How to end a YouTube video: what 188,483 scripts show
A third of videos just stop. Ending on a subscribe ask lines up with 1.20x the same channel's cold endings, while a sign-off or a next-video tease on its own changes nothing.

End your YouTube video with a clear ask, ideally to subscribe, placed in the last 30 seconds. A sign-off or a next-video tease on its own does nothing. In 188,483 long-form English scripts I pulled from the TubeAI research database, videos that ended on a subscribe ask did 1.20x better than the same channel's videos that ended cold. Videos that ended with two or more asks did 1.23x better. A plain "thanks for watching, see you next time" landed at 1.01x, and a tease for the next video with no ask at 0.97x.
The most common ending is no ending at all: 32.2% of scripts simply stop, with no ask, sign-off or tease in the final lines.
- 32.2%of videos end cold60,772 of 188,483 scripts
- 1.20xsubscribe ask at the endvs the same channel's cold endings
- 1.03xcomment ask onlyvs the same channel's cold endings
- 19 of 21niches where an ask ending winssame-channel comparison
How YouTube videos actually end
I took the last 600 characters of each transcript, roughly the final 100 words or 30 to 40 seconds of speech, and looked for four things: an engagement ask (the word "subscribe", a comment request such as "let me know in the comments", or a like request), a tease for the next video, a sign-off ("thanks for watching", "see you next time", "bye"), or none of those. Each video lands in exactly one group.

Chart data
| How the last ~600 characters of the script end | Share of videos (%) |
|---|---|
| Cold ending (no ask, sign-off or tease) | 32.2 |
| Engagement ask (subscribe, comment or like) | 30.4 |
| Sign-off only (thanks, see you, bye) | 16.7 |
| Ask + next-video tease | 12.5 |
| Next-video tease only | 8.1 |
Asks are common: 42.9% of endings contain at least one. Sign-offs are everywhere too. Across all scripts, 41.6% of endings include some form of "see you" and 17.5% say "thanks for watching", often alongside an ask.
Which endings line up with better videos
Comparing channels against each other mixes up ending style with channel style, so I compared each channel with itself. For every channel that had both cold endings and endings of a given type, I took the median outlier multiplier of each group (views relative to that channel's typical video) and divided one by the other.

Chart data
| What the ending contains | Median vs the same channel's cold endings (x) |
|---|---|
| Two or more asks | 1.23 |
| Subscribe ask | 1.2 |
| Like ask | 1.07 |
| Comment ask | 1.03 |
| Sign-off only | 1.01 |
| Next-video tease only | 0.97 |
| Ending ask | Videos | Channels compared | Median vs own cold endings | Channels where it beat cold |
|---|---|---|---|---|
| Two or more asks | 31,962 | 4,681 | 1.23x | 56.3% |
| Subscribe ask only | 22,554 | 4,327 | 1.20x | 55.0% |
| Like ask only | 2,290 | 553 | 1.07x | 51.7% |
| Comment ask only | 24,053 | 5,064 | 1.03x | 51.1% |
Three things stand out. First, the gap between asks is bigger than the gap between asking and not asking: a subscribe ask lines up with 1.20x, while a comment ask alone sits at 1.03x, almost the same as ending cold. Second, the polite close most creators use does nothing measurable on its own. Third, even the best ending beats the channel's own cold endings on only 56.3% of channels, so this is a modest edge, not a switch.
Teasing the next video works only when it comes with an ask. Endings that combined an ask with a tease did 1.17x better than the same channel's cold endings (3,271 channels), while a tease alone came in at 0.97x (3,268 channels).
| Ending style | Share of videos | Channels compared | Median vs own cold endings | Channels where it beat cold |
|---|---|---|---|---|
| Ask + next-video tease | 12.5% | 3,271 | 1.17x | 54.7% |
| Engagement ask | 30.4% | 9,734 | 1.15x | 54.0% |
| Sign-off only | 16.7% | 7,189 | 1.01x | 50.2% |
| Next-video tease only | 8.1% | 3,268 | 0.97x | 49.0% |
Does it hold in every niche?
Almost. I repeated the same-channel comparison inside each niche with at least 150 channels to compare, treating any ask as one group. An ask ending beat a cold ending in 19 of 21 niches.

