What works for new AI story accounts
Data report for the two story-reel channels. Built 2026-10-23 03:49 from free public data. Hover a bar for details; each chart has a table.
284new faceless animated accounts (2025-2026)
31,437videos scored against their own account
22.8%of all videos hit 3x their account median
127real scripts read, hits vs flops
How to read this
Every video is compared with its own account's median, so a tiny new page and a giant one count the same. A "hit" is 3x that median or more. Titles are tagged by simple word rules, so a video can sit in several groups.
What the data says
- Hooks: hits open a specific question in the first 15 words twice as often as flops of the same channel (the most solid finding).
- Reach: in a video's first two weeks, the ones with more likes, shares and saves per view are the ones that travel. Comments per view fall as views grow.
- Titles: on the full set only two features clearly lift the hit rate inside the same account: a number in the title (1.25x, 194 accounts) and animals (1.18x). Space and Why looked ahead on the first 15,000 videos and did not hold; What if and Money show no effect. Title words are a weak lever.
- Niche: what-if accounts get about twice the comments per view; science explainers the fewest. Share and save gaps between niches are noise.
- Length: videos under 20 s hit 3x less often than the same account's others (0.69x); over 90 s more often (1.38x, 106 accounts), likely because TikTok favours longer videos. A lead to test, not a rule.
- Shares: extreme words, animals and survival topics lead on shares inside the same account; videos under 20 s get far fewer shares and saves. Leads, not rules.
- Look: among what-if accounts under 40 days old, realistic 3D grows fastest, on small numbers.
A lead: which look grows fastest among new what-if accounts
Median followers gained per day, what-if accounts under 40 days old only (older accounts grow slower whatever their look, so mixing ages misleads). Small groups: a lead to test, not proof.
Realistic 3D people / crowds
4332 n=5 accts
Toy or low-poly 3D
1489 n=17 accts
Clean 3D diagram + labels
978 n=29 accts
Flat 2D cartoon
652 n=6 accts
Table
| group | value | n | note |
|---|---|---|---|
| Realistic 3D people / crowds | 4332 | 5 accts | |
| Toy or low-poly 3D | 1489 | 17 accts | |
| Clean 3D diagram + labels | 978 | 29 accts | |
| Flat 2D cartoon | 652 | 6 accts |
Which title formats and topics hit 3x more often, inside the same account
Free data, 31,437 videos (TikTok, YouTube, Instagram). 1.0x = the same account's untagged videos. A video hits when it gets 3x its account's median views. Many groups are tested, so about 1 in 20 'clear' results can be chance: trust a result most when the other dataset agrees.
all videos: 1.00x
Has a number (clear)
1.25x n=194 accts
Animals (clear)
1.18x n=59 accts
Body
1.09x n=41 accts
A to B (becomes, turns into)
1.08x n=29 accts
What if
1.07x n=52 accts
Money
1.04x n=64 accts
How
1.04x n=78 accts
Why
1.02x n=66 accts
Extreme word (most, biggest, ever)
1.01x n=69 accts
Asks a question (?)
0.99x n=151 accts
Survival / trapped
0.97x n=52 accts
Says you / your
0.96x n=155 accts
Space
0.90x n=74 accts
Versus
0.87x n=38 accts
POV
0.79x n=14 accts
Table
| group | value | n | note |
|---|---|---|---|
| Has a number (clear) | 1.25x | 194 accts | 95% range 1.12 to 1.34; clear = the range does not include 1.0 |
| Animals (clear) | 1.18x | 59 accts | 95% range 1.03 to 1.36; clear = the range does not include 1.0 |
| Body | 1.09x | 41 accts | 95% range 0.82 to 1.29; clear = the range does not include 1.0 |
| A to B (becomes, turns into) | 1.08x | 29 accts | 95% range 0.85 to 1.29; clear = the range does not include 1.0 |
| What if | 1.07x | 52 accts | 95% range 0.79 to 1.50; clear = the range does not include 1.0 |
| Money | 1.04x | 64 accts | 95% range 0.83 to 1.24; clear = the range does not include 1.0 |
| How | 1.04x | 78 accts | 95% range 0.91 to 1.17; clear = the range does not include 1.0 |
| Why | 1.02x | 66 accts | 95% range 0.87 to 1.20; clear = the range does not include 1.0 |
| Extreme word (most, biggest, ever) | 1.01x | 69 accts | 95% range 0.83 to 1.17; clear = the range does not include 1.0 |
| Asks a question (?) | 0.99x | 151 accts | 95% range 0.88 to 1.14; clear = the range does not include 1.0 |
| Survival / trapped | 0.97x | 52 accts | 95% range 0.75 to 1.21; clear = the range does not include 1.0 |
| Says you / your | 0.96x | 155 accts | 95% range 0.85 to 1.07; clear = the range does not include 1.0 |
| Space | 0.90x | 74 accts | 95% range 0.80 to 1.04; clear = the range does not include 1.0 |
| Versus | 0.87x | 38 accts | 95% range 0.64 to 1.19; clear = the range does not include 1.0 |
| POV | 0.79x | 14 accts | 95% range 0.37 to 1.22; clear = the range does not include 1.0 |
Does length change the hit rate, inside the same account
Free data, videos with a known length. 1.0x = the same account's videos of other lengths. Many groups are tested, so about 1 in 20 'clear' results can be chance: trust a result most when the other dataset agrees.
