YouTube Shorts vs Long Form: 1.91× More Long-Form Views
YouTube Shorts vs long form: median long-form views are 1.91× Shorts (1,261 videos, 17 channels). Shorts win views-per-second (9.95×). Full dated method.
We measured YouTube Shorts vs long form on 1,261 view-bearing uploads (760 Shorts + 501 long-form) from 17 large public channels on September 18, 2026. Median long-form views were 2,363,444. Median Shorts views were 1,234,644.5. Long-form has 1.91× the views of the median Short. Shorts do not get more views on this sample.
On the 16 channels that publish both, only 5 of 16 have a higher Shorts median. The median of those per-channel ratios is 0.50.
Formula, one line, applied to every row: views = post.engagement.views; engagement_rate = (likes + comments) / views (drop the row if views are missing or ≤ 0). We dropped shares; they were null on every YouTube row. Same formula as the 2026 engagement-rate benchmark.
The number people quote instead is feed inventory. YouTube's CEO has said Shorts average over 200 billion daily views. That is how much the Shorts Feed plays, not how a typical upload performs. The academic number is older. The 2024 WebSci paper's same-channel ~110× mean covers January 2021–December 2022, before 3-minute Shorts. Those authors already showed that medians reverse in the 10–30 minute bucket. This post is a dated public sample with medians, p25/p75, and a method you can rerun.
Do YouTube Shorts get more views than long-form videos?
No. On the newest Shorts shelf and the latest Videos-tab page from these channels, the median long-form upload out-views the median Short.
| Format | Videos (with views) | Median views | Mean views | p25 views | p75 views | Median likes | Median comments | Median ER | Median duration |
|---|---|---|---|---|---|---|---|---|---|
| Shorts | 760 | 1,234,644.5 | 21,256,027 | 66,792 | 5,297,422.5 | 41,661 | 508 | 2.73% | 54s |
| Long-form | 501 | 2,363,444 | 15,024,454 | 150,134 | 7,260,249 | 68,783 | 3,200 | 2.18% | 1,280s (21m 20s) |
| Long-form / Shorts | — | 1.91× | 0.71× | 2.25× | 1.37× | 1.65× | 6.30× | 0.80× | 23.7× |
Read the median column, not the mean. Shorts mean views (21,256,027) beat long-form mean views (15,024,454) because the Shorts distribution has a heavier right tail: the max Short in this pull is 1,188,886,864 views (MrBeast); the max long-form is 253,910,617. Quote that mean as the "average" Shorts view count and you are describing those outliers. The typical upload in this sample is the median, and the median goes the other way.
Cadence is the obvious objection. A high-volume Shorts poster fills a 48-item "newest" page in days; a slow long-form poster can still have videos from 2023 sitting on a "latest 30" page (Mark Rober does). Median shelf age is close — Shorts 80 days, long-form 94 days — but that is not the same as a matched window. Restricting both formats to a common published_at range makes the long-form lead larger, not smaller.
| Window | Shorts n | Long-form n | Shorts median views | Long-form median views | Long / Short | Shorts median ER | Long-form median ER |
|---|---|---|---|---|---|---|---|
| All shelf items | 760 | 501 | 1,234,644.5 | 2,363,444 | 1.91× | 2.73% | 2.18% |
| Last 90 days | 406 | 245 | 192,337.5 | 232,624 | 1.21× | 2.61% | 1.64% |
| Last 30 days | 269 | 143 | 49,605 | 200,582 | 4.04× | 2.47% | 1.16% |
| Last 7 days | 86 | 36 | 18,974.5 | 130,258 | 6.87× | — | — |
The 30-day cut (269 Shorts vs 143 long-form) is the fairest age-matched claim: long-form median 200,582 vs Shorts 49,605, a 4.04× lead. Last 90 days is 1.21×. The 7-day long-form cell is n = 36 — directional only, too thin to headline.
Two definition notes before you paste the 1.91× into a slide. First, these are YouTube's public viewCount figures. From August 24, 2026, a public view counts the moment a video starts to play — for Shorts, long-form, and live. That is a play-start count, not the engaged-view number YouTube pays on, and it is not unique viewers or watch time. A 54-second Short and a 21-minute video both add 1 to the numerator we used. YouTube's public views are also not Instagram Reel plays; do not mix the two. Second, this is not a random sample. Seventeen large, mostly English-language, recognizable channels. One page per shelf, newest/latest sort, so these are recent uploads, not all-time catalogs. Treat the table as a major-account benchmark, not a platform-wide average.
Does short-form video content outperform long-form on YouTube?
