Algospeak

Content-Moderation-Evasion Substrate as Platform-Substrate Brand-Strategy

Also known as: Algorithmic Speech · Content Moderation Evasion · Unalive Speech · TikTok Algospeak · Algospeak-Substrate Cycle

Algospeak is the creator-driven linguistic workaround that emerged after platforms started algorithmically suppressing or removing content containing certain words. "Unalive" instead of suicide or kill. "Seggs" instead of sex. "Le$bian" with the dollar sign breaking automated detection. "Mascara" as code for sexual partners. "Pew pew" for guns. The pattern goes back to the early internet but accelerated dramatically on TikTok between 2020 and 2024 as the platform's content-moderation systems matured. Algospeak matters for brands because the rules of speech on the largest cultural platform are now shaped by what algorithms detect, and creators have collectively developed a parallel vocabulary that brands have to recognize whether they participate in it or not.

The intellectual foundation crosses platform studies and content-moderation scholarship. Tarleton Gillespie's Custodians of the Internet (Yale, 2018) established the analytical frame for thinking about platforms as moderators of public speech rather than neutral conduits. Sarah Roberts's Behind the Screen (Yale, 2019) — based on her UCLA Information Studies research — documented the human-labor side of content moderation that the algorithmic layer is steadily replacing. Robyn Caplan's Data & Society work since 2014 has tracked how content-moderation policies translate into creator-economy effects. The contemporary practitioner anchor is Taylor Lorenz's April 8, 2022 Washington Post piece "Internet 'algospeak' is changing our language in real time, from 'nip nops' to 'le dollar bean'" — the article that named the pattern in mainstream press and made the term a working part of platform-strategy vocabulary.

How it works

Algospeak operates on the gap between what platforms say their content rules are and what their algorithms actually detect. The detection layer is faster than the policy layer; the creator-language layer is faster than both.

Algorithmic detection vs. policy intent. TikTok, Instagram, and YouTube have community guidelines that are written in human language — sex, suicide, drugs, weapons, hate. The detection systems are trained classifiers operating on tokens, audio, and visual cues. Creators discovered early that the classifier doesn't recognize "unalive" as suicide, doesn't recognize "le$bian" as lesbian, doesn't recognize "mascara" as a coded reference to sexual encounters. The vocabulary expanded to fill the gap.

Suppression vs. removal. The most-consequential platform action isn't removal; it's reduced reach. Content that triggers a moderation flag often stays up but stops being recommended. Creators can't see the suppression directly — they only see traffic dropping. Algospeak emerged partly as a defensive response to the unverified suspicion that specific words were costing creators reach.

Creator economic incentive. Creators whose income depends on platform reach face concrete financial pressure to evade suppression. The vocabulary spreads not because creators want to be coy but because they have rent to pay. This is what makes algospeak structural rather than stylistic — the underlying force is the platform's commercial relationship with creator labor.

A 2026 wrinkle: AI-mediated content moderation has industrialized the cat-and-mouse cycle. Newer classifiers can detect "unalive" as a suicide proxy, which means the workaround has to evolve. Some words enter and exit algospeak vocabulary in months as platforms catch up. Detection Asymmetry describes how the speed differential between platform detection and creator workaround shapes the vocabulary's churn.

Variants

Death-and-self-harm vocabulary

The most-discussed and most-defended cluster. "Unalive" is the canonical example, used both as verb (to unalive oneself) and as descriptor. The vocabulary serves a real function for mental-health creators who want to discuss suicide and self-harm publicly without their content being suppressed; the function is contested by suicide-prevention researchers who note that the indirection can also obscure resources audiences need to find.

Sex-and-sexuality vocabulary

"Seggs," "spicy," "corn" (for porn), "mascara" (as coded reference), "the deed." The vocabulary is largely playful and the indirection is mostly low-stakes. The dollar-sign-in-LGBTQ-words variant ("le$bian," "ga¥") emerged in response to specific algorithmic suppression of LGBTQ content that platforms publicly denied was happening.

Violence-and-weapons vocabulary

"Pew pew," "unalive" again in violent context, the broader gun-content vocabulary. The variant is most common in news-and-commentary content about violence, where creators need to discuss events without triggering policies designed to suppress glorification.

Drug vocabulary

"Yerba" for marijuana in some contexts, "ouid" as a typographically obscured spelling, broader herbal-and-coded-reference vocabularies. Cannabis content sits in a particularly difficult moderation zone because the legality varies by jurisdiction and platforms tend to apply the most-restrictive rule across all users.

Brand-engagement variant

Brands that try to use algospeak get caught fast. The vocabulary is creator-economic in origin, and brands borrowing it without standing reads as cringe-coded marketing. The few cases where brand engagement has worked are heavily ironic ones (Liquid Death, Duolingo at certain moments) where the brand voice already lives in the same self-aware register.

