The 44-Criteria Virality Matrix: Full Breakdown of All 6 Categories

When we say that ReelsKIT scores a script against 44 criteria, the first reaction is usually: "isn't that just marketing?" Fair question. The term "virality matrix" sounds like a product buzzword. In this guide we break it down to the bone - what's inside, how the weights work, what a niche modifier does, and why the same hook can score 78 in fitness and 51 in business.

Why we need a matrix at all

Most "AI Reels script generators" do one thing: they take your topic, run it through GPT, hand back text. After that - read the tea leaves. Will it go viral? Nobody knows, including the model. Authors of this approach usually say "well, at least it's fast."

The problem is that Reels virality is not a property of a single chunk of text. It's the sum of dozens of small decisions: strength of the first second, rhythm across the timeline, emotional amplitude, voice distinctiveness, the payoff moment, density of visual triggers. Each of those decisions can be scored. If you score them systematically, against a checklist - it becomes clear where the script sags and what to fix.

The virality matrix is that checklist. Not "taste," not guesswork, but 44 atomic criteria, each scored 0 to 5. The weighted sum = your final score from 0 to 100.

The main value of the matrix isn't the score itself - it's the ability to pinpoint: "this scene in your script - 2 out of 5, because there's no emotional amplitude; the hook - 4 out of 5, because there's a curiosity gap but no identity marker."

Six categories and their weights

Weights sum to 100%. Categories aren't placed there "because it looks nice" - they're placed by how much each one actually contributes to real watch retention. Hook weighs the most because nothing else works without it: if the first 1-2 seconds didn't grab them, the other 28 seconds didn't need to be shot.

H (Hook): 9 criteria - 25%

The most loaded category. The hook is what happens in the first 1.5 seconds after autoplay. If in that time the viewer hasn't understood "why should I keep watching," they leave. The platform algorithm sees low first-second retention and stops pushing the video wider.

The 9 Hook criteria:

1. H1. Curiosity gap - is there information asymmetry ("you don't know X, but I'm about to tell you").
2. H2. Identity marker - does the hook address a specific viewer identity ("if you're an entrepreneur," "if you have belly fat").
3. H3. Contradiction / pattern break - breaking expectations, contradicting a common belief.
4. H4. Status hook - social trigger (be above, not be below, stand out).
5. H5. Urgency / loss aversion - loss, deadline, closing window of opportunity.
6. H6. Transformation promise - a "before → after" promise in concrete terms.
7. H7. First-frame density - what the viewer sees on screen at second 0.0 (not just hears, but sees).
8. H8. Verbal compactness - hook length in syllables. The shorter, the higher the meaning density.
9. H9. Author authenticity - does the hook sound like a real phrase from a real person, or an AI template.

H9 is especially important. We added it in Wave 8, after we noticed a pattern: scripts with obvious AI templates ("This will change your life!", "Nobody talks about this, but…") get a high formal score but low retention. Audiences have learned to recognize the AI voice - and they scroll past.

S (Story): 8 criteria - 20%

After the hook you need to hold them. Story is the structure of what happens between hook and payoff.

1. S1. Setup clarity - is it clear in the first 3 seconds "what this is about."
2. S2. Stake escalation - does the stake/tension grow along the timeline.
3. S3. Mid-roll twist - is there an unexpected turn in the middle.
4. S4. Payoff strength - how much the ending rewards the watch.
5. S5. Loop potential - does the video loop back (does the last frame invite a rewatch).
6. S6. Information density - how much new info is delivered per unit time.
7. S7. Logical coherence - no "sag" in the narrative logic.
8. S8. Pacing rhythm - rhythm of scene/point changes (also a Wave 8 addition).

S8 (Pacing rhythm) is usually a weak spot for beginners: they either rattle on without pauses or drag the pauses. Both kill retention. Right rhythm = a "beat" change every 2-4 seconds.

E (Emotion): 8 criteria - 20%

Reels is an emotional medium. If the viewer didn't feel anything - they won't finish it and won't reshare it.

