We pulled the data on 3,997 short-form videos across 109 creators, analyzed over 10 months on both TikTok and Instagram Reels. Every video was tagged by hook archetype, emotional trigger, format, idea type, duration, and CTA style. Then we looked at what actually correlates with views, shares, and saves.
The single most shocking finding:
The 49x Gap
Videos built around Fear as the primary emotional trigger averaged 264,031 views. Videos built around Hope averaged 5,423 views. Same platforms. Same time period. Same analysis methodology. A 49x difference, driven entirely by the emotion the creator chose to activate.
That one stat explains most of what we found. The short-form algorithm in 2026 is not neutral. It rewards specific emotional registers, specific hook structures, and specific runtime bands, and it punishes the stuff most creators default to.
Here's the full breakdown.
3,997
Videos analyzed
109
Creators
10 mo
Data window
2
Platforms (TT + IG)
49x
Fear vs Hope views
31x
90s+ vs 15s views
How we did the study
Every video in the dataset went through the same analysis pipeline. Our proprietary AI watched each video, read the transcript, and tagged it across 30+ structural and creative dimensions, including:
Hook archetype (Hot Take, Investigator, Proof Drop, Contrarian, Fortune Teller, Teacher, Story, etc.)
Primary emotional trigger (Fear, Outrage, Curiosity, Empathy, Humor, Trust, Aspiration, Hope)
Format type (Greenscreen, Talking Head, Voiceover + B-Roll, Mixed, Montage)
Idea type (Commentary, Educational, Storytime, Inspirational, Entertainment, Promotional)
CTA style (Implied, Direct, None)
Duration, engagement rate, and raw view/like/share/save counts
Then we aggregated across every video and looked at which tags paired with outsized views, shares, and saves. No opinions. Just the numbers.
Finding 1: Hot Take hooks beat Story hooks by 20x
The hook is still the most important second of any video. But the gap between the best and worst hook archetypes is larger than most people realize.
Average views by hook archetype (min 20 videos)
Hot Take (a bold, opinionated first sentence) and Investigator (a question or unfolding mystery that pulls viewers in) lead the pack at ~140K avg views. Story hooks, the ones that start with "So the other day..." or "Let me tell you about..." , land at 7,127 average views.
That's roughly 20x worse.
If your video opens with "So basically..." or "Here's what happened..." you have already lost most of your reach. The algorithm decides within the first 1.5 seconds whether to keep pushing the video out. Start in the middle of the take, not at the beginning of the story.
Saves vs views is a different story. Proof Drop hooks (the "screenshot this" moment, a chart, a receipt, a specific number) average 1,761 saves per video, the highest in the dataset by a wide margin. If your goal is save-through-rate for an educational or reference piece, Proof Drop wins.
Finding 2: Fear, Empathy, and Outrage are the three emotions that actually move the algorithm
We tagged every video with its dominant emotional trigger. The ranking by average views:
| Emotional Trigger | Videos | Avg Views | Avg Shares | Avg Saves |
|---|---|---|---|---|
| Fear | 51 | 264,031 | 480 | 393 |
| Empathy | 34 | 170,369 | 2,139 | 975 |
| Outrage | 85 | 168,502 | 616 | 313 |
| Curiosity | 664 | 98,145 | 720 | 532 |
| Humor | 78 | 92,253 | 645 | 276 |
| Trust | 224 | 42,149 | 691 | 1,294 |
| Aspiration | 86 | 29,168 | 255 | 151 |
| Hope | 40 | 5,423 | 91 | 39 |
Two things jump out.
One: Fear, Empathy, and Outrage are a tier unto themselves. Each averages 3-5x the views of Curiosity, which is itself 2x Trust. If you want volume, you need to activate one of the top three.
Two: Different emotions drive different metrics.
Wins on raw views. Something to lose, something to protect, something you didn't know. Best for awareness and reach.
Wins on shares (2,139 avg, 3x anything else). Shared because people relate. Best for distribution.
