Four published studies on YouTube title length agree on nothing. ViewsKit says titles under 30 characters earn roughly twice the views of titles between 80 and 100. Briggsby says 20 to 40 characters is the sweet spot. AIR Media-Tech says it depends on the niche.
Subsub.io’s 120,703-video study reports that top-performing titles average 65 to 70 characters, and a widely cited "3 million video" study, the one behind the claim that 90 to 100 characters outperforms by more than 2x, traces back to a domain that does not currently resolve.
We pulled our own numbers instead of picking a side. 28,947 tracked videos across 707 channels, title length checked against a channel-and-age-normalized performance metric, decile by decile, niche by niche, and cross-tabbed against eight other title patterns. The correlation is 0.0046. Title length does not predict performance in this dataset, not weakly, not conditionally, not in any niche we could test with enough volume to trust.
Four Studies, Four Different Numbers
The studies below are all real, sourced, and reachable, we opened and read every one of them this month. None agree on a number, and one of the most-cited claims in this space traces to a website that does not load.
| Source | Sample | Claim |
|---|---|---|
| Subsub.io (Feb 2026) | 120,703 videos | Top performers average 65-70 chars; "shorter wins, almost always" |
| ViewsKit | 61,838 videos | Under 30 chars ~2x the views of 80-100 |
| Briggsby | 100,000 videos, 75,000 channels | 20-40 chars is the sweet spot |
| AIR Media-Tech | 11 niches | No single number, depends on the niche |
| "3M-video study" (cited by 10xcreator.dev, echoed by ytzolo.com) | Claimed, unverified | 90-100 chars outperforms by 2x+ |
| Our data | 28,947 videos, 707 channels | No relationship, r = 0.0046 |
That fifth row deserves a note. 10xcreator.dev ranks first for "best youtube title length," citing a 3-million-video study and a 90 to 100 character recommendation. The domain does not resolve, on repeated attempts, with or without www.
The same unlinked "3 million video" figure appears on two different ytzolo.com pages, with two different numbers attached to it, a 10-14% lift on one page, "more than 2x" on the other. Nobody citing this claim appears to have read the same study, because there does not seem to be a study to read.
Below those five sits a wider swarm: TopicTree, WildAndFreeTools, InfluencerMarketingHub, and Uppbeat each recommend a different range (40-60, 45-60, 70, 50-60) with zero sample size and zero methodology, stated with the same confidence as the real studies above them. Humbleandbrag is the one exception, it cites Subsub.io’s study directly instead of asserting the number as its own.
Our Method, and Its One Honest Limitation
We pulled every tracked long-form video published since January 2025 with at least one recorded view snapshot: 28,947 videos across 707 channels. For each one, we calculated velocity, views divided by days live as of its most recent snapshot, then divided that by the video’s own channel’s median velocity in the same sample. A video sitting at 1.0x is performing exactly as its channel typically does; a video at 2.0x is doing twice as well as that channel’s norm.
That normalization matters because raw view counts mostly measure channel size, not title length. Without it, a length bucket that happens to contain a couple of large channels looks like a length effect. With it, every video is compared only against its own channel’s baseline, the same logic behind the mean-versus-median gap our video length study found hiding inside raw duration averages.
One limitation, stated plainly: this measures performance, views relative to a channel’s own norm, not click-through rate specifically. CTR requires impression data that only a channel’s own owner can see, and we don’t have that for the 707 channels in this dataset. Neither does anyone else’s study of other people’s videos. What we can measure is whether a video ends up doing better or worse than its channel usually does, a real outcome, just not identical to CTR.
The Actual Curve
Split into ten equal-count deciles by character length, the curve looks like this:
The lowest decile, titles from 1 to 35 characters, sits at 0.981x. The highest single point, 91 to 97 characters, sits at 1.010x. Every other decile lands between 0.995x and 1.000x. That is the entire spread: under 3 percentage points across the full range from the shortest titles in the dataset to the longest.
Squared, that correlation is an r² of roughly 0.00002. Title length statistically accounts for about 0.002% of the variance in how a video performs against its own channel’s norm. That is not a weak effect. That is no effect.
Does It Hold Within Any Niche?
AIR Media-Tech’s answer to the disagreement between studies is that it depends on the niche. We checked, across all 15 niches with enough tracked volume to clear a 30-video floor. Every one came back negligible: tech showed the strongest negative correlation at -0.084, vlogs the strongest positive at 0.065, and the other 13 niches fell somewhere between, none crossing the 0.10 threshold that would count as even a weak relationship.
If title length mattered in some corner of YouTube we don’t cover, we didn’t find it in gaming, cooking, music, comedy, finance, education, sports, travel, beauty, fitness, entertainment, news, or vlogs, the full set of niches large enough to test.
Stop guessing at title length. Score the whole upload instead.SEO Studio checks your title, description, tags, and hashtags against what is ranking in your niche right now, not a length rule of thumb. Free to try.Try SEO Studio →Does a Number or a Question Change the Curve?
A reasonable objection: maybe length only matters in combination with something else. We tested two of the most common somethings.
Splitting the dataset into titles with a number and titles without, then running the same length deciles on each group separately, produces two curves that both hover near 1.00x with no clear trend in either one.
Splitting by question-framed titles versus statements does the same: the question-title curve is noisier, only 1,606 titles in the dataset are question-framed, but it shows no consistent length pattern within it either. Length does not start mattering once a number or a question enters the title. It stays exactly as flat.
