YouTube Ad Formats: What the Creative Must Do Before the Skip

The formats, as Google defines them

Google Ads Help publishes the format definitions; the rows below restate them. Two things in the table decide the creative: where the viewer regains control (the skip point, the swipe, or never), and whether the unit is bought for a view or for an impression, because the first sets what the opening must do and the second sets what the whole piece is for.

The five-second audition

A skippable in-stream ad is two ads. For the viewers who will skip, the first five seconds are the whole ad; for the viewers who stay, they are the opening. Both audiences have to be served by the same five seconds, which means the opening cannot be a build-up. It has to carry the brand or the product visibly, one idea a viewer can restate, and a reason to keep watching that the next fifteen seconds then honour. The reason is the part most spots leave out: a question the rest answers, a demonstration begun, a claim the proof follows. Sound is usually on in this placement, unlike the feed, so the opening line matters as much as the opening frame, but the frame must still work on a television across a room and a phone in a hand, which sets the size of type and product in it.

Six seconds, one idea

A bumper is not a cut-down. Six seconds holds one idea, and a spot written for thirty holds several, so a bumper cut from a longer edit keeps the fragments of an argument and none of the argument. Write the bumper as its own script: the brand visible from the first frame, one image or one line the viewer takes away, sound and picture saying the same thing, and an end card that is part of the idea rather than a signature after it. The discipline pays back into the longer cuts: a concept that cannot be said in six seconds is usually a concept that has not been found yet.

Non-skippable: earn the time you were given

A non-skippable spot removes the audition and adds an obligation. The viewer cannot leave, so the opening does not have to stop a skip, but the whole piece is watched by people who did not choose it, and a spot that wastes their fifteen seconds is remembered for that. Structure it as a complete story with a beginning that sets the situation, a middle that shows the product doing the one thing the ad is about, and an end that says the name and what to do. The 16 to 30 second unit on connected TV is a living-room commercial: no clickable call to action exists on the screen, the response has to live in the picture and the voice, and the technical delivery follows the television specification set out in connected TV commercial specifications.

Shorts is a different brief

Shorts behaves like a vertical feed, not like in-stream: full-screen 9:16, sound often on but never guaranteed, and control returned to the viewer with a swipe from the first frame. The creative that belongs there is the creative that belongs on TikTok and Reels: a first frame that is already the point, on-screen captions, a person or a product in motion inside the first second, and a length set by the idea rather than the maximum. The same concept as the in-stream spot can live here, but as a new cut composed for the vertical frame with the interface's safe areas respected, not a crop of the 16:9 master. Unreal's short-form video production page describes that work.

Skip point

The moment a viewer regains control of the screen: five seconds into a skippable in-stream ad, the first frame of a Shorts ad, never in a bumper or a non-skippable spot. The opening of a piece is written to what happens at its skip point.

Bumper

Google's non-skippable format of up to six seconds, bought on impressions. One idea, brand from the first frame, written as its own script rather than cut from a longer one.

In-stream

Ads that play before, during or after another video, skippable after five seconds or non-skippable to their end. The viewer came for the video, not the ad, which is the premise the creative starts from.

In-feed

A thumbnail and headline in search results, Watch Next or the Home feed that the viewer chooses to click. The thumbnail and the first line are the whole ad until then.

Master and cut

The master is the longest, fullest version of a concept; a cut is a shorter or differently framed version made from the same concept. A bumper and a Shorts version are cuts only if they were planned as such; otherwise they are new pieces.

One brief, four cuts

The efficient way to cover YouTube is to plan the set before anything is produced: which cut each format gets, what each keeps and what each drops. Planned that way, the six-second and vertical versions are designed, not salvaged, and the assets they need (a clean product hero, an end card that works at every length, captions as a separate layer) are produced once.

Reading the result

Each format gives a different read. On a skippable spot the first question is how many viewers chose to keep watching past the skip point, and the retention curve around that point says whether the opening made its promise; the second question is what the viewers who stayed then did. On a bumper and a non-skippable spot there is no choice to read, so the piece is judged on the outcome the campaign was bought for and on the brand measures the account tracks. On Shorts the reads are the feed's: the swipe, the watch time, the response. None of those reads is a promise about the next piece; each is the brief for it, which is the method set out in the creative iteration framework.

What this guide does not tell you

It does not say which format a campaign should buy; that is a media decision made from the objective and the budget. It does not publish view rates, retention figures or any performance result: Unreal publishes no performance statistics and never guarantees advertising performance. And it does not replace Google's documentation: the length rules quoted here are Google's on the review date and are verified again at upload.