This sample is YouTube-only — not a cross-platform study of short form video content on TikTok or Reels — and the format gap is not a vertical constant. The ratio is median Shorts views divided by median long-form views on that channel, sorted Shorts-advantage first.
Only 5 of 16 paired channels have a higher Shorts median
| Channel | Subs | Shorts n | Long n | Shorts median views | Long-form median views | Shorts / long | Shorts median ER | Long-form median ER |
|---|---|---|---|---|---|---|---|---|
| Gordon Ramsay | 22.0M | 48 | 30 | 513,208.5 | 87,123 | 5.89 | 2.84% | 1.37% |
| NASA | 15.1M | 40 | 30 | 693,085.5 | 158,235.5 | 4.38 | 3.20% | — |
| Mrwhosetheboss | 22.9M | 48 | 30 | 6,383,815.5 | 4,298,721.5 | 1.49 | 3.00% | 2.94% |
| Marques Brownlee | 21.3M | 48 | 30 | 5,282,549.5 | 4,682,743.5 | 1.13 | 3.43% | 2.61% |
| MrBeast | 517M | 48 | 30 | 148,993,785.5 | 141,093,047 | 1.06 | 2.07% | 1.89% |
| Linus Tech Tips | 16.9M | 48 | 21 | 1,097,567 | 1,739,086 | 0.63 | 2.39% | 1.72% |
| The Verge | 3.5M | 48 | 30 | 20,399 | 36,987 | 0.55 | 1.53% | 1.07% |
| Cleo Abram | 8.7M | 48 | 30 | 1,885,766 | 3,748,790.5 | 0.50 | 5.28% | 2.67% |
| NBA | 24.5M | 48 | 30 | 28,122.5 | 57,207.5 | 0.49 | 2.61% | 1.07% |
| Dude Perfect | 62.7M | 48 | 30 | 3,430,937.5 | 7,996,141.5 | 0.43 | 1.94% | 1.47% |
| Mark Rober | 82.7M | 48 | 30 | 19,099,268.5 | 46,759,536 | 0.41 | 1.99% | 1.39% |
| Veritasium | 21.3M | 48 | 30 | 3,874,887 | 9,595,434 | 0.40 | 3.00% | 2.62% |
| Kurzgesagt | 25.6M | 48 | 30 | 1,507,437.5 | 4,823,447.5 | 0.31 | 4.66% | 3.83% |
| Vox | 12.7M | 48 | 30 | 21,483 | 104,100.5 | 0.21 | 3.17% | 1.53% |
| Ali Abdaal | 6.7M | 48 | 30 | 25,696 | 165,403 | 0.16 | 3.32% | 2.79% |
| First We Feast | 15.8M | 48 | 30 | 52,313.5 | 879,678.5 | 0.059 | 1.97% | 1.61% |
5 of 16 (31%) have a Shorts median above their long-form median: Gordon Ramsay (5.89), NASA (4.38), Mrwhosetheboss (1.49), Marques Brownlee (1.13), and MrBeast (1.06). The other 11 go the other way. The median of the 16 ratios is 0.50.
Food is the cleanest split inside one category. Gordon Ramsay's clip-and-recipe Shorts crush his long-form (513,208.5 vs 87,123). First We Feast sits on the floor of the table: its Shorts median is 6% of its Hot-Ones-style long-form median (52,313.5 vs 879,678.5). Same vertical, opposite format bet. Channels choose what to post as a Short versus a video. A Short that is a clip of a long-form upload is not an independent trial. This describes recent shelves, not a causal claim that "switching to Shorts" would move Gordon Ramsay's ratio onto First We Feast.
NASA's long-form engagement rate is blank because comment counts did not populate on that Videos page. PewDiePie (109M subs, 30 long-form, median views 2,442,567.5) has no Shorts in this pull and is omitted from the ratio column.
Category cuts with one channel are that channel
We pooled by the category we tagged before the pull. Entertainment is MrBeast only. Do not quote it as a vertical finding.
