Free YouTube comment picker

Draw a giveaway winner from up to 500 recent top-level comments on your video, or up to 1,000 with your own YouTube Data API key. Duplicate entries are filtered out, and the receipt records a seed and entrant count so you can publish the fixed draw inputs for checking.

How to run a giveaway people trust

The complaint under every giveaway announcement is the same: how do we know you did not just pick your friend. A random picker on its own does not answer that, because nobody watched you use it and nothing stops you drawing twenty times until you like the name.

A seed answers it. The seed decides the outcome, so if you post the seed before you draw, anyone can run the same seed against the published entrant list afterwards and get the same winner. Re-rolling becomes visible, because the published seed would no longer match the name you announced.

1. Announce your entry conditions and a seed in the video or a pinned comment.
2. When entries close, paste the link here with that seed.
3. Publish the winner, the seed, and the entrant list CSV.

What counts as an entry

• One entry per channel is on by default, so twenty comments from one viewer count once. Channels are identified by their channel URL, falling back to the display name when the URL is missing.
• Required keyword excludes anyone who did not follow the instruction. It is a case-insensitive match anywhere in the comment, not a whole word — requiring done also admits I am done! and abandoned.
• Minimum words filters out one-character entries left by bots. Words are counted by splitting on spaces, so a Japanese or Chinese comment written without spaces counts as one word however long it is. Leave the minimum at zero for those.
• Replies are never counted. Only top-level comments are read at all, and there is no switch for it, because a reply is usually a conversation rather than an entry.

What the draw actually does

Six steps, in this order:

1. Read up to the ceiling of top-level comments, newest first, in batches of 100.
2. Sort them by comment id. This is the step that makes the draw reproducible: the entrant list stops depending on the order YouTube happened to return.
3. Drop comments that do not contain the required word, if you set one.
4. Drop comments under the minimum word count, if you set one.
5. If one entry per channel is on, keep the first comment from each channel in that sorted order. Sorting is by id rather than by time, so the comment kept is a stable choice rather than necessarily that viewer's first.
6. Shuffle what is left using the seed, and take the winners off the top.

One honest caveat about reproducing a draw. The seed fixes the shuffle, not the entrant list. If comments arrive or are deleted between your draw and someone else's check, their entrant list differs and so can the winner. That is why the receipt records the entrant count and the time, and why publishing the entrant CSV alongside the seed is what actually settles an argument.

What you get back

The draw receipt has four fields: the seed, the draw time in UTC to the second, the number of comments scanned, and the number of eligible entrants. Those are the four numbers worth pasting into your announcement.

The entrant list downloads as a CSV with three columns — position, author and comment_id — named giveaway-entrants-<video id>-<seed>.csv. It is UTF-8 with a byte-order mark, so channel names in any script open correctly in Excel.

Two things it deliberately is not. position is the row number in the eligible-entrant list, in the sorted order above — it is not the draw ranking, and row 1 is not the winner. And the file does not carry the comment text; if you want the entries themselves as evidence, run the same video through the comment exporter, which gives nine columns including the text and the like count.

A worked example

Sample data, not a real giveaway. Three rows of an entrant file:

position,author,comment_id
"1","Mara K","UgxAbC1"
"2","Field Notes","UgxDeF2"
"3","Tomas R","UgxGhI3"

Published next to it: seed 4f9c2ab71d0e5c83, drawn 2026-08-14T18:02:11+00:00, 412 comments scanned, 168 eligible entrants. An entrant with those five things can re-run the draw against the published entrant list and check the name you announced.

The limits, in numbers

The ceiling is the one limit that can change a result rather than just annoy you. Comments are read newest first, so a giveaway that drew more entries than the ceiling would be drawn from the most recent 500 — which is a different draw from the one you announced. The tool says when it hit the ceiling instead of drawing quietly; if it does, either add your own API key for 1,000 or say plainly which entries were in scope.

How it compares to the other pickers

Figures read from each tool's own page on 24 August 2026, and they will drift.

