Data Study

What 76 Top TikTok Ads Reveal About Attention

We took 76 verified top performers from TikTok Creative Center — ads with proven CTR and watch-time results — and ran every one through the AdCortex™ brain-engagement model. The results contradict a surprising amount of standard creative advice. Every number below links to the actual ads in the benchmark library, so you can check the work.


Finding 1: "winning" ads vary 55× in predicted engagement

Every ad in this set earned its place in TikTok's Top Ads showcase, yet the engagement index ranged from 57.3 (StickyLint's 42-second demo) down to effectively zero (a 15-second apartment tour) — a 55× spread among proven winners, median 30.9. The takeaway isn't that the low scorers are bad ads; it's that platform performance and neural attention measure different things, and knowing where your creative sits on the second axis before you spend is exactly the point of ad pre-testing.

Finding 2: length is a non-factor (r = 0.04)

The set runs from 5.3 seconds to 4 minutes 21 seconds, and runtime explains essentially none of the engagement variance. The shortest ad scored 8/100; the longest scored 9/100 — both near the bottom, for reasons that had nothing to do with length. Meanwhile the top 20 ads ran a median of 28 seconds against 14 seconds for the bottom 20: strong creative earns its runtime rather than being punished for it. The full breakdown is in ad length vs. engagement.

Finding 3: top ads build — the median peak lands 57% in

Only 17 of the 76 ads hit their attention peak in the first third of the runtime; 31 peaked in the middle third and 28 in the final third. The pattern among the strongest performers is consistent: a competent open, then escalation to a mid-video payoff. Privacycase (#2) peaks at 16.8s of 20s; Comfrt (#3) at 30.6s of 42s. The inverse pattern — opening at your peak and decaying — is what the bottom three ads all share. More on this in hooks ranked by brain data.

Finding 4: scripted ads out-engage music-only ads

59 of the 76 ads have a real script or voiceover; 17 are carried by music and visuals alone. The scripted group's median engagement index was 33.3 against 26.3 for the music-driven group. Voice gives the brain a second channel to track — but the exceptions are instructive: waterdrop's music-only ad scored 89/100 on pure visual pacing. Full comparison in music vs. voiceover.

Finding 5: platform CTR tier told us nothing about neural engagement

39 ads in the set carry TikTok's own CTR-percentile tier ("Top 6% CTR" etc.). Correlation with brain engagement: r = −0.03 — noise. Ads in TikTok's top-10%-CTR tier had a median engagement index of 36.4; ads in the much weaker top-34%-and-below tier scored 38.1. A high CTR means the first frame won the click; it says nothing about whether the following 30 seconds hold a brain. That gap is the subject of high CTR vs. brain engagement — and it's the same story as high CTR, low conversions.

Method, honestly stated

Each ad was scored by the AdCortex™ model — trained on fMRI brain scans to predict visual-cortex activation — producing a second-by-second attention curve and a single engagement index (details in the methodology). Correlations are Pearson r across the 76 ads. This measures predicted attention, not conversions; sample sizes for sub-groups are stated inline, and where n is small we say so. These 76 ads are also the exact benchmark every PreTestAds score is measured against, so "your ad scored 72" means it out-engaged 72% of this set.

See where your ad lands against these 76

Upload your creative and get the same attention curve and percentile score — first analysis free, no credit card.

Score My Ad Against the Benchmark