all videos: 1.00x
over 90 s (clear)
1.38x n=106 accts
60-90 s
1.02x n=118 accts
20-30 s
0.96x n=55 accts
30-45 s
0.91x n=71 accts
45-60 s
0.87x n=76 accts
under 20 s (clear)
0.69x n=50 accts
Table
| group | value | n | note |
|---|---|---|---|
| over 90 s (clear) | 1.38x | 106 accts | 95% range 1.20 to 1.64; clear = the range does not include 1.0 |
| 60-90 s | 1.02x | 118 accts | 95% range 0.90 to 1.15; clear = the range does not include 1.0 |
| 20-30 s | 0.96x | 55 accts | 95% range 0.80 to 1.14; clear = the range does not include 1.0 |
| 30-45 s | 0.91x | 71 accts | 95% range 0.72 to 1.10; clear = the range does not include 1.0 |
| 45-60 s | 0.87x | 76 accts | 95% range 0.64 to 1.07; clear = the range does not include 1.0 |
| under 20 s (clear) | 0.69x | 50 accts | 95% range 0.54 to 0.93; clear = the range does not include 1.0 |
Which viewer actions go with reach (per view, inside the same account)
Rank correlation between each action per view and views, paid TikTok data (1550 videos, 60 accounts). Positive = videos that got more of it per view also got more views. It only holds while a video is new.
under 14 days old 14+ days old
likes
+0.36 n=921
+0.04 n=629
shares
+0.36 n=921
-0.01 n=629
saves
+0.28 n=921
-0.03 n=629
comments
-0.26 n=921
-0.36 n=629
Table
| group | under 14 days old | n | 14+ days old | n | note |
|---|---|---|---|---|---|
| likes | +0.36 | 921 | +0.04 | 629 | |
| shares | +0.36 | 921 | -0.01 | 629 | |
| saves | +0.28 | 921 | -0.03 | 629 | |
| comments | -0.26 | 921 | -0.36 | 629 |
What people share more, inside the same account
Paid TikTok data: shares per view of tagged videos vs the same account's other videos (1.0x = no difference). Many groups are tested, so about 1 in 20 'clear' results can be chance: trust a result most when the other dataset agrees.
all videos: 1.00x
Animals (clear)
2.53x n=12 accts
Why
1.73x n=6 accts
Body
1.34x n=6 accts
length 20-30 s
1.26x n=5 accts
Survival / trapped (clear)
1.23x n=9 accts
Extreme word (most, biggest, ever) (clear)
1.23x n=11 accts
Says you / your
1.18x n=36 accts
Space
1.14x n=20 accts
length 45-60 s
1.11x n=8 accts
length over 90 s
1.11x n=20 accts
What if
1.09x n=14 accts
A to B (becomes, turns into)
1.08x n=8 accts
Asks a question (?)