Where the others are ahead: Comment Picker has filters this does not — date range, like count, a cap on entries per person — and can draw across up to five videos on its paid tier. Hadzy publishes a proof page it hosts itself, which is easier for an entrant to open than a seed they have to re-run. Both are reasonable choices.

What is different here is the free ceiling, fifty winners without paying, and the entrant list as a file you hold rather than a page somebody else hosts. The longer comparison is in the Comment Picker write-up.

How the seed makes a draw checkable

Every draw records a seed, and the seed is what turns "trust me" into something a viewer can test. It is the starting value for the random selection: the same seed against the same entrant list always produces the same winners, and a different seed produces different ones.

That matters because the usual objection to a creator-run giveaway is not that the draw was rigged, but that nobody can tell either way. Publishing the seed and entrant CSV alongside the winner, together with the entrant count and draw time, gives a sceptical viewer the fixed inputs needed to re-run the draw.

The practical version is short. Announce the closing time before you draw. Draw once. Post the winner with the seed, entrant CSV, number of eligible entrants, and draw time. If someone disputes it, they have the fixed inputs needed to check.

Mistakes that make a giveaway look rigged

None of these require bad intent, which is why they are worth naming.

• Redrawing because the winner did not reply. Understandable, and it looks exactly like drawing until you get an answer you like. Say up front how long a winner has to respond and what happens if they do not.
• Changing the entry rule after entries are in. Adding "must be subscribed" after the fact both invalidates the draw and, if you enforce it, runs into YouTube's rules on making rewards conditional on engagement.
• Drawing from a partial list without saying so. If the comment count exceeded what was read, a draw from the most recent 500 is not a draw from everyone. This tool tells you when it hits that ceiling for exactly this reason.
• Announcing only the winner. A name with no entrant count, no time and no seed is unfalsifiable, which is a strange thing to aim for when the alternative is free.

The legal side varies by country and is worth checking separately — the rules that apply to YouTube giveaways covers the platform's own requirements, and proving a draw was fair goes further into the verification side.

Common questions

How do I prove a YouTube giveaway draw was fair?

Publish the seed, draw receipt and entrant CSV alongside the winner. The seed fixes the shuffle for that recorded entrant list, so re-running it against the same list produces the same winners. A later API read can differ if comments change.

Can I stop one person entering a giveaway many times?

Yes. One entry per channel is on by default, so a viewer who comments twenty times counts once. Channels are identified by their channel URL, and the comment kept is the first one in the sorted order the draw uses, which is a stable choice rather than necessarily their first in time.

Can I require entrants to comment a specific word?

Yes. Set a required keyword and a minimum word count, and comments that do not meet both conditions are excluded before the draw.

Does the comment picker need a YouTube API key?

No. It reads up to 500 comments free with no signup. Supplying your own YouTube Data API key raises the ceiling to 1,000, which matters for videos with very large comment sections.

Are replies counted as giveaway entries?

No, and there is no option to include them. Only top-level comments are read, because a reply is usually a conversation rather than an entry. If your entry mechanic depends on replies, this is the wrong tool for it.

How many winners can I draw at once?

Between one and fifty, in a single draw from one video. Drawing fifty needs no account and no payment.

Can I pick a winner from a YouTube Short?

Yes. Paste the /shorts/ link and it is handled the same way as a watch page, because a Short is an ordinary video as far as the API is concerned. Live and embed URLs work too, as does a bare video id.

What is in the entrant list file?

Three columns: position, author and comment_id. Position is the row number in the eligible-entrant list, not the draw ranking, so row one is not the winner. The file does not carry the comment text; export the video through the comment exporter if you need the entries themselves.

Is this allowed under YouTube's rules?

Reading public comments through the official API is fine. The rules you need to read are the ones about the contest itself — YouTube requires that you run it lawfully and make clear it is not sponsored by YouTube. That is covered in what YouTube comment automation covers and in the write-up on what gets channels banned.