Yes, ChatGPT can read JSON files you upload, and can parse and analyse them using its code interpreter. However, large files such as your own conversations.json export usually exceed the context window or upload limits, so converting them first is more reliable.
Short answer
Yes, ChatGPT can read JSON files you upload, and can parse and analyse them using its code interpreter. However, large files such as your own conversations.json export usually exceed the context window or upload limits, so converting them first is more reliable.
The short answer is yes. The useful answer depends on what the JSON is and how big it is, because the two ways ChatGPT can handle a file behave very differently.
Two mechanisms, often confused
Reading it into context. The file's text is placed in the conversation, and the model reads it like any other input. Limited by the context window. Good for small files where you want the model to reason about the content.
Analysing it with code. ChatGPT writes and runs Python against the file in a sandbox. The file itself never enters the context window, only whatever the code prints. This handles far larger files, since the constraint becomes upload size and execution time rather than tokens.
Which one you get depends on file size and how you phrase the request. Asking to "read" a file nudges toward context; asking to "analyse" or "parse" nudges toward code.
Practical limits
Upload limits sit around 512 MB per file, but that is rarely the binding constraint. The real ceilings are the context window for direct reading, and sandbox memory and time for code analysis.
As rough guidance, JSON up to a few hundred kilobytes is comfortable in context. Into the megabytes, code analysis becomes necessary. At tens or hundreds of megabytes, even code analysis becomes slow and often fails on memory.
Why your ChatGPT export usually fails
People frequently try uploading their own conversations.json and asking ChatGPT to summarise their history. This rarely works, for three compounding reasons.
Size. A moderately used account produces 40 MB to 300 MB. That is far past the context window and pushes the sandbox hard.
Structure. The file is not a list of conversations. Each conversation is a tree of message nodes linked by parent and child references, because editing a prompt or regenerating a response branches the thread. Code written against the naive assumption of a flat list produces duplicated messages from abandoned branches and wrong ordering. ChatGPT frequently makes exactly this assumption.
Density. Even when parsing succeeds, the useful text is buried under metadata, which wastes whatever context is available.
The result is usually a confidently wrong summary, which is the worst outcome.
The reliable approach
Convert the export into readable documents first, then work with those.
ChatExports reads conversations.json, walks each tree along its active path, and writes per-conversation files in Markdown, plain text, Word, PDF, or CSV. Processing runs locally in your browser, so the file is never uploaded anywhere.
With that done, several things become possible that were not before. You can upload individual conversation files to ChatGPT, which are small enough to sit comfortably in context. You can upload the CSV and ask for analysis across conversations, which is a natural fit for code interpretation. And you can search the whole archive locally with proper tools rather than hoping a model summarises it correctly.
Getting good results from JSON uploads generally
For JSON files that are a reasonable size, a few things help.
Say what the structure is rather than making the model infer it, especially for nested or unusual shapes. Ask for analysis rather than reading when the file is more than a few hundred kilobytes, so it uses code. And ask it to show the parsing code, so you can check the assumptions it made about structure before trusting the output.
Frequently asked questions
What size JSON file can ChatGPT handle?
Up to a few hundred kilobytes reads comfortably into context. Larger files need code analysis, which handles megabytes but struggles past tens of megabytes.
Why can ChatGPT not summarise my conversations.json?
It is usually too large for the context window, and its tree structure means naively written parsing code produces duplicated and out-of-order messages. Convert it to documents first.
Can ChatGPT edit a JSON file and give it back?
Yes, using code interpretation it can modify a file and return the result as a download, provided the file is small enough to process in the sandbox.
Is it safe to upload JSON to ChatGPT?
Uploaded files go to OpenAI's servers and are subject to your data controls settings. For your own chat export, converting locally avoids sending the whole history back to OpenAI in a new form.