| Category | Channels in paired set | Shorts n | Long n | Shorts median views | Long-form median views | Shorts / long | Shorts median ER | Long-form median ER |
|---|---|---|---|---|---|---|---|---|
| Food | 2 | 96 | 60 | 324,512.5 | 255,105.5 | 1.27 | 2.40% | 1.54% |
| Entertainment | 1 (MrBeast) | 48 | 30 | 148,993,785.5 | 141,093,047 | 1.06 | 2.07% | 1.89% |
| Tech | 3 | 144 | 81 | 3,540,623.5 | 3,957,726 | 0.89 | 2.85% | 2.52% |
| Sports | 2 | 96 | 60 | 943,214 | 1,762,993 | 0.54 | 2.22% | 1.26% |
| Explainers | 1 | 48 | 30 | 1,885,766 | 3,748,790.5 | 0.50 | 5.28% | 2.67% |
| News | 2 | 96 | 60 | 20,900.5 | 44,322 | 0.47 | 2.07% | 1.22% |
| Science | 3 | 136 | 90 | 4,026,549.5 | 9,595,434 | 0.42 | 2.54% | 1.99% |
| Education | 1 | 48 | 30 | 1,507,437.5 | 4,823,447.5 | 0.31 | 4.66% | 3.83% |
| Productivity | 1 | 48 | 30 | 25,696 | 165,403 | 0.16 | 3.32% | 2.79% |
| Gaming | 0 paired (PewDiePie long-form only) | 0 | 30 | — | 2,442,567.5 | — | — | 5.60% |
Food (two channels) is the only multi-channel pool where Shorts median views beat long-form, and that pool is Gordon Ramsay plus First We Feast pulling in opposite directions. Tech is close (0.89). Education is Kurzgesagt (0.31). Productivity is Ali Abdaal (0.16). Explainers is Cleo Abram (0.50). Gaming has no paired ratio.
Entertainment looks like a Shorts win until you notice the channel count is 1. We requested handle=RyanTrahan as a second entertainment account. It resolved to @Riiyan (display name Ryan, 37 subscribers, zero videos), not the 20M+ creator. Handle collisions are a real lookup failure mode: always check author.followers after a handle resolve, or the entertainment row is MrBeast wearing a category label.
The 2024 WebSci paper already flagged the same pattern on a much larger, older sample: entertainment Shorts dominated the mean view gap; education and political Shorts did not. A 2026 major-account shelf is not a 9.9-million-video research grant, but the category warning still applies. Shorts vs long-form on YouTube is a per-channel question.
What do the short-form video statistics actually show?
Mean and median disagree — that is why recycled short form video statistics say Shorts win. Views-per-second disagrees with views-per-upload. Likes and comments disagree with each other. If you pick the wrong summary, you can make either format "win."
Mean views say Shorts win. The median says they don't.
Shorts mean views: 21,256,027. Long-form mean views: 15,024,454. Shorts mean is higher (ratio 0.71× long-form / Shorts). Shorts median: 1,234,644.5. Long-form median: 2,363,444. Long-form median is higher (1.91×).
Both distributions are right-skewed. Shorts more so. One MrBeast Short in this sample has 1,188,886,864 views. That single row does a lot of work on the Shorts mean. The median does not care. Lead with the median because it describes the typical upload on these shelves. A chart titled "Shorts get more views" that is built on means is a chart of outliers.
Platform-wide daily-view totals have the same shape problem, just coarser. 200 billion daily Shorts views is how much inventory the Feed served that day. It does not answer whether your next Short will out-view your next long video. Kapwing and others recycle that inventory figure into "short-form is bigger." Inventory is not per-upload performance.
A 54-second Short is 9.95× more viewed per second of runtime
Shorts do win a real metric. Divide views by duration_seconds.
| Format | Median views / second | Median duration | Median views |
|---|---|---|---|
| Shorts | 23,710 | 54s | 1,234,644.5 |
| Long-form | 2,383 | 1,280s | 2,363,444 |
| Ratio | 9.95× Shorts | 0.04× | 0.52× |
A Short in this sample is about 24× shorter (54s vs 1,280s) and 9.95× more viewed per second of runtime. Total views still go to the long video. If the KPI is views per minute of content produced, Shorts win. If it is views per upload, long-form wins.
That split is the closest this harvest gets to a strategy answer. It is still not a revenue answer. Named-creator interviews in Digiday (February 2025) put Shorts RPMs around $0.15–$0.20 against long-form around $3–$6. Those are self-reported interviews, not this sample. We did not multiply our view medians by those RPMs to invent a revenue table. Public views are not what YouTube pays on. More public views are not automatically more money.
Long-form collects 6.3× the median comments
Likes and comments are the rest of the question.
| Format | Median likes | Median comments | Comments as % of likes |
|---|---|---|---|
| Shorts | 41,661 | 508 | 1.2% |
| Long-form | 68,783 | 3,200 | 4.7% |
Long-form median likes are 1.65× Shorts (68,783 vs 41,661). Long-form median comments are 6.3× (3,200 vs 508). Comments as a share of likes: Shorts 1.2%, long-form 4.7%. Shorts are a like-surface. Long-form is where the comment section fills.
