Algospeak
A reader's summary of the coded vocabulary — “unalive,” “seggs,” “le dollar bean” — that creators built to slip past automated content moderation without knowing exactly which words trigger it, its deliberate echo of Orwell's Newspeak, and the April 2022 press coverage that carried it from creator slang into general vocabulary.
The term at a glance
Algospeak names the practice of substituting code words, deliberate misspellings, or near-homophones for terms that automated content moderation systems are believed to penalize — reducing a video's reach, demonetizing it, or excluding it from search and recommendation surfaces. It is not one fixed vocabulary but an ongoing, decentralized adaptation: creators guess at what triggers suppression, since platforms rarely publish their keyword lists in full, and successful substitutions spread through a platform's creator community faster than any single list of them can be compiled.
Origin
The word is a portmanteau of “algorithm” and “speak,” and the second half is a pointed borrowing: it echoes Newspeak, the shrinking, state-controlled vocabulary in George Orwell's Nineteen Eighty-Four, engineered to make certain ideas difficult to even phrase. Applying that frame to platform moderation casts recommendation algorithms as a kind of automated censor that creators' language has to route around — a comparison creators themselves reached for, since the underlying behavior (adopting substitute words to avoid an unaccountable, opaque suppression mechanism) is structurally similar even though no single authority designed the vocabulary being avoided.
History and context
The practice preceded its name by a couple of years, developing organically among TikTok creators as short-form video platforms scaled up automated moderation and demonetization in the late 2010s and early 2020s. Substitutions like “unalive” for kill or suicide, and euphemisms for sexual-health and sexual- assault topics, spread through creator communities as informal, crowd-tested workarounds long before there was a settled word for the phenomenon. The Washington Post's April 2022 feature on “algospeak” — documenting terms from “nip nip” to “le dollar bean” — is the most-cited point at which the practice and its name reached a mainstream audience outside the platforms where it had already been ordinary vocabulary for some time.
Main ideas
A portmanteau built on a deliberate literary echo
Algospeak fuses "algorithm" and "speak," and the second half is not an accident — it deliberately calls back to Orwell's Newspeak, the controlled vocabulary in Nineteen Eighty-Four designed to make certain thoughts hard to express at all. The borrowed suffix frames the practice as a language reshaping itself under surveillance, even though here the reshaping is done by users, not an authority.
The vocabulary is inventive by necessity
Words get replaced for one reason: platforms' automated moderation systems flag or suppress certain terms, so creators substitute near-homophones, deliberate misspellings, or unrelated code words for topics the algorithm is likely to penalize — "unalive" for kill or suicide, "SA" for sexual assault, "seggs" for sex, and single-use community coinages like "le dollar bean" standing in for words the platform is presumed to be listening for.
It targets reach penalties, not just outright removal
Much of algospeak is aimed at a softer form of moderation than deletion: reduced recommendation, demonetization, or exclusion from search and "For You" surfacing, the same family of consequence covered by shadowbanning. Creators often can't tell which specific word triggered a drop in views, so substitutions spread as a precaution as much as a proven workaround.
Mainstream press coverage arrived years after the practice
The coded vocabulary had been circulating among creators, especially on TikTok, for a couple of years before it broke into general awareness. The Washington Post's April 2022 piece on "algospeak" is the most-cited point at which the phenomenon and its name reached readers outside the platforms where it was already ordinary practice.
It spreads faster than any list of banned words
Because the underlying moderation rules aren't published in full, and because a successful workaround gets copied instantly across a platform's creator base, algospeak evolves faster than any single glossary can track — a code word that works today can be flagged tomorrow, prompting another substitution.
It affects legitimate speech, not only rule-breaking content
Because moderation systems key on keywords rather than intent, creators discussing sexual health, mental health, historical violence, or moderation itself in good faith are caught by the same filters as bad-actor content, which is the main reason the workaround vocabulary extends well beyond anything actually against a platform's stated rules.
Critique
- Nobody can confirm which words actually trigger suppression. Because moderation criteria aren't published, much of algospeak is precautionary folk knowledge rather than confirmed workaround — creators substitute words based on shared suspicion and anecdote, not disclosed rules.
- It degrades plain discussion of serious topics. Mental health, sexual health, and violence are all topics people have legitimate reasons to name directly; routing that speech through euphemism to dodge a keyword filter makes ordinary, rule-compliant conversation harder to have clearly.
- The vocabulary itself can date or exclude a conversation. Terms specific to one platform's creator community age quickly and can be unintelligible outside it, meaning the workaround solves a reach problem for the in-group while adding a comprehension barrier for anyone else.
Impact
Algospeak is the linguistic symptom of the same moderation machinery covered elsewhere in this series: Shadowbanning names the reach-limiting mechanism creators are adapting their language to avoid, and the underlying discomfort with an opaque, unaccountable system shaping what can be said runs through Dead Internet Theory as well — both describe platform infrastructure that users can feel the effects of without being able to see or verify how it actually works.