Short answer: pick the fields, set a row count and a seed, then copy the JSON or CSV. The same seed always regenerates the same rows, and every address is an example.com one so nothing in the output can reach a real inbox.
What “seeded” buys you
Test fixtures and screenshots need stable data: a table that reorders on every reload makes a visual diff useless. A seeded generator produces the same values every time, so the fixture can be committed and reviewed like code.
Where the line is
This is prototype data. It has no relationships — no order belongs to a customer, no date follows another — because inventing plausible joins is a database problem, not a text problem. When a prototype needs real relationships, generate the rows here and wire the links in code.