Engagement rate uses the same likes-plus-comments over views formula as the 2026 benchmark. It slightly favors Shorts: median 2.73% vs 2.18%. Viewers who land on a Short like or comment at a slightly higher rate. They leave far fewer comments in absolute terms. ER n is 724 Shorts and 471 long-form — 36 Shorts and 39 long-form rows lack comments (NASA's long-form comments were all null), so those rows drop out of the rate and stay in the view tables. Comment counts on YouTube's public surface are often rounded (rows like 21,500 or 1,000); likes and views in this pull look exact. The likes term dominates the ER numerator anyway.
The pattern in one sentence: Shorts convert a view to a like a bit more often, and almost nobody stays to comment.
How did we measure this — and can you reproduce it?
Yes. The question was whether YouTube Shorts get more views than regular long-form videos on a like-for-like public sample. Headline metric: median views per video, Shorts vs long-form. Median because both distributions are right-skewed.
Classification, written before the pull. Shelf-primary. Every item from GET /v1/youtube/channel/shorts is a Short. Every item from GET /v1/youtube/channel/videos is long-form. YouTube documents those two surfaces as disjoint. Sanity flags, not the classifier: URL contains /shorts/, or duration_seconds <= 60. We did not use the 60-second rule as the definition. YouTube Shorts are square or vertical videos up to 3 minutes on standard channels for uploads on or after October 15, 2024. This sample's Shorts max out at 180 seconds, median 54 seconds, and several Shorts use youtube.com/watch?v= URLs rather than /shorts/. #Shorts is not the classifier either.
Sample. Convenience sample of 17 notable public channels (3.5M to 517M subscribers) plus one requested handle that resolved to the wrong account. Mixed category. Not random; a major-account benchmark.
| Handle | Display name | Category | Subscribers |
|---|---|---|---|
| MrBeast | MrBeast | entertainment | 517,000,000 |
| pewdiepie | PewDiePie | gaming | 109,000,000 |
| MarkRober | Mark Rober | science | 82,700,000 |
| DudePerfect | Dude Perfect | sports | 62,700,000 |
| kurzgesagt | Kurzgesagt – In a Nutshell | education | 25,600,000 |
| NBA | NBA | sports | 24,500,000 |
| Mrwhosetheboss | Mrwhosetheboss | tech | 22,900,000 |
| GordonRamsay | Gordon Ramsay | food | 22,000,000 |
| mkbhd | Marques Brownlee | tech | 21,300,000 |
| veritasium | Veritasium | science | 21,300,000 |
| LinusTechTips | Linus Tech Tips | tech | 16,900,000 |
| FirstWeFeast | First We Feast | food | 15,800,000 |
| NASA | NASA | science | 15,100,000 |
| Vox | Vox | news | 12,700,000 |
| CleoAbram | Cleo Abram | explainers | 8,720,000 |
| AliAbdaal | Ali Abdaal | productivity | 6,690,000 |
| TheVerge | The Verge | news | 3,520,000 |
| RyanTrahan | (not used — handle collision) | entertainment | — |
handle=RyanTrahan resolved to @Riiyan (37 subscribers) and contributed zero videos. PewDiePie's Shorts shelf returned 0 items (0 credits). Paired analysis uses the 16 channels with both shelves populated. PewDiePie's 30 long-form videos stay in the pooled long-form column; dropping them moves the long-form median from 2,363,444 to 2,267,024 and does not change the direction of the result.
Pull. One page per shelf, newest/latest sort (not popular — popular would bias toward all-time hits). includeExtras=true on /channel/videos so likes and comments populate. Plus one /v1/youtube/channel call per handle for subscriber counts. No second page. Date pulled: 2026-09-18, 22:04:21–22:05:31 UTC, through the SocialCrawl API.
Endpoints and credits.
GET /v1/youtube/channel— 1 creditGET /v1/youtube/channel/shorts?sort=newest— 1 creditGET /v1/youtube/channel/videos?sort=latest&includeExtras=true— 1 credit
51 credits billed across 54 calls: 18 profile + 16 billed Shorts pages + 17 billed Videos pages. Empty and 404 shelves refunded. YouTube video endpoint catalog: /platforms/youtube.
Rows dropped. We dropped 9 of 1,270 pulled items with no view count (no honest denominator). Engagement rate additionally needs likes and comments.
