ChatGPT Conversation Summary
Paste a public ChatGPT share link and get an instant statistical overview of the whole conversation: how many messages each side sent, total words and characters, an exact o200k token count, reading time, and how many times the chat branched from edits and regenerations. Everything is computed in your browser — nothing to install, no signup.
The stats are computed in your browser — nothing is logged or saved.
This tool works only with public share links — the chatgpt.com/share/… URLs created with ChatGPT's Share button. Regular chatgpt.com/c/… chat URLs are private to your account and can't be read here.
Why Analyze a Conversation
Full stats in one paste
One link in, a complete overview out: message counts per side, words, characters, reading time — no scrolling through the chat, no manual counting.
Real BPE token count
The token tile runs the actual o200k_base tokenizer — the GPT-5 / GPT-4o encoding — in your browser, not a length estimate. Pair it with our Token Counter & Cost Calculator for arbitrary text.
Branch detection
Edits and regenerations fork a conversation into a tree. The summary counts those forks — a view of the chat's history that ChatGPT itself never shows.
Reading time
Total words at a 200-wpm reading pace tells you at a glance whether a shared conversation is a two-minute skim or a half-hour read.
What the Numbers Mean
The counts cover the transcript the share link actually displays: the active branch of the conversation, with ChatGPT's hidden tool and system turns filtered out — the same messages you see when you open the link. Words are counted per message and summed, characters count the raw message text, and reading time divides the word total by a 200-words-per-minute pace.
| Number | Read it as |
|---|---|
| Messages | Turns on the displayed path — not every message the conversation ever held. |
| Words / characters | The transcript as text. Useful for sizing a write-up you are about to lift out of the chat. |
| Tokens | What a model reads and an API bills, via the real o200k_base encoding. Roughly 4 characters of English each — paste the text into the Token Counter to turn the number into money. |
| Branch points | How many times the conversation forked — one fork per edited message or regenerated reply. A high number means much of the thinking is on paths the share link doesn't show. |
| Reading time | Words ÷ 200. A 30-minute chat with a model is often a 4-minute read once the back-and-forth is on one page. |
Frequently Asked Questions
What are branch points?
Every time you edit a message or regenerate a reply, ChatGPT keeps the old version and starts an alternate branch — the conversation becomes a tree, and the share link displays only the active path. The branch-points count is the number of places where that tree forks, so it shows how much editing and regenerating happened behind the transcript you see.
Is the token count exact?
For OpenAI's current models, yes — it runs the real o200k_base BPE tokenizer in your browser, the same encoding the GPT-5 family and GPT-4o use, not a characters-divided-by-four estimate. It counts the visible transcript text; ChatGPT's own context also includes system messages and formatting you can't see, so actual API usage would run slightly higher.
Can it read my private chats?
No. It works only with public share links (chatgpt.com/share/…) created with ChatGPT's Share button. A regular chatgpt.com/c/… chat URL is private to the account that owns it and cannot be read by this tool — or by anyone else without your login.
Is anything stored?
No. The share link is fetched once through our server proxy (browsers can't request chatgpt.com directly), the stats are computed in your browser, and nothing is logged or kept. Close the tab and every number is gone.
