Skip to main content
RoastIQBuyerLensHugoPricingBlogAbout
Book a demoSign inStart free →
Industry trend · n=1,110 · published 18 May 2026 · 6 min read

Sound-off performance, quantified.

Half the feed is muted. The penalty for sound-on-dependent creative is not a vague "best practice", it is −11 to −15 Get Noticed points, by surface, measured against a 1,110-ad cohort. The four moves that survive muted, the categories that get punished hardest, and the checklist that prevents it.

Sample · n=1,110 TikTok mute · 52% Meta-Feed mute · 64% YT Shorts mute · 38% Confidence · HIGH
Oussama Nakhil portrait
Oussama Nakhil · Founder & CEO
Multiple years buyer-side: NielsenIQ insights, then L'Oréal Groupe in global marketing insights. Every chart pinned to model_version and benchmark_pool_version v.2026-05.
64%
Meta-Feed plays muted · highest of 4 surfaces
−15
Worst-case Get Noticed penalty (Meta-Feed)
4
Things that survive a muted watch

01 · Mute is the default.

Across the four short-form vertical surfaces in our cohort, the median ad view is muted. The exact share varies by surface and by user behaviour, but the order is unambiguous: Meta-Feed (64%) > TikTok For You (52%) > Instagram Reels (49%) > YouTube Shorts (38%). Meta-Feed leads because the surface autoplays muted by default for the largest share of users, and because the feed context (scrolling alongside friend content) primes the muted habit. Shorts trails because the surface inherits viewers mid-session from longer-form YouTube, where sound expectation is structurally higher.

The mute rates above are weighted from platform-published figures where available and modelled against our pool of 1,110 ads where they are not. The shape is stable across every cohort version we have shipped since 2025-Q1. Mute is not an edge case in the planning conversation; it is the working assumption.

Meta-Feed muted
64%
Autoplay-muted by default · most users
TikTok muted
52%
For You feed · median session
Reels muted
49%
Slightly below TikTok · shared paradigm
Shorts muted
38%
Inherits sound-on from long-form sessions

02 · The penalty curve.

The Get Noticed penalty for sound-on-dependent creative, defined as cuts where the spoken value proposition, the brand reveal, or both, depend on audio comprehension to land, is concentrated by surface. Meta-Feed punishes hardest at −15 points, followed by TikTok (−12), Reels (−11) and YouTube Shorts (−7). The penalty matches the mute rate ordering and is driven almost entirely by Get Noticed, with secondary effects on Sell Proposition (where the spoken claim is muted) and Brand Impact (where the spoken brand mention is muted).

// Get Noticed penalty for sound-on creative, by surface (n=1,110)

Same cut, scored against each surface's benchmark sub-pool. Higher bar = larger penalty. Meta-Feed punishes hardest.
Meta-Feed
−15
TikTok · For You
−12
Instagram · Reels
−11
YouTube · Shorts
−7

Two implications follow. First, Meta-Feed and TikTok cuts should be built sound-off as the default and have audio added as a bonus track. The reverse, building sound-on and "checking" muted at the end, produces the −15 penalty by design. Second, YouTube Shorts is the only surface where audio-led creative is structurally viable, and even there, the bottom-quintile audio-led cut underperforms by margin large enough to make sound-off design the safer default brief.

03 · By category, the penalty diverges.

The penalty is not evenly distributed across categories. Categories whose creative tradition leans on voice-over (auto, finance, telco) get punished hardest; categories where the visual grammar already does the work (beverage, snacks, beauty) absorb the mute with minimal damage. The category-by-category penalty mirrors the seven-failure-mode catalogue we track in the pipeline, "sound-on dependency" is one of the seven systematic failure modes detected and flagged.

CategorySample (n)Get Noticed penaltySound-on dependency rateVisual brand cue (median latency)
Beverage112−722%0.9s
Beauty156−931%1.4s
Snacks88−928%1.0s
Telco124−1354%2.1s
Finance91−1466%4.8s
Auto82−1571%4.2s
The compounding

"Auto and finance pay twice. The category brands late (see Report 02) and depends on voice-over to land the proposition. Mute compounds the cue-late penalty. The result is structural, not a brief failure but a craft tradition mismatched to the medium."

04 · Four things that survive muted.

Across the cuts in our cohort that score in the top quintile of Get Noticed on Meta-Feed and TikTok, four creative moves are present significantly more often than chance. None of them is novel. All of them are systematically under-briefed on cuts inherited from sound-on traditions.

  • Caption that promises, not narrates. Top-third pinned caption stating the value, not the narrator's words. The caption should be redundant to the audio, not a transcript of it. Cuts where the caption alone delivers the proposition score +9 Get Noticed vs cuts with transcript-style captions.
  • Visual brand cue inside 1.5s. A distinctive non-logo cue (colour wash, mascot, packaging silhouette, see Report 02) that lands the brand handshake without audio. Cuts with visual cue under 1.5s carry a +11 Brand Impact lift on muted surfaces.
  • Face mid-expression. A human face mid-action with readable emotion, not a static neutral expression. Faces survive muted because emotional reading is sub-second and pre-verbal. Static talking-head openers without expression cost a median −8 Get Noticed.
  • Audio as bonus, not load. Music for atmosphere, sonic logo for closure, sound design for texture, none of it load-bearing. If the cut "doesn't work" with audio off, the cut doesn't work on Meta-Feed.

