Upload up to 3 sources — PDFs, Word docs, slides, spreadsheets, text files, Google Drive files, or a URL — pick the JSON Export tile, and Scholarly generates a .json file with consistent keys and records in a top-level array, saved to your library and ready to download. The Free plan includes one lifetime AI creation and one upload (8MB or 32 PDF pages). Premium ($30/mo or $144/yr) allows 10 AI creations per week and Laureate ($99/mo) allows 80.
Asking a chatbot for JSON usually returns a code block with an explanation above it, a caveat below it, and key names that drift from one answer to the next. Every one of those needs to be stripped or normalised before a parser will accept it. The JSON Export tile produces the file, not a message about the file.
Because the generator reads your uploaded sources, the records are drawn from your document — the clauses in your agreement, the SKUs in your catalogue, the studies in your review. Where the document is silent, the field is left empty so the gap is visible in the data rather than filled with a plausible guess.
The result is a real .json file in your Scholarly library, stored next to the source it came from. Download it into a script or pipeline, and when the source changes, generate again with the same instructions to get the same shape.
The JSON tile is deliberately narrow. These are the boundaries to plan around.
Developers and analysts use the JSON Export tile to get documents into pipelines: vendor reports into a warehouse, contract obligations into a tracker, product specs into a catalogue service, policy documents into a searchable index. Name the keys once and every document in the same family comes out in the same shape.
Teams plans add shared libraries, so the source documents, the generated JSON, and the SOPs that describe the process stay together where the whole team can reach them.