From the scan data

The honesty ledger: what the scan says it cannot tell you

By the neuro-viral team · 28 Oct 2026 · 5 min read

Every v2neuro scan carries a second-by-second trust score next to the predicted cortical response curve. We do not usually publish the trust score on its own, so here it is, plainly, for three recognizable ads already on our site: Apple's 1984, Nike's You Cannot Stop Us, and Old Spice's The Man Your Man Could Smell Like. For each one we name the single second the model is most confident about and the single second it is least confident about, and we say what a low trust score means: a second we make no claim about.

What the trust score is

The trust score is not the predicted response itself. It is a separate per-second number, 0 to 100, that says how far the model's second-by-second prediction can be relied on at that moment, given how well-documented that brain network is and how typical the second looks to the encoder. A high trust score does not mean the predicted response is high; a low trust score does not mean it is low. The two are independent reads on the same second.

v2neuro groups trust into three bands: higher (the model is on familiar ground), medium (usable, but weaker), and lower (indicative only). Across these three clips, no single second reached the higher band. The most confident second in each clip sits in the medium band. That is the honest ceiling on these three ads, not a flaw we are hiding.

Apple, 1984: most confident at 0:18, least confident at 0:58

In Apple's 1984 ad (59.8 seconds, 60 one-second steps), the most confident second is 0:18, trust score 50 of 100, in the medium band, with the visual network leading. The least confident second is 0:58, trust score 4 of 100, in the lower band. We make no claim about what the predicted response is doing at 0:58; a lower-band second is one the model itself flags as indicative only.

The boring explanation fits here: 0:58 is the last recorded second of the scan. Edge seconds, where the clip is about to end, are exactly where a frame-by-frame model has the least context to work with. The lowest-confidence second being the final second is not a surprising result about this ad; it is a known shape of how the model runs out of runway.

Nike, You Cannot Stop Us: most confident at 0:09, least confident at 0:58

In Nike's You Cannot Stop Us (60.1 seconds, 61 one-second steps), the most confident second is 0:09, trust score 50 of 100, medium band, visual network leading. The least confident second is again the clip's last recorded second, 0:58, trust score 5 of 100, lower band.

Same boring explanation as Apple: the weakest-trust second in this ad is also the final second of the scan. Two of three clips here put their least confident second at the very end, which looks like a property of where a clip stops, not of what either ad shows.

Old Spice, The Man Your Man Could Smell Like: the one that breaks the pattern

Old Spice's ad (32.2 seconds, 33 one-second steps) does not follow the other two. The most confident second is 0:20, trust score 50 of 100, medium band, visual network leading, same shape as the other ads. But the least confident second is 0:11, trust score 14 of 100, lower band, in the middle of the clip, not at the end, with the language network leading.

That rules out the simple end-of-clip explanation for this one. We do not have a second candidate explanation to offer from the data we have on hand; we are naming the second and the band, not guessing at a cause we cannot check.

What this does not tell you

None of this is a measurement of any viewer. The predicted cortical response and its trust score come from v2neuro, a brain-encoding model, applied to the video file. Nobody watched these ads for this piece. A trust score is a statement about the model's own confidence at a given second, not a statement about the ad's quality, and it does not tell you anything about how an audience responded.

We publish the lower-band seconds for the same reason we publish the higher-band ones: a scan that only showed its confident seconds would not be a scan you could check.

Run your own clip

Run your own video against this same scan and see its own trust score, second by second: https://neuro-viral.com/?utm_source=blog&utm_medium=honesty_ledger&utm_campaign=free_scan. If you have a second-by-second question about your own clip, reply and tell us which second you want checked.

Method box

Clips and uids: famous_apple1984 (Apple, 1984, Super Bowl XVIII, 60 seconds), famous_nike (Nike, You Cannot Stop Us, 61 seconds), famous_oldspice (Old Spice, The Man Your Man Could Smell Like, 33 seconds). All three are owner scans already on neuro-viral.com, not customer data.

Rule used to pick the named second: for each clip we read every second's value in that scan's trust.json, dropped any second flagged below_gate (none were dropped here; all seconds in all three scans had a valid trust value), and took the highest and lowest remaining value. Bands (higher / medium / lower) come from the thresholds in each scan's own trust.json legend (0.55 / 0.30 / 0.0).

What this data cannot show: why Old Spice's low-confidence second falls mid-clip instead of at the end; whether trust score relates to anything an audience did; and anything beyond these three clips. This is a note on three ads, not a study.

Frequently asked questions

What does a low trust score mean on a v2neuro scan?

It means the model flags that second as a weaker basis for its predicted-response number, either because that brain network is less documented or because the second looks atypical to the encoder. A low-band second is one we make no claim about.

Which second had the highest trust score in each of these three ads?

Apple's 1984: 0:18 (50/100, medium band). Nike's You Cannot Stop Us: 0:09 (50/100, medium band). Old Spice's The Man Your Man Could Smell Like: 0:20 (50/100, medium band). None reached the higher band on any of the three clips.

Which second had the lowest trust score in each ad, and why?

Apple and Nike both have their lowest trust score on the clip's final recorded second (0:58 in both), which fits the boring explanation that a model has the least context right before a clip ends. Old Spice's lowest trust score is at 0:11, mid-clip, which that explanation does not cover.

Related reading: how to test video ads before you launch · what makes a video hook work?.

Latest breakdowns: Switch Sports Resort trailer: predicted brain response · Percy Jackson season 3 trailer (Disney+): where the brain response peaks · Nobody Wants This season 3 trailer (Netflix): where the brain response peaks · Avengers: Doomsday Trailer: Where the Predicted Brain Response Peaks · Last One Laughing UK Halloween Special trailer (Prime Video): brain response · Cyberpunk 2077 Edgerunners 2 trailer: brain response scan.