When it breaks

The primary failure is brand cringe deployment. Brands that sprinkle algospeak into copy without earning the register get treated as out-of-touch. The vocabulary is creator-native and the audience reads brand uses as borrowed culture. Tourist Marketing describes the structural pattern.

The second is vocabulary obsolescence. Words enter and exit algospeak vocabulary as platform classifiers update. Brands that build campaigns around specific terms ("unalive Tuesday!" "seggs sale!") risk being permanently dated to a particular six-month window of platform-detection state. The vocabulary churn is faster than brand-calendar planning cycles.

The third is platform-rule conflict. Platforms periodically crack down on algospeak directly when the vocabulary becomes legible enough that the policy layer wants to close the gap. Content using the workaround can be retroactively flagged. Brands that built creator partnerships around algospeak content carry exposure when the rules shift.

The most expensive failure is trust collapse with audiences who depend on the vocabulary. Mental-health, LGBTQ, and abortion-access creators rely on algospeak to circulate information audiences need. Brands that deploy the same vocabulary commercially risk diluting the protective function — making it more legible to platform classifiers and harder for the original users to evade detection.

In the wild

Played straight. A creator-economy brand deploys algospeak in voice that matches its existing register. Liquid Death, Duolingo, certain creator-facing tools (Linktree, Beacons) sit closest to this lane.

Inverted. A brand explicitly avoids algospeak and uses platform-permissive language directly. Most enterprise and most luxury brands sit here without thinking about it.

Subverted. A brand comments on the algospeak phenomenon directly — work that addresses the platform-moderation gap as raw material. Rare and tricky.

Averted. A brand declines the category entirely. Default for B2B and infrastructure brands.

Canonical examples

TikTok content-moderation system (2020 onward)

ByteDance's TikTok is the platform where algospeak evolved fastest, partly because the For You Page algorithmic distribution is unusually consequential to creator success and partly because TikTok's content-moderation rules have been more aggressive than peer platforms. TikTok had approximately 1.5B+ monthly active users globally as of 2024 . The platform's rule changes (the 2020 self-harm policy expansion, subsequent ban on suicide-method depiction, ongoing LGBTQ-content moderation controversies) drove the vocabulary's evolution. Canonical case of a platform's moderation policy directly producing a parallel speech vocabulary.

"Unalive" (2020 onward)

The most-cited algospeak word. Adopted across mental-health TikTok roughly mid-2020 as suicide-and-self-harm content faced increasing suppression. The hashtag #unalive has accumulated approximately 4B+ views as of 2024 . The word now carries specific suicide-prevention vs. suppression-evasion tension — researchers including those at the American Foundation for Suicide Prevention have raised concerns that the vocabulary makes it harder for at-risk users to find direct mental-health resources. Canonical case of an algospeak word that became a contested object of public-health debate.

"Le$bian" and the dollar-sign LGBTQ vocabulary (2021 onward)

Adopted across LGBTQ TikTok after creators experienced specific suppression of their content. TikTok publicly denied suppressing LGBTQ creators while researchers including those at GLAAD documented that specific terms triggered demotion. The dollar-sign workaround spread as defensive infrastructure. Canonical case of a vocabulary developed in direct response to platform behavior the platform was officially denying.

Tarleton Gillespie, Custodians of the Internet (Yale, 2018)

The foundational scholarly text on platform content moderation. Gillespie's work at Microsoft Research synthesized a decade of platform-policy analysis into the framework that subsequent algospeak research builds on. The book has accumulated thousands of academic citations and is the standard reference in platform-studies courses. Canonical case of academic work establishing the analytical frame practitioners later use.

Sarah Roberts, Behind the Screen (Yale, 2019)

Roberts's UCLA-based research on the global content-moderation labor force documented the human-cost side of the moderation system that algospeak responds to. The book traced the work to commercial moderation firms in the Philippines and elsewhere where moderators viewed traumatic content for low wages. Worth naming because the algospeak phenomenon partly reflects the system's structural under-investment — both human moderators and algorithmic ones produce coarse-grained decisions creators have to route around.

Taylor Lorenz, "Internet 'algospeak' is changing our language" (Washington Post, April 8, 2022)

The article that named the pattern in mainstream press. Lorenz had been covering creator-economy and platform dynamics at The Atlantic, NYT, and Washington Post; this piece pulled together the linguistic phenomenon she'd been observing across platforms. The piece is the standard citation point for the pattern's mainstream legibility. Canonical case of a journalist naming a pattern that practitioners had been operating inside without formal vocabulary for it.