Checking Our Numbers Against the Closest Real Study
Subsub.io’s study is the one worth taking seriously here, real methodology, a real sample, and it measures several of the same patterns we did. Lined up against our own numbers, here is where we agree and where we don’t:
| Angle | Subsub.io | Ours |
|---|---|---|
| Length | "Shorter wins," top performers average 65-70 chars | No relationship, r = 0.0046 |
| Questions | 6-8% of top titles, largely ineffective | Agree: 5.5% of titles, r = -0.009 |
| Numbers | "Still powerful," 32-40% of top performers | Barely: 40.1% prevalence, r = 0.021 |
| Brackets | Slightly outperform average | Essentially zero, r = 0.002 |
| Emoji | Large effect, but specific to Shorts | Matches on long-form: r = 0.013, we excluded Shorts entirely |
| Capital letters | Strategic emphasis beats all-caps screaming | Cruder measure (has/doesn’t), r = 0.0005 |
The most telling row is the first one. Subsub.io’s own reported "top performer" average, 65 to 70 characters, sits close to our own overall population median of 66 characters, not the top slice, every video we tracked. If shorter titles genuinely outperformed longer ones, the top performers’ average length should sit meaningfully below the population average. It doesn’t, in either dataset. That is what a null result looks like even when a study is confident enough to headline "shorter wins."
The One Pattern Worth a Second Look
Length isn’t the only thing we tested. Eight structural patterns, numbers, leading numbers, question framing, brackets, year tags, colons, all-caps words, and emoji, each checked the same way: prevalence, a median performance split, and its own correlation.
| Pattern | Prevalence | Median lift | Correlation |
|---|---|---|---|
| Year tag (2026, etc) | 8.9% | +9.5% | 0.024 |
| Contains a number | 40.1% | +1.3% | 0.021 |
| Starts with a number | 6.0% | +5.1% | 0.017 |
| Has an emoji | 24.0% | 0% | 0.013 |
| Has a colon | 8.0% | -1.1% | -0.012 |
| Question-framed | 5.5% | -0.4% | -0.009 |
| Has brackets/parens | 7.2% | 0% | 0.002 |
| Has an all-caps word | 37.5% | 0% | 0.001 |
Only one clears even a modest bar. A year tag in the title correlates at 0.024, still technically weak, but the median lift is 9.5%, the largest of anything in this table by a wide margin. Everything else, numbers, brackets, colons, all-caps words, emoji, sits at a correlation under 0.02, indistinguishable from noise.
Frequently Asked Questions
What is the optimal length for a YouTube title?
Across 28,947 videos in our own tracked dataset, there isn’t one. The correlation between title length and performance is 0.0046, statistically negligible, and every length decile from short to long sits within about 2% of a channel’s typical performance. Pick the length the idea needs, not a target character count.
How long can a YouTube title be before it gets cut off?
YouTube allows 100 characters. Search results and suggested feeds typically truncate somewhere around 60 to 70 characters depending on device, so anything past that risks losing its ending in some placements. That is a display constraint, not a performance one: our data shows titles past 70 characters performing the same as titles under 35.
What should I avoid in YouTube titles?
Based on what we measured, avoid over-optimizing for length specifically, it is not the lever it is treated as. Two patterns did show a slight negative direction: a colon in the title (r = -0.012) and question-framed titles (r = -0.009). Neither is strong enough to call a rule, but neither helped either.
How do I know if my YouTube title is good?
Not by its length. Our data found title length, word count, brackets, capital letters, and emoji all had a negligible relationship with performance. A good title is judged by whether it accurately sets up what the video delivers, something no character count measures.
How many words should my YouTube title be?
The median title in our dataset runs 11 words, and word count correlates with performance about as weakly as character count does (r = 0.0050). A commonly cited figure elsewhere is eight words, sourced from a 2026 Subsub.io study of 120,703 videos, but our own measurement, run against tracked upload and view data over time, does not find word count moving performance either.
What is the YouTube 7 second rule?
It is a retention idea, not a title-length one: a video needs to hook the viewer within its first seven seconds or lose them to the next recommendation. That is a real and separate problem from what a title does. A title’s job is to earn the click; the first seven seconds decide whether that click was worth it, and no title length fixes a weak opening.
Does putting a number or a year in my title help?
Marginally, and mostly for year tags. A number in the title showed a 1.3% median lift (r = 0.0214), negligible. A year tag like 2026 was the strongest single signal in the whole study at a 9.5% median lift, still a weak correlation (r = 0.024) but the only one worth a second look.
What should I optimize instead of title length?
Whatever explains the huge variance we did find: individual videos in our dataset ranged from roughly half a channel’s typical performance to more than double it, and title length accounts for almost none of that spread. That points toward topic selection, thumbnail, and the video’s first seconds, not the character count of the title.
Write the Title for the Idea, Not the Counter
Twenty-eight thousand videos, and title length explains close to none of what separates a video that outperforms its channel’s own norm from one that doesn’t. The videos in our dataset range from roughly half a channel’s typical performance to more than double it, and virtually all of that spread comes from somewhere character count cannot reach.
The honest use of a character counter is making sure a title doesn’t get cut off, not chasing a number a study told you to hit. Write the title the idea needs. If it is 24 characters, publish it at 24. If it needs 95 to say the specific thing correctly, publish it at 95.
The data says neither choice is costing you anything, and the four studies arguing about which one wins were arguing about noise. Our own title formula guide is being corrected to point here instead of the unsourced "eight words" figure it previously carried.