1. E1. Emotional amplitude - range of emotions from start to end.
2. E2. Surprise triggers - frequency of "wow" moments.
3. E3. Relatability - recognizability of the situation/problem.
4. E4. Vulnerability - author's personal vulnerability (drives trust).
5. E5. Tension release - is there a release after tension builds.
6. E6. Humour hits - number of funny moments (where applicable to the niche).
7. E7. Anger / injustice - is there "righteous anger" (a reshare trigger).
8. E8. Hook-payoff resolution - does the payoff close the emotional loop opened by the hook (Wave 8).

E8 is critical. If the hook promised "you'll learn how to X" and the payoff didn't deliver - the viewer feels cheated. This is usually what sags in videos with a strong hook and a weak ending.

V (Voice): 7 criteria - 15%

"Voice" is how the script sounds, separately from what is said. Voice makes content recognizable and makes people follow.

1. V1. Distinctiveness - does the voice differ from typical AI / typical blogger in the niche.
2. V2. Personal anchor - are there personal details, ties to the author.
3. V3. Speech rhythm match - does it sound like spoken language, not written text.
4. V4. Vocabulary range - vocabulary wider/narrower than typical for the niche.
5. V5. Conviction - how confidently each statement lands.
6. V6. Anti-template phrases - absence of platitudes and "speaker phrases."
7. V7. Counter-intuitive twist - how much the voice "surprises" with non-standard phrasing (Wave 8).

V6 and V7 hit especially hard. The banned-phrase filter in ReelsKIT cuts phrases like "in this video I'll tell you," "many don't know," "let's break it down" - these are the "AI template of 2024" that viewers are already tired of.

Vi (Visual): 6 criteria - 10%

Reels is a visual medium, but the matrix scores only what's described in the script (shot list, frames, B-roll). The actual shoot is your zone.

1. Vi1. Shot variety - variety of shots (close/medium/wide).
2. Vi2. B-roll density - how often the visual changes.
3. Vi3. Text overlay strategy - are there visual captions and where they sit.
4. Vi4. Visual hook in frame 1 - what's in the first frame beyond the sound.
5. Vi5. Visual payoff - is there a visual release at the end.
6. Vi6. First-frame thumbnail strength - how well the first frame works as a thumbnail (Wave 8).

A (Audio): 6 criteria - 10%

1. A1. Music tension curve - does music match the drama curve of the script.
2. A2. SFX hits - accents on key moments.
3. A3. Silence usage - are pauses used as a tool.
4. A4. Voice tone variation - does tone/volume change across the timeline.
5. A5. Trend music match - is there a trending sound relevant to the niche.
6. A6. Niche vocabulary signal - does speech carry "insider" markers for the niche (Wave 8).

Niche modifiers: why the same score means different things

The 44 criteria are the baseline. But the importance of each category depends on the niche. For example:

• Fitness / wellness: Visual × 1.4 (movement demos needed), Voice × 0.9.
• Business / SaaS: Story × 1.3 (argument structure needed), Visual × 0.7.
• Beauty / fashion: Visual × 1.5, Audio × 1.2, Story × 0.8.
• Education: Story × 1.4, Hook × 1.1, Visual × 0.8.
• Comedy / lifestyle: Emotion × 1.3 (humour weight), Hook × 1.2.

That means the same script with average visual strength gets 78 in business and 61 in beauty. This isn't "unfair" - it reflects how the niche actually works.

How to read the score and what to fix

When ReelsKIT returns a final score - say, 67/100 - the number itself isn't the most useful part. Useful is the category breakdown:

H: 22/25 ✅ (strong hook)
S: 12/20 ⚠️ (story sagging - no mid-roll twist)
E: 14/20 ⚠️ (flat emotion - low amplitude)
V: 11/15 ✅
Vi: 5/10 ❌ (poor visual - no B-roll strategy)
A: 3/10 ❌ (audio completely ignored)

From this breakdown it's immediately clear what to fix first: add a mid-roll twist to the story, write out a B-roll shot list, plan musical accents. For each problem category ReelsKIT proposes concrete edits - not generic "improve the hook."

Then multi-turn iteration kicks in: "Improve again" doesn't start from scratch - it takes the current version + last scores and fixes the weak spots. Usually 2-3 iterations bump the score by 15-25 points - that's already the difference between "won't fly" and "has a shot."

If you're specifically interested in the hook category alone - we have a sister product HookStar, which generates hooks only, but in 5-10 variants at once, across 6 scientific types. Useful when the script is already there and the hook isn't tight yet.