05 · The muted-watch approval checklist.

The simplest operational fix is a four-item checklist that every short-form vertical cut must pass before approval. Watch the cut muted, with the phone held at the angle a thumb-scroller actually holds it. Score it against four binary questions:

  • (1) Does the value proposition land in the first 2 seconds, muted? If the answer requires audio, the cut fails Meta-Feed and TikTok. Re-brief.
  • (2) Is the brand recognisable from a non-logo distinctive cue inside 1.5 seconds? If the brand is only the wordmark or only the end-card, the cut fails. Add a colour wash, packaging silhouette, mascot, or recurring talent inside the hook window.
  • (3) Is the top-third caption readable in under 600ms? Test with a phone held at arm's length, ambient light, viewer who is half-scrolling. If reading the caption takes longer than the median scroll-stop, the caption fails its job.
  • (4) Does the audio add or carry? If you mute the cut and the meaning collapses, the audio is carrying. Re-cut so the audio adds, atmosphere, closure, texture, but the meaning stands alone.
The discipline

"Watch every cut muted, phone held at thumb angle, before you watch it sound-on. Most of your distribution will see it this way. Brief for the median viewer, not the conference-room reviewer."

06 · Four worked examples.

EX-01 Luxury auto · Cinema sizzle reel, fails muted REBUILD · 38
Surface · Meta-FeedAudio · Engine roar + VOVisual brand cue · 5.2sCaption · None

A 30-second cinema sizzle reel re-cut at 15 seconds for Meta-Feed. The proposition lives in the voice-over ("from the most engineered driving machine in the world…"); muted, the cut is a sequence of moody drone shots with no proposition and a badge at the end. Get Noticed 36, Brand Impact 41. The same vehicle, re-cut as a creator-led "first drive" piece with caption-first composition and the badge at 1.0s, scored 67 on the second pass. Same product, different brief.

EX-02 Finance · VO-led explainer, fails muted REBUILD · 42
Surface · Meta-FeedAudio · Narrator monologueVisual brand cue · 6.4sCaption · Transcript-style

The textbook finance failure mode. A 22-second cut built around a narrator explaining a savings product, with a transcript-style caption that runs five lines deep. Muted, the cut is a sequence of stock-footage scenes with unreadable captions. Get Noticed 33, Sell Proposition 38. The category-typical "let me tell you about saving money" opener costs more on Meta-Feed than on any other surface.

EX-03 Beauty · "Caption-promise", survives muted SCALE · 74
Surface · TikTokAudio · Lo-fi ambientVisual brand cue · 1.2sCaption · "skin reset in 3 nights"

A UGC-native skincare cut with a top-third caption that delivers the proposition at frame 1 ("skin reset in 3 nights"). The audio is lo-fi ambient, atmosphere, not load. Visual brand cue (the brand's signature serum colour wash) at 1.2s. Get Noticed 78, Brand Impact 71. Muted score is within 2 points of sound-on score, which is the marker of a cut built sound-off-first.

EX-04 Emotional CSR cut, fails muted REBUILD · 44
Surface · Meta-FeedAudio · Score + emotional VOVisual brand cue · End-card onlyCaption · None

A CSR narrative that depends entirely on emotional voice-over to land its purpose claim. Muted, the cut is a sequence of beautiful but ambiguous human-portrait shots with a logo at the end. The narrative is invisible. Get Noticed 40, Sell Proposition 39. CSR cuts have the largest sound-on-to-sound-off score gap of any sub-category in our pool: the medium they are emotionally designed for is the medium where mute is the default.

07 · Implications by stakeholder.

// For the CMO

Mandate a muted-watch approval gate on every short-form vertical cut. The gate is four questions and takes under two minutes per cut. The penalty for missing it averages 12 composite points across your spend; the gate pays back from the first quarter it is enforced. Audit auto, finance, telco and CSR cuts first.

// For the brand team

Your brand handshake cannot live in the spoken brand name. Build the visual cue library, colour, mascot, packaging silhouette, recurring talent, and brief at least one into every cut at < 1.5s. The wordmark works as confirmation; the distinctive cue does the early recognition work.

// For the agency

Stop briefing "sound-on hero, sound-off fallback." That order produces the −15 penalty by design. Brief sound-off-first: the cut must work muted, then audio adds atmosphere and closure. The same shoot supports both, the order of the brief, not the budget, is the lever.

// For the media buyer

Pull spend on cuts where Get Noticed scores within 3 points of the floor on Meta-Feed. These are silent bleeders, the cut looks fine on completion (the small share of viewers who hear it complete) and bleeds Get Noticed across the muted majority. Reallocate to caption-first cuts before increasing budget.

