Choose the sample space before the output
These tools do not all randomize the same kind of thing. A letter or word draw takes entries from a finite collection without replacement; a password treats each character position as an independent draw; a UUID v4 fixes format bits and randomizes the remaining 122 bits. A date or time is selected from an inclusive sequence of calendar days or minutes. The useful first question is therefore what may be selected, not how many times to press the button.
A small example of different rules
Three letters drawn from A through Z have 26 × 25 × 24 ordered outcomes because a chosen letter is removed. Three password characters from a 63-character set have 63³ outcomes because repetition remains possible. A two-way matrix tie gives each tied leader probability 1/2, but an option below the maximum has probability zero. Similar-looking random outputs can answer materially different questions.
Match the check to the result
For a date, count the days between approved endpoints and check that the displayed date is inside them. For a UUID, inspect the 4 version marker and the variant nibble. For a prompt, verify that it contains its character, action, setting, and constraint. For a group decision, retain the candidate list and scoring rule. These checks show whether the stated mechanism was followed; they do not make a result suitable for every purpose.
Know where a randomizer ends
Random selection can remove a visible preference from an eligible set, create a creative constraint, or make a fresh technical identifier. It cannot assess consent, safety, factual accuracy, account security, or the quality of a decision matrix. Put those human and domain rules in place before the draw, then save the input when the result matters beyond a game or private exercise.