"Seggs" and the sexuality vocabulary cluster (2020 onward)

The lighter-stakes part of the algospeak universe. Used playfully across general-interest TikTok, the vocabulary is less defended than the death-and-self-harm cluster because the underlying content is less consequential. The cluster's churn rate is high — words enter and exit fashion as platform classifiers update.

Cannabis-and-drug vocabulary (multi-platform)

"Ouid" (typographically obscured weed), "yerba," herbal-and-coded vocabularies have circulated across TikTok, Instagram, and Reddit for years. The category is structurally hard for platforms to moderate because legality varies across user jurisdictions; the algospeak vocabulary functions as compliance scaffolding more than evasion. Canonical case of a moderation gray zone producing stable parallel vocabulary.

Algospeak is what creator economies do when the rules of speech are written by algorithmic detection rather than published policy. The vocabulary is faster than the platforms it operates against, and it's faster than the brand-marketing calendars that try to engage it. The smart move for brands is to recognize the vocabulary, understand which clusters are protective infrastructure (mental-health, LGBTQ, abortion-access) versus play vocabulary (sex, drugs at low stakes), and avoid borrowing either casually. The contemporary frontier is AI-driven content moderation, which is steadily closing the gap between platform policy and platform detection — and which is producing the next generation of workarounds in real time.

Related insights

Algospeak operates inside Cultural Momentum as a 2020-onward platform-moderation-driven creator linguistic pattern. Closest cousins are Algorithmic Curation (entry 63), which describes the recommendation-system layer that suppression operates within, and Chronically Online Discourse (entry 140), which describes the broader online-vernacular pattern algospeak sits inside. Brat Summer (entry 124), Demure Trend (entry 125), Underconsumption Core (entry 126), Loud Budgeting (entry 127), Stealth Wealth (entry 128), Mob Wife Aesthetic (entry 129), Eras Tour Economy (entry 130), Vibe Shift (entry 131), Dark Academia (entry 132), AI Companions (entry 133), Dumb Phone Movement (entry 134), Soft Life Movement (entry 135), Microtrend Velocity (entry 136), Hot Girl Culture (entry 137), Gorpcore (entry 138), Recession Indicator Meme (entry 139), NPC Streaming (entry 142), Soft Launch (entry 143), Quiet Quitting (entry 91), Brain Rot Aesthetic (entry 92), and Vibecession (entry 93) round out the contemporary cycle landscape. Microtrend Velocity (entry 136) describes the compressed-cycle dynamic algospeak vocabulary churn participates in. Detection Asymmetry describes the speed differential between platform detection and creator workaround that drives the vocabulary's evolution. Manufactured Authenticity describes the failure mode when brands deploy algospeak without standing. Tourist Marketing describes the broader pattern of brands borrowing creator-coded language without earning the register. Costly Signals and Commitment Durability describe the operational backing required when a brand wants algospeak engagement to feel earned. Authenticity Marketing is harder to pull off in algospeak contexts because the vocabulary is by design coded for in-group reception. Capital Inflation and Authenticity Inflation describe the long-run dilution as brand uses multiply. Influencer Marketing (entry 54), Creator-Brand Fit, and Creator-Owned Brands describe the practitioner channels. Generational Cohort Marketing (entry 77) describes how algospeak reads to Gen Z (native speakers) versus older cohorts (often opaque). Heritage Brand Positioning (entry 51) usually doesn't engage algospeak at all. Founder Mythology (entry 72) shows up when creator-coded brand voices lean on founder-as-creator authenticity. Synthetic Parasocial (entry 44) overlaps when AI-generated creator content uses algospeak vocabulary. Counter-Positioning (entry 74) describes how disruptor brands use platform-vernacular in ways incumbent brands can't. Crisis Communications (entry 80) covers the cleanup when brand algospeak deployment goes badly. Cancel Culture describes the reputational dynamics. Brand Personality (entry 83) and Naming Strategy (entry 87) describe the architectural choices brands face when their voice intersects with algospeak. Memetic Marketing, Spreadable Media, and Word of Mouth Marketing (entry 79) describe the diffusion mechanics of the vocabulary itself. Subcultural Capital describes the in-group recognition algospeak both relies on and produces. Cialdini Influence Principles (entry 99) — particularly unity — describes the in-group identity mechanic. Earned vs Paid Media (entry 89) describes the credibility differential between organic creator algospeak and brand-purchased uses. Marketing Mix Modeling (entry 84) struggles with algospeak attribution because the vocabulary is platform-specific. Signaling Theory gives the formal frame: algospeak is a separating-equilibrium vocabulary that distinguishes platform-fluent in-group members from out-group brand interlopers, and the cost of crossing the boundary uninvited is high. The pattern is that contemporary brand strategy operating on TikTok, Instagram, or YouTube has to recognize the parallel vocabulary even when not using it, because it shapes what words mean inside the conversation the brand is trying to enter.