What CSV is
CSV is a text file where each line is a row and commas separate the fields. There is no type system, no formatting and no second sheet. Every value is text until something downstream decides otherwise.
Almost every database, analytics tool and spreadsheet can read and write it, which is why exports default to it.
What JSON is
JSON is a text format for nested data: objects with named keys, arrays, strings, numbers, booleans and null. It has no date type and no comments.
Every web API, most configuration files and anything that talks to JavaScript.
What changes when you convert CSV to JSON
Reading CSV means deciding three things the file does not state: which character separates fields, how quotes work, and what encoding the bytes are in. Tablizer sniffs the delimiter from the first few lines, handles quoted fields containing commas and newlines, and reads the file as UTF-8.
You choose the shape. An array of objects is the default and the one most code expects. A 2D array is smaller. Column arrays suit charting libraries. A keyed object suits lookups by id.
Flat rows become records. Each header becomes a key, and each row becomes one object holding those keys. A CSV with columns id, name and role produces objects with an id, a name and a role, in that order.
Blank cells need a decision, because CSV has one way to say nothing and JSON has three. An empty field can become an empty string, a null, or an omitted key. Empty string is the default because it round-trips: convert back to CSV and you get the file you started with.
Duplicate headers would collide as object keys, so the second one is suffixed. Two columns both called name become name and name_2 rather than one silently overwriting the other.
What carries over from CSV to JSON
CSV records 2 things about a table that JSON has no way to hold.
JSON wants a type for each column, and CSV does not record one, so each column is typed from what its values look like. A column of digits that should stay text - a zip code, a phone number, a leading-zero id - is the usual thing to check afterwards.
JSON has no header row. The column names are repeated as keys on every record instead, which is why the result is bigger on disk than the table it came from.
CSV can be read one record at a time. JSON has to be complete before it means anything, so the whole table is held in memory while it is written and a file that gets cut off part way through is not partly usable - it is unusable.
The identifier 007 comes out of the JSON as 007. Reading it as a number would have made it 7, and an id that changes value is worse than one that stays text.
The role "Analyst, data" survives with its comma, in one cell rather than split across two. That is the first thing to check in any converted table, and the usual place a JSON file goes wrong.
The quotes around Jonah "Jo" Pryce are escaped with a backslash in the JSON.