Wins on comments and watch time. The fight in the comments is the point. Best for engagement.
Wins on saves (1,294 avg). People bookmark what they believe. Best for building authority and return viewers.
Pick your emotion based on what you're trying to optimize. Most creators default to Aspiration ("here's how to become X") or Hope ("it gets better"), which are the two worst-performing emotional categories by a mile.
Finding 3: Greenscreen beats plain Talking Head by 2.6x
Format matters more than most creators admit. The highest-performing formats in the dataset all share one trait: visual context behind the creator.
Average views by format type (min 30 videos)
Greenscreen formats (overlaying the creator onto a screenshot, article, chart, or piece of media) average 150K+ views. Plain Talking Head averages 56K. Montage videos (the format that used to dominate Instagram) averages 12K.
Why? Greenscreen gives the viewer two things to look at in one frame: the creator's face for emotional register, and the source material for context. It's a higher information-density format, and the algorithm rewards watch time, which rewards information density.
The practical takeaway
If you're making commentary, news, or hot-take content, stop filming yourself in front of a blank wall. Screenshot the article, the tweet, the chart, or the clip you're reacting to, put it behind you with Greenscreen, and point at the parts that matter. Your average view count will roughly double.
Finding 4: Long videos win. 90s+ beats 0-15s by 31x.
This one is the most counterintuitive finding in the dataset. Every creator "best practice" guide for the last five years has said "keep it short." The data says the opposite.
| Duration | Videos | Avg Views | Avg Shares | Avg Saves |
|---|---|---|---|---|
| 90s+ | 777 | 170,228 | 1,329 | 1,002 |
| 61-90s | 1,880 | 126,299 | 809 | 429 |
| 31-60s | 760 | 27,797 | 167 | 164 |
| 16-30s | 319 | 31,998 | 118 | 44 |
| 0-15s | 261 | 5,406 | 7 | 12 |
Videos over 90 seconds average 170K views. Videos under 15 seconds average 5K. That's not a rounding error. That's a 31x gap across 777 videos in the 90s+ bucket, which is a real sample.
The reason is simple: watch time is the single most important signal the algorithm uses in 2026, and a 90-second video with 70% completion rate produces 63 seconds of watch time, while a 10-second video with 100% completion produces 10 seconds of watch time. The long video wins the algorithm even if fewer people finish it.
The sweet spot for most creators is 60-90 seconds (1,880 videos, 126K avg views). Long enough to build watch time, short enough that you can pack information density across the full runtime. Go to 90s+ only if you have enough substance to sustain it.
Finding 5: Commentary beats Educational, Storytime beats Myth-Busting
Idea type is the "what is this video actually about" tag. Here's the ranking.
Average views by idea type
Commentary beats everything. Pure Educational content sits in the middle. Promotional (the "here's why you should buy my thing" video) averages 26K views. Myth-Busting, which was a dominant format in 2023-2024, has collapsed to 8K average views in this dataset.
The pattern: viewers in 2026 want a point of view, not a lecture. Commentary-led content blends education into opinion, which is why the Educational + Commentary hybrid also performs well at 124K avg views.
Finding 6: Implied CTAs beat direct ones by nearly 2x
This one surprised us. Conventional wisdom says "always have a clear CTA." The data disagrees.
| CTA Type | Videos | Avg Views | Avg Shares |
|---|---|---|---|
| Implied | 485 | 138,650 | 584 |
| None | 243 | 99,432 | 599 |
| Direct | 205 | 72,379 | 445 |
Videos with an implied CTA (the video itself is the CTA, the payoff leaves the viewer wanting more, the comment section becomes the call to action) averaged 138K views. Videos with a direct CTA ("follow for more," "go to my link," "comment X below") averaged 72K, almost exactly half.
Direct CTAs signal to the algorithm that the creator is trying to extract an action, which depresses watch-time on the back half of the video. Implied CTAs preserve watch time and let the audience act on their own terms.
