TikTok vs Reels vs Shorts: what caption length actually works

Caption length on short-form video is one of those things every creator thinks they know but almost nobody has actually measured. We pulled a sample of 1,000 short-form videos (≈330 per platform) from accounts in the 5k–250k follower range across lifestyle, fitness and small-business niches, and graded them against three outcomes: watch-through rate, comments-per-1k-views, and saves-per-1k-views.

Three patterns held up across niches.

TikTok: 80–140 characters wins

On TikTok, captions in the 80–140 character range outperformed both shorter and longer captions on every metric except raw view count. Very short captions (under 40 chars) over-indexed on view count but under-indexed on comments and saves — the algorithm pushed them broadly but the audience didn't engage.

What works in that 80–140 zone is a setup — one line that gives the viewer a reason to keep watching past the 3-second mark. "POV: you just got the promotion you've wanted for three years and you're trying to act normal." That's 95 characters. It's a complete frame, it raises a question (what happens next?), and it tells the algorithm something about who should see this.

The 140+ caption zone on TikTok hurts. Long captions get truncated in the feed UI and the "more" tap rate is under 4%. Anything past the cutoff is functionally invisible.

Reels: 70–110 characters wins, with a twist

Reels is the most ruthlessly short-form of the three. Optimal range is 70–110 characters — about 15% shorter than TikTok. The twist is that Reels has a much more punitive penalty for off-topic captions. If your caption doesn't describe what's in the video, the model under-distributes the Reel.

The pattern that works on Reels: a clear, descriptive one-liner plus one hashtag. "Three small kitchen upgrades under $30 that actually got used. #renterfriendly" is the canonical shape. No question, no CTA, no emoji clutter. The Reel does the work.

The Reels caption is mostly a retrieval signal: it tells the model what the video is about so it can match it to interested viewers. Treat it like alt text for an image, not like a hook.

YouTube Shorts: longer is fine, hashtags matter

Shorts behaves differently. Captions in the 150–280 character range performed best on saves and comments. YouTube ranks Shorts partly using metadata it inherits from the YouTube ecosystem — title text, description, channel topic. The caption (which becomes the Short's title) carries more retrieval weight than on TikTok or Reels.

Three hashtags is the sweet spot. One trending tag (#shorts, niche-relevant), one specific tag, one branded tag. Anything past three diminishes; anything past seven hurts.

The hook is the first 4 words

Across all three platforms, the strongest single predictor of watch-through was the first 4 words of the caption. We hand-graded each caption for whether it had a question, a number, a specific noun, or a contrarian claim in the first 4 words. Captions that did saw 38% higher comment rates and 24% higher save rates on average.

Generic openers like "When you...", "POV:", "Here's how...", and "Why I..." still work but only when paired with a specific second clause. "POV: your mom asked what you do for a living" is a strong hook. "POV: it's a good day" is filler.

What this means for the generator

Our generator now defaults to platform-specific caption length when you pick TikTok, Reels or Shorts. If you want to override the default, you can — but the suggested length is calibrated to this data, not to a generic "social media" template. The fastest way to write a winning short-form caption is to pick the platform first, let the length adapt, and then rewrite the first 4 words yourself.

Rule of thumb

If you only remember one rule per platform: - TikTok: write a setup, not a summary. 80–140 chars. - Reels: describe the video so the algorithm knows what it is. 70–110 chars, one hashtag. - Shorts: treat the caption like a title. 150–280 chars, three hashtags.

Same video, three captions. It's annoying, but the data is consistent enough that we now write three caption variants for every short-form upload we plan internally.

Methodology note: sample collected April–May 2026 from public posts on accounts in the 5k–250k follower range. We controlled for niche and aspect ratio. This isn't a controlled experiment — it's an observational sample — but the effect sizes were large enough that we're comfortable publishing the directional finding.

Last updated: 2026-06-12.