1.02x n=24 accts
POV
1.01x n=11 accts
Has a number
0.99x n=28 accts
length 60-90 s
0.98x n=28 accts
length 30-45 s (clear)
0.80x n=7 accts
How
0.79x n=12 accts
Money
0.63x n=10 accts
length under 20 s (clear)
0.27x n=4 accts
Table
| group | value | n | note |
|---|---|---|---|
| Animals (clear) | 2.53x | 12 accts | 95% range 1.77 to 4.66; clear = the range does not include 1.0 |
| Why | 1.73x | 6 accts | 95% range 0.84 to 3.10; clear = the range does not include 1.0 |
| Body | 1.34x | 6 accts | 95% range 0.87 to 1.90; clear = the range does not include 1.0 |
| length 20-30 s | 1.26x | 5 accts | 95% range 0.00 to 1.57; clear = the range does not include 1.0 |
| Survival / trapped (clear) | 1.23x | 9 accts | 95% range 1.04 to 1.77; clear = the range does not include 1.0 |
| Extreme word (most, biggest, ever) (clear) | 1.23x | 11 accts | 95% range 1.06 to 1.50; clear = the range does not include 1.0 |
| Says you / your | 1.18x | 36 accts | 95% range 0.99 to 1.46; clear = the range does not include 1.0 |
| Space | 1.14x | 20 accts | 95% range 0.97 to 1.58; clear = the range does not include 1.0 |
| length 45-60 s | 1.11x | 8 accts | 95% range 0.79 to 1.78; clear = the range does not include 1.0 |
| length over 90 s | 1.11x | 20 accts | 95% range 0.91 to 1.36; clear = the range does not include 1.0 |
| What if | 1.09x | 14 accts | 95% range 0.73 to 1.36; clear = the range does not include 1.0 |
| A to B (becomes, turns into) | 1.08x | 8 accts | 95% range 0.53 to 1.52; clear = the range does not include 1.0 |
| Asks a question (?) | 1.02x | 24 accts | 95% range 0.87 to 1.37; clear = the range does not include 1.0 |
| POV | 1.01x | 11 accts | 95% range 0.57 to 1.78; clear = the range does not include 1.0 |
| Has a number | 0.99x | 28 accts | 95% range 0.78 to 1.13; clear = the range does not include 1.0 |
| length 60-90 s | 0.98x | 28 accts | 95% range 0.87 to 1.35; clear = the range does not include 1.0 |
| length 30-45 s (clear) | 0.80x | 7 accts | 95% range 0.40 to 0.91; clear = the range does not include 1.0 |
| How | 0.79x | 12 accts | 95% range 0.70 to 1.16; clear = the range does not include 1.0 |
| Money | 0.63x | 10 accts | 95% range 0.51 to 1.19; clear = the range does not include 1.0 |
| length under 20 s (clear) | 0.27x | 4 accts | 95% range 0.00 to 0.65; clear = the range does not include 1.0 |
What people save more, inside the same account
Paid TikTok data: saves per view of tagged videos vs the same account's other videos (1.0x = no difference). Many groups are tested, so about 1 in 20 'clear' results can be chance: trust a result most when the other dataset agrees.
all videos: 1.00x
Survival / trapped (clear)
1.35x n=9 accts
Why
1.28x n=6 accts
Animals
1.28x n=13 accts
Body (clear)
1.25x n=6 accts
Extreme word (most, biggest, ever)
1.16x n=11 accts
length over 90 s (clear)
1.16x n=20 accts
Has a number (clear)
1.16x n=28 accts
Says you / your
1.15x n=36 accts
Space
1.09x n=20 accts
Asks a question (?)
1.05x n=27 accts
POV
1.05x n=12 accts
A to B (becomes, turns into)
0.98x n=8 accts
length 60-90 s
0.97x n=30 accts
length 45-60 s
0.96x n=8 accts
How
0.96x n=12 accts
What if
0.88x n=14 accts
length 30-45 s
0.86x n=8 accts
Money
0.84x n=10 accts
length 20-30 s
0.76x n=7 accts
length under 20 s (clear)
0.58x n=4 accts
Table
| group | value | n | note |
|---|---|---|---|
| Survival / trapped (clear) | 1.35x | 9 accts | 95% range 1.07 to 1.48; clear = the range does not include 1.0 |
| Why | 1.28x | 6 accts | 95% range 0.97 to 1.40; clear = the range does not include 1.0 |
| Animals | 1.28x | 13 accts | 95% range 0.97 to 1.54; clear = the range does not include 1.0 |
| Body (clear) | 1.25x | 6 accts | 95% range 1.15 to 1.46; clear = the range does not include 1.0 |
| Extreme word (most, biggest, ever) | 1.16x | 11 accts | 95% range 0.96 to 1.59; clear = the range does not include 1.0 |
| length over 90 s (clear) | 1.16x | 20 accts | 95% range 1.00 to 1.27; clear = the range does not include 1.0 |
| Has a number (clear) | 1.16x | 28 accts | 95% range 1.01 to 1.36; clear = the range does not include 1.0 |
| Says you / your | 1.15x | 36 accts | 95% range 1.00 to 1.37; clear = the range does not include 1.0 |
| Space | 1.09x | 20 accts | 95% range 0.99 to 1.24; clear = the range does not include 1.0 |
| Asks a question (?) | 1.05x | 27 accts | 95% range 0.98 to 1.28; clear = the range does not include 1.0 |
| POV | 1.05x | 12 accts | 95% range 0.79 to 1.32; clear = the range does not include 1.0 |
| A to B (becomes, turns into) | 0.98x | 8 accts | 95% range 0.71 to 1.14; clear = the range does not include 1.0 |
| length 60-90 s | 0.97x | 30 accts | 95% range 0.88 to 1.05; clear = the range does not include 1.0 |
| length 45-60 s | 0.96x | 8 accts | 95% range 0.82 to 1.50; clear = the range does not include 1.0 |
| How | 0.96x | 12 accts | 95% range 0.83 to 1.36; clear = the range does not include 1.0 |
| What if | 0.88x | 14 accts | 95% range 0.74 to 1.09; clear = the range does not include 1.0 |
| length 30-45 s | 0.86x | 8 accts | 95% range 0.75 to 1.10; clear = the range does not include 1.0 |
| Money | 0.84x | 10 accts | 95% range 0.70 to 1.11; clear = the range does not include 1.0 |
| length 20-30 s | 0.76x | 7 accts | 95% range 0.62 to 1.20; clear = the range does not include 1.0 |
| length under 20 s (clear) | 0.58x | 4 accts | 95% range 0.11 to 0.93; clear = the range does not include 1.0 |
Which niches get talked about
Median comments per 1,000 views by niche (paid TikTok). Only the what-if (high) and science (low) gaps pass a permutation test; share and save gaps between niches are noise.
