Markdown is markup you can read with your own eyes, and the format language models read best. JSON is a data tree of objects, lists, numbers and strings. pandoc sits between the two, and this page says exactly what happens to an MD file on the way to becoming a JSON one.
What runs when an MD file becomes JSON
pandoc in two passes, on a machine we rent and watch. Your MD file is uploaded once, pandoc runs once, the JSON comes back, and neither file is kept. MD to JSON is one of the 3649 pairs that engine was probed on with a real file, which is why it has a page here and why the pairs the probe could not prove do not.
What survives from the MD into the JSON
Lists from the MD file, nesting included, plus code blocks, block quotes and links, all re-expressed in JSON.
Images: pandoc extracts them out of the MD file into a working directory and re-embeds them in the JSON, so nothing ends up pointing at a file that no longer exists.
What MD to JSON costs you
Tables. The JSON writer cannot carry one, and the probe caught this the hard way: a CSV file, which is nothing but a table, came back as a well formed and completely empty JSON with a zero return code. The engine now refuses that case outright rather than handing you the emptiness.
If a model is going to read the JSON
Structure is what survives from the MD file, and structure is what a model needs. MD in, JSON out, with the heading tree intact instead of flattened into one long paragraph.
MD to JSON, measured rather than promised
Running MD to JSON against the real engine with a real file, pandoc wrote 1,726 bytes of JSON in under a tenth of a second, machine otherwise idle. That is one MD file on one day and not an average, which is why the number is given together with the file that produced it.
The MD to JSON verdict was reached by a pandoc round trip, meaning the output was read back by pandoc and had to still contain a witness word planted in the source, which proves the content travelled and not merely the container. Which check was used matters, because they do not all prove the same thing, and a status code of 200 proves nothing whatsoever about whether the JSON file has anything inside it.
The same MD file, sent somewhere else
Other outputs the probe measured out of an MD file, so the cost of choosing JSON can be read against something. Sizes do not compare across engines, because each family was probed with its own MD witness file.
MD to PDF: 25,764 bytes in 1.9 seconds, verified by a format signature.
MD to DOCX: 10,106 bytes in 0.1 seconds, verified by a pandoc round trip.
MD to EPUB: 5,034 bytes in 0.1 seconds, verified by a pandoc round trip.
MD to TXT: 264 bytes in under a tenth of a second, verified by reading the text back.
MD to HTML: 4,018 bytes in 0.1 seconds, verified by a pandoc round trip.
MD to RTF: 687 bytes in under a tenth of a second, verified by a pandoc round trip.
Other ways into JSON, and what they measured
Among the published routes into JSON, MD is the fourth largest output of the 8 measured. The witness files differ, so this ranks the probe run and not your document.
ODT to JSON: 1,761 bytes in 0.1 seconds.
RTF to JSON: 1,743 bytes in 0.1 seconds.
TXT to JSON: 1,343 bytes in under a tenth of a second.
CSV to JSON: 110 bytes in under a tenth of a second.
YAML to JSON: 110 bytes in under a tenth of a second.
What we will not pretend about MD to JSON
An MD file over 25 MB is refused before the upload finishes rather than after it, so you do not wait for a rejection.
An MD to JSON run that passes 60 seconds is killed, and the pandoc process is killed with it. A run left behind would sit on one of the machine's two cores until somebody noticed.
32 of the 935 pairs probed in the family that serves MD to JSON failed, and this pair is not one of them. They fail for reasons worth knowing: a writer that cannot carry what the document is made of, or output the engine could not read back. They are counted here rather than hidden, because a pair that fails quietly is worse than one that fails loudly.
pandoc never writes the JSON through a TeX engine here, and it never writes PDF at all: the eleven engines it would need are not on this machine and a TeX distribution weighs several gigabytes. Anyone who wants a PDF out of an MD file goes through LibreOffice or calibre, both of which genuinely can.
One MD file at a time, chosen in the browser. There is nothing else to set up and nothing else on offer.
MD to JSON: what people ask
What actually converts my MD file to JSON?
pandoc does it, in two passes, on a machine we rent and watch. Not a browser trick and not somebody else service: the MD file is uploaded once, pandoc runs once, the JSON comes back, and neither file is kept afterwards.
How long does MD to JSON take?
On the file the probe used, pandoc took under a tenth of a second and wrote 1,726 bytes of JSON. That is one real measurement on one real MD file, not an average and not a promise about yours: a larger MD takes longer, and past 60 seconds the run is stopped.
What do I lose going from MD to JSON?
The one to know about first: Tables. The JSON writer cannot carry one, and the probe caught this the hard way: a CSV file, which is nothing but a table, came back as a well formed and completely empty JSON with a zero return code. The engine now refuses that case outright rather than handing you the emptiness.
Are the tables in my MD file still tables in the JSON?
No, and that is a property of JSON rather than a defect of this path. JSON cannot hold a table, so an MD file built around one is the wrong candidate for this pair.
Is MD to JSON free?
There is a free allowance every month, and one MD to JSON conversion costs half a credit against it. When the allowance runs out the tool says so and stops, rather than quietly handing you a worse JSON.