Sanity. 0 Videos-tab URLs contained /shorts/. 3 Videos-tab items had duration_seconds <= 60 (minimum 45 seconds); they stayed long-form because they came from YouTube's own Videos shelf. 11 NASA Shorts-shelf items predate 2024 (oldest 2012-10-31, a recategorized clip). YouTube's Shorts tab is not a pure "videos this channel published as Shorts this year" feed. Excluding NASA entirely does not flip the headline.
Formula, applied identically.
views = post.engagement.views
likes = post.engagement.likes
comments = post.engagement.comments
engagement_rate = (likes + comments) / views # drop if views missing or <= 0
views_per_sec = views / duration_seconds # drop if duration missing or <= 0
Shares stay out. YouTube does not publish a public share count on these rows. The API also returns computed fields, including an engagement_rate that would add shares when they exist. We did not use that field for the headline — shares are always null here, so the two formulas coincide.
Subscriber figures are YouTube's rounded public counts (author.followers; three significant figures above 1,000). Public views are not unique viewers and not watch time. No causal claim. One page per shelf is recent uploads, not the full catalog.
Reproduce it. Classify by endpoint (Shorts shelf vs Videos tab). For each video with post.engagement.views > 0, take views, and engagement_rate = (likes + comments) / views. Report the median per class. 51 credits for the 18-handle, one-page design (17 intended channels + 1 colliding handle). Swap the handle list for a niche — cooking, coding, sports teams — and the same method is a custom benchmark instead of a recycled "Shorts get more views" chart. The Explorer lets you see the response before writing a single line. First-call setup is in the API quickstart.
curl "https://www.socialcrawl.dev/v1/youtube/channel?handle=mkbhd" \
-H "x-api-key: $SOCIALCRAWL_API_KEY"
curl "https://www.socialcrawl.dev/v1/youtube/channel/shorts?handle=mkbhd&sort=newest" \
-H "x-api-key: $SOCIALCRAWL_API_KEY"
curl "https://www.socialcrawl.dev/v1/youtube/channel/videos?handle=mkbhd&sort=latest&includeExtras=true" \
-H "x-api-key: $SOCIALCRAWL_API_KEY"Frequently asked questions
Do YouTube Shorts get more views than regular videos?
No, not on this sample. Median long-form views were 2,363,444 vs Shorts 1,234,644.5 (1.91×). Last 30 days: 200,582 vs 49,605 (4.04×). 5 of 16 paired channels go the other way. Mean Shorts views are higher because of outliers — do not quote the mean as the answer.
Do YouTube Shorts get more likes than long-form videos?
No. Median likes were 41,661 on Shorts vs 68,783 on long-form (1.65× long-form). Shorts do slightly win rate: median engagement rate 2.73% vs 2.18%, using (likes + comments) / views.
Do YouTube Shorts get more comments than long-form videos?
No. Median comments were 508 on Shorts vs 3,200 on long-form (6.3×). Comments as a share of likes: 1.2% vs 4.7%. Long-form is where the comment section fills.
Should I post YouTube Shorts or long-form videos?
A measured trade-off, not coaching. Views per upload favor long-form on this major-account sample. Views per second of runtime favor Shorts (9.95×). Comments favor long-form. Engagement rate favors Shorts by a little (2.73% vs 2.18%). Channel strategy flips the view gap (Gordon Ramsay 5.89 vs First We Feast 0.059). Public views are not RPM; Digiday's 2025 named-creator interviews put Shorts around $0.15–$0.20 against long-form around $3–$6, and that is their reporting, not this harvest. Subscriber conversion was not measured.
How long is a YouTube Short vs a long-form video?
Officially, a Short is square or vertical and up to 3 minutes, for standard-channel uploads on or after October 15, 2024. This sample classified by YouTube's own shelves, not duration: Shorts median 54 seconds (max 180 seconds); long-form median 1,280 seconds (21 minutes 20 seconds). Three Videos-tab items were ≤60 seconds and stayed long-form. #Shorts is not the classifier.
Related posts
TikTok video search API: region on 30/30, 1 credit
TikTok video search API: 30/30 cooking rows labeled at 1 credit. region=ES still returned 5 non-ES videos — filter country client-side on post.ext.region.
TikTok user search by country: 3 KR hits, 15 credits
3 KR cooking accounts at 15 credits, all location KR. Omit country= and TikTok user search is 1 credit with location null. country=ZZ returns a free 400.
Instagram Reels Stats API: 12 Reels, 2 Credits
Instagram Reels Stats API: 12 public reels with views, likes, shares, and saves for 2 credits. Graph Insights still needs a Page you own. Live 2026-09-15.