story
0.98 n=4 accts
what-if / simulation
0.85 n=26 accts
survival story
0.47 n=3 accts
shared-reel accounts
0.44 n=5 accts
tech / how it works
0.34 n=4 accts
science explainer
0.22 n=11 accts
Table
| group | value | n | note |
|---|---|---|---|
| story | 0.98 | 4 accts | p = 0.137 vs other niches (noise) |
| what-if / simulation | 0.85 | 26 accts | p = 0.002 vs other niches (real) |
| survival story | 0.47 | 3 accts | p = 0.791 vs other niches (noise) |
| shared-reel accounts | 0.44 | 5 accts | p = 0.666 vs other niches (noise) |
| tech / how it works | 0.34 | 4 accts | p = 0.367 vs other niches (noise) |
| science explainer | 0.22 | 11 accts | p = 0.024 vs other niches (real) |
How hits are told vs flops of the same channel
Share of scripts with each feature: 63 hits vs 64 flops from 15 channels.
hits flops
A specific question in the first 15 words (solid, p=0.01)
49% n=63
25% n=64
Opens with a question
24% n=63
11% n=64
Opens with background / context
21% n=63
28% n=64
Tells a life story (weak, p=0.10)
16% n=63
28% n=64
Spoken call to action at the end
19% n=63
28% n=64
Has a twist
19% n=63
20% n=64
Table
| group | hits | n | flops | n | note |
|---|---|---|---|---|---|
| A specific question in the first 15 words (solid, p=0.01) | 49% | 63 | 25% | 64 | |
| Opens with a question | 24% | 63 | 11% | 64 | |
| Opens with background / context | 21% | 63 | 28% | 64 | |
| Tells a life story (weak, p=0.10) | 16% | 63 | 28% | 64 | |
| Spoken call to action at the end | 19% | 63 | 28% | 64 | |
| Has a twist | 19% | 63 | 20% | 64 |
What held up after an independent review
A separate agent recomputed every headline finding from the raw files and tried to break it.
| finding | verdict | what the data supports |
|---|---|---|
| Realistic 3D grows fastest | weakened | age drives followers/day; only a lead among new what-if accounts |
| What if / Money titles help | weakened | no effect inside the same account; on the full 31K set only a number in the title (1.25x) and animals (1.18x) are clear; Space and Why did not hold |
| 20-30 s videos hit most | refuted | one account made it; on the full set under 20 s is worse (0.69x) and over 90 s better (1.38x), a lead |
| Shares and saves go with reach | weakened | true only in a video's first ~2 weeks |
| Title words don't move shares | refuted | the old method was 1.00 by construction; some tags do move shares |
| Niche share and save rates differ | weakened | only comments differ for real (what-if high, science low) |
| Instagram shares 2-5x TikTok | weakened | 0.8-1.8x on the same channels; IG shares and likes rise with views |
| Some comment types mark a hit | weakened | comment mix is about the same on hits and normal videos |
| A question in the first 15 words | confirmed | 31 of 63 hits vs 16 of 64 flops, p=0.01, holds leaving out any channel |
| Script QC gate | weakened | clean wording keeps 19 of 21 hits, stops 14 of 21 flops; a held-out test is running |
Limits
Instagram hides views without a login, so Instagram has about 12 recent reels per account. YouTube lists carry no dates. Follower counts are summed across platforms. Title tags are word rules, not a person's judgement. This shows what goes with hits, not proof of cause: the real test is posting and measuring.