been building for hours — that hit this wall. The conversations that most need rescuing are exactly the ones ChatGPT won't let you share.

What to do right now: write a handoff by hand

You don't need a share link to keep going. What you need is a handoff — a short reboot prompt that gives a fresh chat everything it needs to pick up where the cooked one left off. A fresh conversation is also faster and sharper, because it isn't dragging thousands of old tokens behind it.

Open a new ChatGPT chat, paste the template below, and fill in the four blanks from your old conversation. Keep it tight — you're capturing decisions and constraints, not transcribing the whole thing.

I'm continuing a conversation from another chat. Here's where we are — pick up from this and don't re-litigate settled points.

GOAL
<what we're actually trying to produce or solve, in one or two sentences>

LOCKED DECISIONS (don't reopen these)
- <decision 1 we already made and are keeping>
- <decision 2>
- <decision 3>

CONSTRAINTS I keep having to repeat
- <tone, format, tech stack, length, audience, things to avoid>

CURRENT STATE / WHERE WE STOPPED
<the last thing we were working on and the immediate next step>

Confirm you've got this, then continue from the next step.

That's the whole trick. Four fields — goal, locked decisions, repeated constraints, current state — carry roughly ninety percent of the useful context in a long thread. The fresh chat doesn't need the full history; it needs to know where you landed and what not to argue about again.

If you can still scroll the old chat

The conversation is still there — 413 only blocks sharing it, not reading it. So while you fill in the template, scroll back through the thread and copy out the parts that matter:

1. Scroll up to find the moments where a decision got locked in, and paste those into "Locked decisions".
2. Grab any constraint you found yourself repeating — a format, a tone, a rule the model kept forgetting — into "Constraints".
3. Copy the last working output or the exact spot you stopped into "Current state".
4. Skip the dead ends, the false starts, and anything you already abandoned. They only slow the new chat down.

Paste the filled-in handoff into a new chat and you're moving again — often faster than before, because the fresh thread isn't carrying all that weight.

Can Uncook do this for me?

Honest answer, so you don't waste a click: Uncook works from a share link today, and if ChatGPT returns 413 you can't create that link. So Uncook's share-link flow can't ingest a conversation that's already 413'd right now. For that specific cooked thread, the hand-written handoff above is your path.

Where Uncook helps: for the many conversations that are long and painful but haven't yet crossed the 413 line — the ones where sharing still works — Uncook reads the whole thing and writes the handoff for you, so you don't have to do the scrolling-and-copying by hand. It's the faster version of everything above.

And for the 413 case itself, a browser extension is on the way that captures the conversation directly in the page, so a broken share link won't matter — it'll read what's on screen rather than needing a link at all. We're not putting a date on it, but it's the future fix for exactly this problem.

Related: Continue a ChatGPT conversation after you hit the limit · How to hand off an AI conversation to a fresh chat