How long should a video ad be: the platform draws three of the lines for you

Google Ads Help states that a skippable in-stream ad can be skipped after 5 seconds, that non-skippable in-stream ads are 60 seconds or shorter, and that bumper ads run between 5 and 6 seconds (Google Ads Help, 2026). YouTube Help states that Shorts can be up to 3 minutes long (YouTube Help, 2026).

Those are not trivia for a specs page. Each one is a structural instruction.

The skip point is the only runtime you are guaranteed. Everything after second five is a runtime you have to earn, and the way you earn it is by having said something worth staying for before it arrives. If the claim has not landed by then, the rest of the ad is a file nobody watched.

A bumper is a different craft rather than a compressed one. Five or six seconds carries one idea attached to one brand. Squeezing a thirty second structure into it produces an ad that is technically complete and practically unreadable.

A ceiling is permission, not advice. Shorts allowing 3 minutes tells you what is accepted, and accepted is a long way from watched. Read every published maximum as the outer wall of the room, never as the recommended seat.

One more constraint sets length and nobody publishes it. If the ad has to work with the sound off, the captions carry the argument, and reading is slower than listening. A script that reads comfortably aloud in twenty seconds often needs twenty-five on screen.

The advice to cut is usually right. The reason people give for it is usually wrong, and the wrong reason is what makes teams cut the wrong seconds.

Cutting works when it removes runtime that was doing nothing. The setup before the claim. The second example proving the same point. The pause before the instruction. The branded opening that nobody asked for.

Cutting fails when it removes the proof, and it usually does, because the proof is the longest part of any honest ad. A claim takes two seconds. A demonstration that makes the claim believable takes eight.

You can watch this happen in a review. Someone says get it to fifteen, and what comes out is the demonstration, because it is the biggest single block. The ad now hits the number and convinces nobody, and the report will file it under creative did not work.

There is a category test underneath it. An obvious product with a visible benefit does not need thirty seconds and never did. A considered purchase with a mechanism to explain cannot be compressed into fifteen. It can only be made shorter and less understood, which reads identically in a dashboard and is not the same thing at all.

This is the trade-off stated as arithmetic, and once you see it you stop optimising for the wrong number.

Completion rate divides the people who reached the end by the people who started. Shorten the ad and the number climbs, whether or not the ad got better. A six second ad that sells nothing posts a completion rate a forty second ad cannot reach, and the metric is doing precisely what it was defined to do.

So completion rate is a diagnostic, not a target. It answers one question well: among ads of roughly the same length, which one is holding people. Compare a fifteen second ad's completion to a forty-five second ad's completion and you have compared two different questions and learned nothing.

Two ads of the same length can be judged on how far people got. Two ads of different lengths can only be judged on what happened after the view.

The real cost of the longer ad is simpler than the metric makes it look. More people leave before the instruction, so a smaller share of the money you spent reached the end of the argument. That is worth paying when the argument needs the seconds and wasteful when it does not, and only one of those two situations is visible in a completion column.

When a length target arrives from outside, cut in this order and stop as soon as you hit the number.

The honest answer to how long a video ad should be is that you find out, for your product, at your price, against your audience. Two lengths of the same argument. Same hook, same offer, same proof, different runtime, run against each other.

That answer is useless advice when the second length costs another booking. Which is why most teams have a house length rather than an answer: the second version was never affordable, so the question was never asked.

Cutroom is one way to make the second version cheap. Film one talking take of up to 3 minutes, and it transcribes every word, writes the hook, places b-roll on the exact words that need showing, times and styles the captions, trims silence and filler, and hands back a finished 9:16 MP4. Change the pace or delete a line on the transcript and the cut rebuilds around it, which is what makes a second runtime a decision rather than a project. A batch is 100 credits and export adds 20 credits per finished minute, so a 30 second ad is about 110 credits. Lite is $19.99 a month for 1,250 credits, about 11 of those, cancel in one click. Two limits worth naming before you plan around it: the source take stops at 3 minutes, and exports are 9:16 MP4 only.