Chart data
| Niche | Ask ending vs same channel's cold endings (x) |
|---|---|
| Crafts & Handmade | 1.42 |
| Food | 1.37 |
| Business | 1.34 |
| Personal Development | 1.34 |
| Homesteading & Off-Grid | 1.32 |
| Automotive | 1.31 |
| Home & DIY | 1.31 |
| Health & Fitness | 1.31 |
| Lifestyle & Personal | 1.3 |
| Beauty & Makeup | 1.28 |
| Financial Markets | 1.21 |
| Product Reviews & Demos | 1.18 |
| Entertainment | 1.16 |
| Tech | 1.15 |
| Gaming | 1.15 |
| Sports & Recreation | 1.15 |
| Relationships & Family | 1.13 |
| Politics & Government | 1.12 |
| Travel & Adventure | 1.1 |
| Commentary & Analysis | 0.93 |
| Reaction Videos | 0.92 |
The edge is biggest in practical, how-to niches: Crafts (1.42x), Food (1.37x), Business and Personal Development (1.34x each). Gaming, the largest group with 2,119 channels, sits at 1.15x. Commentary & Analysis (0.93x) and Reaction Videos (0.92x) lean the other way. My guess is that viewers in those formats come for the take and leave when it is over, but the data cannot confirm why.
How to end a YouTube video
Land the payoff first
Finish delivering what the title promised before anything else. The ask works when the viewer has just gotten what they came for.
Make one clear ask, and make it subscribe
A subscribe ask lined up with 1.20x the same channel's cold endings. A comment ask alone barely moved the needle at 1.03x.
Pair a second ask or a tease with it, not instead of it
Two or more asks reached 1.23x, and an ask plus a next-video tease 1.17x. A tease on its own came in at 0.97x.
Keep the sign-off short
"Thanks for watching" is fine as manners, but it adds nothing measurable. Do not let it replace the ask or drag out the last 30 seconds.
Where else in the video to put the subscribe ask is covered in when to ask viewers to subscribe. For the other end of the script, see how to start a YouTube video, and for how long the whole thing should run, how long a YouTube video should be.
How I measured it
I took English transcripts from the TubeAI research database for videos whose YouTube ID starts with A or B (about 1 in 32 IDs, effectively a random sample), on channels that pass TubeAI's creator curation. I kept long-form videos between 3 minutes and 3 hours with at least 1,500 characters of transcript, published at least 30 days before the query so view counts had time to settle. That left 188,483 videos from 85,860 channels, published through Sep 26 2026 (median publish date June 2025), with transcripts collected between Dec 24 2025 and Oct 25 2026.
The ending is the last 600 characters of the transcript. An ask is the whole word "subscribe", a comment request ("comment", "comments" or "let me know") or a like request ("like button", "hit like", "leave a like" and similar). A tease names the next video, episode, week or part. A sign-off is "thanks for watching", "see you", "catch you", "until next time" or "bye". Performance is the outlier multiplier: a video's views divided by the typical views of the channel's neighboring uploads.
Frequently asked questions
What is the best way to end a YouTube video?
Deliver the payoff, then make one clear ask, ideally to subscribe. In 188,483 scripts, ending on a subscribe ask lined up with 1.20x the performance of the same channel's videos that ended cold.
Should I say thanks for watching at the end of a video?
It does not hurt, but on its own it does not help either. Endings with only a sign-off performed at 1.01x of the same channel's cold endings.
Does asking viewers to comment at the end help?
Barely. A comment ask on its own came in at 1.03x of the same channel's cold endings, against 1.20x for a subscribe ask.
Should I tease my next video at the end?
Only together with an ask. An ask plus a next-video tease reached 1.17x of the same channel's cold endings; a tease alone was 0.97x.
Do most YouTubers end their videos with a call to action?
No. 32.2% of the long-form English scripts I checked end with no ask, sign-off or tease at all, and 42.9% contain an ask in the final lines.
