There is no single best YouTube ad length. Placement and objective decide it: 6 second bumpers for unskippable reach, 15 to 30 seconds for almost anything asking for a click or a purchase, and longer cuts only where retention holds up. The one fixed deadline is the skip button on in-stream, which appears at 5 seconds, and everything about a skippable script should be built backwards from it.
Which YouTube ad length should you use?
Length is a function of two variables, not one. What placement is this running in, and what is it supposed to achieve. Pick the row, not the number you liked last time.
| Placement | Objective | Length we ship | Why that length |
|---|---|---|---|
| Skippable in-stream | Sales or leads | 20 to 30 seconds | Room for claim, evidence and ask after the skip has been survived |
| Skippable in-stream | Consideration | 30 to 60 seconds | Watch time is the goal, so let the argument breathe |
| Non-skippable in-stream | Reach with a message | 15 seconds | The full 15 is guaranteed, so use all of it |
| Bumper | Reach and recall only | 6 seconds | One idea, one visual, no offer |
| Shorts | Direct response | 15 to 30 seconds | Vertical, scroll-driven, needs an evidence beat |
| Shorts | Retargeting | 10 to 15 seconds | They already know you, skip the setup |
| In-feed video | Click to watch | 20 to 45 seconds | The click is a choice, so the payoff has to be there |
The pattern worth noticing: the two placements with the most volume, skippable in-stream and Shorts, converge on 15 to 30 seconds while getting there for opposite reasons. In-stream needs enough runway to justify the viewer's decision not to skip. Shorts needs to be short enough that the thumb never gets bored. Same numbers, different logic, and it changes what goes in second three.
What has to happen before the 5 second skip point?
On skippable in-stream you are effectively buying a 5 second ad, with an option on the rest. Everything you say after second five is only heard by people who chose to stay. So the first five seconds have exactly three jobs.
State who this is for. Not the brand, not the mood, the audience. "If you run a warehouse with more than ten vans" disqualifies most of the impression and keeps the person who matters.
Make one specific claim. A number, a timeframe, a price, a comparison. Vague openings get skipped because there is nothing to be curious about.
Show the product. Not a logo card, not a title sequence. The thing itself, in use, so the viewer knows what category they are in before they decide.
What does not belong in the first five seconds: a logo sting, a slow establishing shot, a name introduction, or a question the viewer has no reason to answer. Median hook rate on YouTube in-stream sits at 22 percent according to the platform medians in our 2026 fatigue benchmark, lower than Meta Reels at 28 percent, and median video CTR is 0.42 percent. Those numbers say the skip is aggressive and that most of your budget is spent on people who leave. Structure accordingly.
Where completion rate fits
Once someone stays past five seconds, completion rate becomes the thing to watch, because it tells you whether the middle of the ad is doing work or drifting. A 30 second cut with strong view rate and poor completion usually has a sag somewhere between second eight and second fifteen, which is where writers put the part they were reluctant to cut. Trim there before you touch the opening.
Why a 6 second bumper cannot carry an offer
Bumpers are genuinely useful and constantly misused. Six seconds, unskippable, cheap on a CPM basis, and completely unable to carry a direct response argument. The arithmetic is simple: a claim takes roughly two seconds to land, evidence takes another two to three, and a call to action needs about two. That is already over budget before you have shown the product.
What a bumper does well is one idea repeated into memory. A single product shot with a single price. A category association. A reminder for someone already in your retargeting pool. What it does badly is a discount code, a multi-step offer, a feature list, or anything requiring the viewer to remember a URL.
The failure mode we see most often is a 30 second script compressed to 6 seconds by speeding up the voiceover. You end up with an ad that is technically legible and persuades nobody, and because bumpers are unskippable it looks fine in reporting until you check what happened downstream. If your only measurable objective is purchases, bumpers are the wrong buy, and it is more honest to say that than to pretend a length can be optimised into a different job.
How do Shorts and in-stream reward opposite things?
| Skippable in-stream | Shorts | |
|---|---|---|
| Frame | 16:9, 1920x1080 | 9:16, 1080x1920 |
| The threat | Skip button at 5 seconds | Scroll at any second |
| Opening | Disqualify and claim | Motion and a reason to stay |
| Pacing | Front-loaded, then steady | Even, with a beat every 2 to 3 seconds |
| Length ceiling in practice | Retention, not the platform | Roughly 35 seconds |
| Voiceover load | High, sound is usually on | High, but captions must stand alone |
The practical consequence: an ad built for in-stream is front-loaded to survive a single decision point, while an ad built for Shorts has to keep clearing small decision points all the way through. Running the same file in both means one of them is wrong. We go into the vertical side in more depth on the Shorts ad maker page.
Testing length as a variation axis
Length deserves the same treatment as a hook: a variable you read, not a preference you hold. The mistake is testing length while also changing the script, which gives you two answers tangled together.
A clean length test looks like this. Take one script and one argument. Produce a 15, a 20 and a 30 second cut of it, where the shorter versions lose the least important beat rather than getting faster. Run them against the same audience with the same bid strategy, and read view rate first, then completion rate, then conversions.
One caution from experience: length interacts with vertical. Higher consideration categories tolerate longer cuts, and our benchmark's own vertical figures show how differently these decay, with B2B SaaS taking about 28 days to a 40 percent CTR decline against 9 days for food and beverage. A 45 second in-stream ad that works for enterprise software will not work for a snack brand, and the reason is not the length.
Treat each cut as a genuine ad variation rather than a trim, because that is how the platform treats it. The benchmark's operational finding is that throughput of unique concepts predicts campaign longevity better than the quality of any single ad, and it puts a typical active campaign at 8 to 20 live variations. Length gives you a legitimate way to add a few of those without inventing new claims.
Where we fit, and what we do not do
Genyad is our product. It takes footage you already own, tags every clip, and builds each variation as a fresh script, shot selection, voiceover and caption set, so producing the same argument at 15, 20 and 30 seconds is three variations at one credit each rather than three edit sessions. Exports cover 9:16, 4:5, 1:1 and 16:9 at 1080p on self-serve plans, 4K on Enterprise, no watermark anywhere.
It does not publish to Google Ads, Meta or TikTok, so campaign setup stays yours. There are no AI avatars, no static banners, no product feed rendering and no predicted performance score, because the honest answer to which length will win is that you have to spend a little money to find out.
Frequently asked questions
What is the best length for a YouTube ad?
For anything measured on clicks or purchases, 15 to 30 seconds on both skippable in-stream and Shorts. Use 6 second bumpers only for reach and recall, and 15 second non-skippable when the whole message has to be guaranteed. Longer than 30 seconds is defensible when completion rate stays healthy and the objective is consideration rather than immediate conversion.
When does the skip button appear on YouTube ads?
After 5 seconds on skippable in-stream. That makes the first five seconds the only part of the ad most of your impressions will actually see, so it needs to disqualify the wrong viewer, make one specific claim and show the product. Anything ornamental in that window is paid-for waste.
Can a 6 second bumper drive conversions?
Rarely, and not on its own. Six seconds is enough for one idea and one image, not for a claim plus evidence plus a call to action, so bumpers work best as reinforcement for people already in a retargeting audience. If purchases are the only metric you are judged on, spend the budget on 15 to 30 second cuts instead.
Should I test different lengths of the same ad?
Yes, and keep the script constant while you do it. Produce the same argument at 15, 20 and 30 seconds, where the shorter versions drop the weakest beat rather than speeding up the read, and run them against the same audience. Read view rate first, then completion rate, then conversions, because a length that wins on attention can still lose on revenue.