Visual scoring

What it is

Visual scoring evaluates the aesthetic register of an ad, its production level, its legibility for sound-off viewers, and whether the visual package matches the inferred Schwartz awareness stage of the audience it's shown to. It is a separate report section from hook analysis and script breakdown — it addresses the look of the ad rather than what the ad says.

Three register tiers account for most DTC creative:

• Ugly-native — looks like organic content. Shot on a phone, no music sting, no brand logo in frame until late or never. Feed-native, low production signal. The Hott principle (Honest, Original, True, Transparent) describes why this register earns thumbstop and trust with cold Problem- and Solution-Aware audiences: it doesn't look like an ad, so the scroll reflex doesn't fire.
• Mid — creator UGC with light production. Competent camera, some color grading, a creator who's clearly "on" rather than candid. Most TikTok-native paid creative lives here.
• Produced — scripted, multi-angle, brand music, logo bug, polish. Fits Product-Aware and Most-Aware audiences where the brand's production quality reinforces authority and trust. On cold traffic it typically signals "ad" before the hook fires.

The register isn't good or bad in isolation. Produced creative on Most-Aware retargeting is often correct. Produced creative on cold prospecting is often the most expensive mistake a DTC media buyer makes.

How AdRevila scores it

One Gemini model call watches the entire video. The visual scoring section of the report outputs:

• Register classification — ugly-native, mid, or produced, with a confidence note when the ad shifts register mid-video (common in hybrid creative that opens ugly-native and cuts to polished B-roll)
• Stage match — does the visual register match the Schwartz stage implied by the hook and body? A mismatch is flagged with the specific tension ("ugly-native hook, produced body — register shift at 4s likely causes dissonance with cold audience")
• Sound-off score — 1–10. Measures whether on-screen text appears early enough, is legible at feed size, covers the primary thesis of the ad, and is timed to the persuasion structure (proof text arriving at the social-proof beat, not randomly)
• Thumbstop visual audit — separate from hook-strength: does the visual design of the opening frame create an interrupt? Color, motion, framing, aspect ratio conformity or violation

The model is not applying a look-up table of production-quality rules. It's watching what a viewer watches, in sequence, and scoring the visual experience as it unfolds.

How to read it

The register classification and stage-match verdict are the lead. An ugly-native ad that stage-matches correctly for cold Solution-Aware traffic needs no visual intervention — the register is doing the job it's supposed to do. A produced ad flagged for register mismatch on cold traffic is the starting point for a creative brief, not an indictment of the production value.

The sound-off score is the most commonly underweighted metric in DTC creative review. Teams watch ads with audio on, which means they experience the ad as a small fraction of Meta's actual feed serves. A 6/10 sound-off score with the note "narration-only thesis in seconds 3–12, no on-screen text equivalent" means roughly half the impressions are receiving no persuasion argument in that window. The fix is almost always additive: subtitle the key line, put the social-proof number on screen, add a text overlay at the moment the solve lands.

Register shifts mid-video are worth flagging separately because they're common and underrecognized. An ad that opens ugly-native — creator in a natural setting, no brand marks — and cuts to produced B-roll of the product at second 6 often experiences a hold-rate drop at exactly that cut. The viewer arrived for the organic-looking content; the register shift signals "ad" at the moment the hook has earned their trust. Keeping the register consistent through the body is a small production decision with outsized hold-rate implications.

Why it matters

Visual register is one of the cheapest levers in DTC creative optimization. Switching from a produced cut to an ugly-native variant of the same script often costs a few hundred dollars and a UGC creator day rate. The thumbstop and CVR lift on cold traffic — where the produced version is triggering the scroll reflex before the hook fires — can be significant.

The harder case is teams that don't realize they have a register problem. They see their produced creative performing at 22% thumbstop and test three new hooks, all produced. The hook isn't the issue; the register is communicating "ad" before any hook fires. Visual scoring surfaces that diagnosis explicitly, which makes the test more targeted: one ugly-native variant against the same body and CTA, same budget, same audience. The register test takes one round. Without the diagnosis it takes three.

Sound-off legibility is the same category of problem: invisible until scored. An ad team that watches their creative with audio on has no signal that 40-50% of their impressions are receiving a silent experience with no on-screen argument. The visual score makes that gap specific.