Short-Form Clip Benchmarks 2026: What Creators Export, Edit and Post

What this measures. Every number on this page describes what creators did: which AI clips they exported, how they edited them, and where they posted them. None of it measures audience views, likes or retention. These are benchmarks of creator behaviour, and they show correlation, not cause. The data: 1,770,924 AI clips, 351,491 exports and 181,902 posts on quso.ai between January and September 2026. The methodology is at the end.

Key takeaways

1. 01Position beats length. The top-ranked clip in a video is exported 31.7% of the time; clips ranked 11th or lower, 7.6%.
2. 02Length barely matters. Export rates sit between 14.4% and 16.2% from under 30 seconds to over 60 seconds.
3. 0386% of exported clips keep the AI's suggested length, to within one second. The median export runs 41.6 seconds.
4. 04Accounts that publish through quso tend to publish widely: 83.5% posted to two or more platforms.

1. AI clips cluster at 30 to 60 seconds

The median AI-suggested clip runs 40.6 seconds (10th percentile 25.6 s, 90th percentile 60.0 s). A typical long video yields a median of 9 clips (10th to 90th percentile: 2 to 24).

2. Clip length barely changes whether a clip gets exported

Across all 1.77 million AI clips, 15.5% were exported. The export rate hardly moves with length: 14.4% under 30 seconds, 15.6% at 30 to 45 seconds, 16.2% at 45 to 60 seconds and 15.6% at 60 seconds or more. The same flat pattern holds among top-ranked clips only (29.6% to 32.3%). When each account is weighted equally, the order of the bands changes. We read this as no consistent length effect: in this data, length does not predict which clips get exported.

3. List position is the strongest signal of which clip gets exported

quso lists a video’s clips in order of their AI score. The clip in first place is exported 31.7% of the time. Clips ranked 2nd to 5th are exported 21.9% of the time, 6th to 10th 13.1%, and 11th or lower 7.6%: a 4.2x drop from top to bottom. The pattern held in both periods we checked separately (January to July, and August). This shows what creators pick, not why. Many creators probably review the list from the top, so position and quality are entangled.

4. Within the same position group, the score adds little

Viewed on its own, the AI score looks decisive: clips scoring 95 or more are exported 25.6% of the time, against 11.0% for clips under 80. But high scorers are also the ones listed first. Compare clips in the same position group and the gap almost disappears. For clips ranked 2nd to 5th, the export rate is between 21.6% and 22.3% in every score band. For clips ranked 11th or lower, it is between 7.4% and 8.1%. Our reading: much of the score’s apparent effect comes through the ordering, though these are position groups, not single positions, so some score effect may remain inside them. This data cannot tell you whether a higher-scored clip is a better clip.

5. 86% of exported clips keep the AI’s suggested length

Of 351,491 exports, 86.2% kept the AI’s suggested clip length to within one second. When creators do change the length, they lengthen a clip about as often as they shorten it. 7.2% were extended, by a median of 15.5 seconds, and 6.6% were shortened, by a median of 10.7 seconds. The median exported clip runs 41.6 seconds (10th percentile 25.3 s, 90th percentile 61.1 s).

6. 9 in 10 exports are vertical

91.6% of exports are 9:16: 88.7% standard portrait and 2.9% portrait split-screen. Landscape 16:9 accounts for 7.1% and square 1:1 for 1.4%. Portrait is the editor’s first option, so most of this is a default that creators kept. About one export in twelve was switched to landscape or square.

Format	Share of exports	Accounts using it
9:16 portrait	88.7%	25,766
9:16 split-screen	2.9%	936
16:9 landscape	7.1%	1,833
1:1 square	1.4%	554

7. Captions stay on, and heavier edits are rare

Animated captions are on in 94.3% of exports. They are on by default, so this mostly shows that creators rarely turn them off, not that they chose them. Filler-word and silence removal was used on 1.8% of exports (631 accounts). Only 0.5% of exports changed the playback speed. The most common change was 1.25x (143 accounts); every faster speed had fewer than 50 accounts, so we don’t break those out.

8. Accounts that publish through quso tend to publish widely

Among the 726 accounts that published clips through quso’s publishing flow this year (a small subset of the 26,890 exporting accounts), 83.5% posted to two or more platforms and 32.2% to five or more. Most of them used YouTube.

We report accounts rather than posts because posting is concentrated: the most active tenth of accounts made 57% of all posts. By post count the order is similar: YouTube 23.7%, TikTok 19.5%, Instagram 18.3%, Facebook 17.3%, LinkedIn 8.6%, X 8.0% and Pinterest 4.5%.

9. One video in three is not in English

By transcript language, 65.5% of the videos creators processed were in English. The rest were led by Hindi (11.5%), Spanish (5.5%), Portuguese (3.5%) and Russian (2.1%). This describes quso’s own user base, which has a large share of creators in India, so it is not an estimate for the market as a whole.

Language	Share of videos	Accounts
English (all variants)	65.5%	238,573
Hindi	11.5%	47,188
Spanish	5.5%	21,676
Portuguese	3.5%	14,363
Russian	2.1%	8,815
French	1.5%	5,551
Indonesian	1.2%	4,921
Urdu	1.2%	5,307
Arabic	1.1%	4,779

10. Given a choice, auto-posters pick the fastest cadence

quso’s TikTok auto-post, launched in August 2026, posts a queue of clips at a fixed interval. The default is every 12 hours. Even so, 52.1% of campaigns chose every 3 hours, the fastest option (467 accounts). 35.2% kept 12 hours (381 accounts). 12.7% chose every 6 hours or slower (99 accounts chose 6 hours; the daily and two-day options had too few accounts to report separately). The feature is six weeks old and TikTok-only, so treat this as an early reading.