// Three things to change tomorrow
01
Watch every cut muted before approval
Phone at thumb angle, ambient light, four-question checklist. The fastest fix in the report.
02
Move the value proposition to the caption
Top-third pinned, readable in <600ms. Audio becomes bonus, not load. The caption is now load-bearing.
03
Audit auto, finance, telco and CSR cuts first
The four sub-categories with the largest sound-on-to-sound-off gap. Highest leverage for the lowest brief change.

08 · Method.

Method. 1,110 ads sourced from Meta Ad Library, TikTok Creative Center, TikTok Ad Library, Google Ads Transparency Center, and manual editorial curation between 2025-Q4 and 2026-Q2 capture. Each scored through the RoastIQ pipeline against benchmark pool v.2026-05 (n=2,047) via Vertex AI Gemini 2.5 Flash/Pro multimodal + Google Video Intelligence (shot/label) + Speech-to-Text (transcription) + TranSalNet-class saliency, with Zod-validated structured output. Sound-on dependency is detected via a paired audio-only / vision-only scoring pass on each cut, with the delta thresholded against an annotated 180-cut training set (Cohen's κ 0.82). Mute rate figures are platform-published where available (Meta, TikTok) and modelled against our pool otherwise (Reels, Shorts). Confidence label HIGH on TikTok and Meta-Feed (n > 500 per surface, ρ +0.31 OOS TikTok engagement). MEDIUM on Reels and Shorts (sub-cohorts under 300). 5 KPIs: Beat the Skip (25%) · Get Noticed (20%) · Brand Impact (20%) · Sell Proposition (20%) · Build Brand (15%). We do not scrape.

What we do not claim. The penalty figures are model predictions calibrated against public engagement and click-intent signals. They do not measure in-market sales, ROAS, attributed conversion or brand recall. Reels and Shorts confidence is MEDIUM because the platform-specific cohort is smaller than the 500-ad threshold for HIGH confidence.

FAQ.

What percentage of short-form ad views are muted?
52% of TikTok For You feed plays, 64% of Meta-Feed plays, roughly 49% of Instagram Reels, and 38% of YouTube Shorts. Meta-Feed is highest because the surface autoplays muted by default for the largest share of users. Mute is the default, not the edge case.
What is the composite penalty for sound-on creative?
−15 Get Noticed points on Meta-Feed, −12 on TikTok, −11 on Reels, −7 on YouTube Shorts. Secondary effects on Sell Proposition (spoken claim missed) and Brand Impact (spoken brand mention missed). The penalty matches the mute-rate ordering across all four surfaces.
Which categories lose the most when muted?
Auto (−15), Finance (−14) and Telco (−13), categories with creative traditions built around voice-over explainers. Beverage (−7), Beauty (−9) and Snacks (−9) lose the least because their visual grammar already does the proposition work.
What survives a muted watch?
Four moves: a caption that promises rather than narrates, a visual brand cue inside 1.5s, a face mid-expression with readable emotion, and audio treated as bonus rather than load. Cuts that pass all four close the sound-on-to-sound-off gap to under 3 points.
Is YouTube Shorts really easier on sound-off?
Slightly. The Shorts mute rate is structurally lower (38%) because the surface inherits viewers from longer-form sessions where sound expectation is higher. The −7 penalty is the smallest of four surfaces, but Shorts is not a free pass. Bottom-quintile Shorts cuts still underperform on Build Brand.
What is the muted-watch approval checklist?
Four binary questions: (1) Does the value proposition land in the first 2 seconds, muted? (2) Is the brand recognisable from a non-logo cue inside 1.5s? (3) Is the top-third caption readable in under 600ms? (4) Does the audio add or carry? If audio carries, the cut fails Meta-Feed and TikTok by design.
What is the confidence label on this report?
HIGH on TikTok and Meta-Feed (n > 500 per surface, ρ +0.31 OOS TikTok engagement). MEDIUM on Reels and YouTube Shorts where the sub-cohort sample is under 300. Sound-on dependency detection validated at Cohen's κ 0.82 against a 180-cut annotation set.

Sources. All ads in the underlying pool are sourced from official public transparency tools (Meta Ad Library API, TikTok Creative Center, TikTok Ad Library, Google Ads Transparency Center) or manual editorial curation with documented licence terms. We do not scrape. Pipeline. Vertex AI Gemini 2.5 Flash/Pro (multimodal) + Google Video Intelligence (shot/label) + Speech-to-Text (transcription) + TranSalNet-class saliency, all scored against benchmark pool v.2026-05 (n=2,047) via Zod-validated structured output. Validation. Held-out OOS Spearman ρ +0.31 TikTok engagement (n=700), +0.30 TikTok CTR (n=691), +0.32 YouTube view counts (n=403). Construct. We predict engagement and click intent against public behavioural outcomes, not sales, ROAS, attributed conversion or brand recall. Reproducibility. Every chart is pinned to model_version and benchmark_pool_version.