The perfect video: every winning attribute, in one post
We asked the obvious question: is there a video in the dataset that checks every single box above? Top-tier hook archetype, top emotional trigger, Greenscreen format, Commentary idea, 90s+ runtime, Implied CTA?
There is. And it hit 750,062 views.
The 750K Video
Creator: @ev_handd (Instagram Reels) Views: 750,062 | Likes: 67,544 | Comments: 1,849 Hook archetype: Investigator | Emotional trigger: Outrage Format: Greenscreen | Idea type: Commentary Duration: 174.6 seconds | CTA: Implied
The opening line:
This Georgia High School football team is getting absolutely shafted by their own state.
@ev_handd · 750K views
Break down what's happening in that one sentence:
Investigator hook: the word 'shafted' implies a story, an injustice, and a deeper explanation the viewer has to stay for.
Outrage trigger: someone is getting screwed by someone else. You feel the unfairness in the first 3 seconds.
Commentary idea: not a news report, not an explainer. A clear point of view from the creator.
Greenscreen format: the creator overlays article screenshots and game footage behind him as he breaks down what happened.
174 seconds long: enough runtime to build the outrage arc and land the payoff. Deep in the 90s+ bucket.
Implied CTA: the video ends on the unresolved situation. No 'follow me,' no 'link in bio.' The comments do the work.
Every single finding above is executed inside that one video. That's not an accident. The creators who consistently do numbers are not guessing, they are running the same pattern again and again with different subject matter.
What to actually do on Monday
If you want to rebuild your content around what the data shows, here is the five-step playbook:
Open with a Hot Take or Investigator hook. No 'so basically,' no 'let me tell you about.' The first sentence should be a claim or a pulled-in question.
Pick one emotion from the top three (Fear, Empathy, Outrage). Build the entire video around activating that specific register.
Film it in Greenscreen. Screenshot your source material, put it behind you, point at the parts that matter.
Aim for 60-90 seconds minimum, 90-180 seconds if the subject supports it. Shorter is not better.
End on the payoff, not the ask. Let the comment section be the CTA. The algorithm rewards the watch-time you preserve.
This is exactly the analysis The Content Labs runs on your own videos
Every single stat in this article came out of The Content Labs' analysis engine. The same system that tagged these 4,000 videos is the one that, for every user on the platform, analyzes every video on your account, every video on your competitors' accounts, and builds a content strategy out of what's actually working.
What you get inside The Content Labs
- Every one of your videos tagged by hook archetype, emotional trigger, format, idea type, and CTA style, so you can see which of your own patterns are actually carrying your reach.
- Your top competitors tagged the same way, so you can spot the hooks and formats that are working in your niche that you haven't tried yet.
- A weekly content strategy generated from your data plus your competitors' data, with specific video ideas that hit the winning attributes from this study.
- A 30-day content calendar with hook text, format, emotional trigger, and runtime spec for every single video, filled in by an AI agent that has already read everything in your niche.
- The Chemist, a chat assistant that has full context on your content, your competitors, and your strategy, so you can ask it things like "what hook should I use for this idea" and get an answer rooted in your actual data.
If you read this entire article nodding your head, you already understand the framework. The Content Labs runs the framework for you, every day, across your whole account, without you having to tag 4,000 videos by hand.
FAQ
How big was the dataset? 3,997 individual short-form videos across 109 creators, analyzed between June 2025 and April 2026. Roughly 53% Instagram Reels, 47% TikTok.
Is this data public? The aggregated findings are. The individual creator-level data belongs to the creators who use The Content Labs. This article only uses aggregate statistics across all videos in the dataset.
How can I get this kind of analysis on my own account? Sign up for The Content Labs. We run the same tagging engine on every video you post, plus your competitors' videos, and generate a weekly strategy and 30-day calendar built on the findings.
Will the numbers change over time? Yes. Hook archetypes and emotional triggers shift in relative performance as the platforms change. The framework holds. The specific winners inside the framework will keep moving. Which is exactly why "analyze once and publish a listicle" does not work, and why ongoing data-driven content strategy does.