Video Ad Length: What the Data Says in 2026

"Keep it under 15 seconds" is the most repeated and least examined rule in video advertising. I've watched a 64-second founder ad outconvert its 15-second cut by 2x for a $300 product, and I've watched a 6-second price-drop ad beat everything else in an account during a sale week. Both results are completely predictable once you stop asking "how long should video ads be?" and start asking the real question: how long does this argument need for this audience at this price point — and does the retention curve show viewers staying with it?

The data — platform benchmarks, retention curves in my own accounts, and what the length distribution of long-running ads in ad libraries reveals — points to a decision framework, not a magic number. Here it is.

What the aggregate numbers actually show

Strip away the folklore and the load-bearing facts look like this:

• Completion rates fall off a cliff with length. Ads under 10 seconds routinely see completion above 50%; by 30 seconds typical completion sits in the 20–30% band; past 60 seconds you're often under 10%. No surprise.
• But completion isn't the objective — conversion is. The viewers who survive to second 45 of a 60-second ad are dramatically more qualified than the average 15-second completer. Longer ads act as self-selecting filters: worse attention metrics, frequently better cost-per-qualified-action.
• Long-running winners skew short for cheap products and long for expensive ones. Spend an hour in any ad library looking at creatives that have run 90+ days (the profitability proxy): impulse-priced products cluster at 10–20 seconds; considered purchases ($150+) cluster at 30–75 seconds. The market has already voted with budget, and the vote is "it depends on price."
• Attention is front-loaded regardless of length. The steepest drop is seconds 0–3 on every platform and every duration. A 60-second ad doesn't get a more patient audience; it gets the same impatient audience and must re-earn attention every few seconds. Which is why the opening matters more than the runtime — the hook-body-CTA structure holds at every length; only the body stretches.

Length by placement: the table

Platform norms are real — not because platforms magically change human attention, but because placements set expectations and inventory formats reward different durations:

Placement	Sweet spot	Ceiling worth testing	Notes
TikTok / Reels feed	9–21s	~34s	Native pacing; loops reward tight endings
Meta feed (FB/IG)	15–30s	60s	Widest tolerance band of any placement
Stories	6–15s	15s	Tap-through culture punishes slow builds
YouTube skippable in-stream	20–45s	3min+	First 5s unskippable = your real ad
YouTube Shorts	15–35s	60s	Retention loop dynamics like TikTok
Bumpers / non-skip 6s	6s exactly	—	One claim, one image, no argument

Two placement-specific notes that change decisions. On skippable YouTube, you effectively run a 5-second ad with an optional continuation — front-load the brand and claim before the skip button, then argue to the self-selected remainder. On TikTok and Reels, the loop matters: a 12-second ad that loops cleanly gets watched 1.5x on average, which quietly beats an 18-second ad watched 0.8x. Format norms per platform go deeper than length; the format cheatsheet covers safe zones, ratios, and caption placement.

Length by objective and price: the decision rules

Placement sets the range; objective and price pick the point. The rules I actually use:

• Retargeting: go short (6–15s). The audience knows you. A reminder, the offer, urgency, done. Long retargeting ads re-argue a case the viewer already heard, and pay for the privilege.
• Cold traffic, impulse price (<$75): 15–30s. Enough for hook, one argument, one proof, CTA. This is the band where "under 15 seconds" folklore is closest to right.
• Cold traffic, considered price ($150+): 30–75s. The purchase requires an actual case — mechanism, objection handling, proof. Short ads here generate cheap clicks that die on the landing page; the longer ad moves the persuasion into the ad where you control it. Past ~90 seconds you're in VSL territory with different rules entirely.
• Offer/sale ads: as short as the offer. "40% off ends Sunday" needs 6–12 seconds. During peak auction weeks this compounds — short offer ads are also cheaper to refresh against fatigue, which is exactly the dynamic that decides Black Friday accounts.
• App installs and lead gen: 15–25s. The conversion is low-commitment; match it with a low-commitment ask.

The real unlock: length is now a testable variable

Here's what changed in 2026, and why this post isn't just another benchmarks list. Historically, teams picked one length per concept because every additional cut was edit time — so "best length" stayed folklore. With AI production, generating a 15-second, 30-second, and 60-second version of the same concept is marginal effort: same script skeleton with a compressed or extended body, same presenter and product held consistent by reference-to-video models, batch-generated in an AI video generator in an afternoon. Models like Flux 3 video handling longer durations with native audio mean even the 60-second cut doesn't require stitching.

That turns length from a guess into a standing test:

1. Ship every new concept at two lengths — your placement's sweet spot and one length up. (15s + 30s for feed; 30s + 60s for considered purchases.)
2. Read cost-per-qualified-action, not completion. The long cut will lose completion every time; that's the format, not the verdict.
3. Check the retention curve on the long cut. A cliff at one specific second is a content problem at that second — fix or trim to it. A smooth decay means the length itself is fine.
4. Trim to the data. If 70% of conversions attributed to the 60-second cut show watch times under 40 seconds, your real ad is 40 seconds wearing a 60-second costume. Cut it.

Retention curves beat rules of thumb

One habit worth stealing: stop reading length benchmarks monthly and start reading your own retention curves weekly. Three patterns and their meanings:

• Cliff at 3s: hook problem, not length problem. No duration decision matters until this is fixed.
• Cliff at a mid-ad moment: something specific dies there — usually a slow demo section or a repeated point. Surgical fix; the first-frame and thumb-stop work upstream determines how many people even reach it.
• Strong hold, weak conversion: the ad entertains without arguing. Often a sign to go shorter and more direct, spending less on the show and more on the claim.

The teams that win the length question in 2026 don't have a number. They have a testing habit and cheap variants, and the number falls out per concept.

FAQ

What is the best length for a video ad in 2026?

There's no single number: 9–21 seconds for TikTok/Reels feed ads, 15–30 for Meta feed, 6–15 for retargeting and offer ads, and 30–75 seconds for considered purchases over ~$150. Price point and objective move the answer more than platform folklore does.

Do longer video ads ever outperform short ones?

Regularly, for higher-priced or complex products — longer ads filter for qualified viewers and do the persuasion before the click, so they often win on cost-per-acquisition while losing on completion rate. Judge lengths by CPA, never by completion.

How do I test different ad lengths without reshooting?

Generate them. Keep the same script skeleton and compress or extend only the body, hold the presenter and product consistent with reference-to-video models, and batch the 15/30/60-second cuts from one concept in an AI video generator. Then split-test on cost-per-qualified-action.

Why is completion rate misleading for ad length decisions?

Because it mechanically favors short ads: a 6-second ad "completes" easily while converting nobody. The viewers who stay through a long ad are self-selected and more qualified, so a 20% completion rate can carry a better CPA than an 80% one. Pair completion with retention-curve shape and conversion cost.

Does ad length affect delivery costs?

Not directly in the auction, but indirectly yes: shorter ads fatigue slower per impression, refresh cheaper, and loop on short-form placements (boosting effective watch time). During expensive auction periods, short offer ads are the practical choice for rapid rotation.

Stop guessing at runtimes: generate the same concept at two lengths in the AI video generator and let your CPA pick. Free credits daily.