Algospeak: How algorithms are transforming the way we communicate

As social media platforms increasingly rely on algorithms to moderate content, online communities are changing how they speak to evade their censors. Sensitive or suggestive topics such as suicide and sex have become “unalived” and “seggs.” When words are caught in the system, emojis act as loopholes — the watermelon emoji being used by pro-Palestinian creators as a symbol for Gaza or the grape emoji as a byword for sexual assault.

This new terminology is trickling offline and into the mainstream — a phenomenon linguist and content creator Adam Aleksic has dubbed “algospeak.” In his new book of the same name, he breaks down how this form of coded speech develops, its fringe influences, and what it means for the future of communication, on and offline.

The following conversation has been edited for length and clarity.
-----

Reuters: How do you define algospeak?

Adam Aleksic: Algospeak is the concept that algorithms are affecting how we speak. You can't say “kill” on TikTok because the word is suppressed, (so) creators instead chose to use words like "unalive" — it comes from a meme (in the animated series "Ultimate Spider-Man," in which the character Deadpool says “unalive” to avoid saying "kill"). Now, we have kids offline writing essays about Hamlet contemplating “unaliving” himself and that's an example of how this algospeak is bleeding into our real life. It's also maybe the most surface-level we can get with algorithms because we can just point to this and be like, “Oh, this is clearly algorithms causing language change." I think it's every aspect of language change. It's where words come from; it's how words spread.

Usage of these kinds of terms, especially offline, is often derided as “brain rot.” Do you see the rise of algospeak as related to that?

I want to separate the idea that language rots your brain. A brain rot to me is this meme aesthetic that people use. Like repeating words ad absurdum, that's brain rot.

I do think (brain rot) is a commentary on why we are getting so many of these words. It makes the words funny so you repeat them even more, which makes them even funnier. And why are they funny in the first place? Because they are there and they are there because of the algorithm — that's what brain rot as a meme aesthetic is to me. But it also points to how creators perpetuate these words into virality. Once we see a word is trending we'll hop onto that word.

Can you share an example from your research where algospeak actually changed the meaning or a tone of a conversation online?

Algospeak is so much more than (a response to) censorship. One example that I find fascinating is the word "preppy." To people (who are) older Gen Z and above, it means academic attire — Ivy League, Brooks Brothers, Ralph Lauren, that kind of aesthetic. Now, if you ask any middle school girl what the word preppy means to them, it's bright pink clothing with smiley faces on it.

How did that happen? It's partially just normal semantic drift; you go from these upscale brands like Ralph Lauren to more mid-tier retailers marketing to younger children. Preppy becomes associated with what middle school girls like to wear and all of a sudden we now have, since middle school girls like bright pink clothing, preppy becomes that word.

I think algorithms accelerated that. I use the example where several TikTok stores have popped up advertising preppy clothing, and they did that because they know this word was trending in that demographic. They hijacked a trending word on the algorithm for a chance for increased algorithmic visibility, for algorithmic optimization, and by using that definition of preppy, they pushed preppy further to the mainstream.

What is the most unsettling algospeak term you've come across?

I have a chapter on incel slang and I think it's pretty unsettling that a lot of words that middle schoolers are currently using — from “sigma" to the suffixes like "pilled,” “maxing" and "mewing” — these come from the manosphere. Sometimes (these terms) directly come from or are popularised by incel circles — this violent, misogynistic group of involuntary celibates. It seems concerning that their language reached the mainstream. I'm not trying to be an alarmist here; I think most middle schoolers don't know where these words come from. They just think it's a funny word to say with their friends, and that's honestly reassuring and refreshing.

Is this kind of fringe influence over mainstream language common?

Yeah, of course. We don't all just spontaneously start using a word at once. It starts with groups that have a shared need to invent new slang, and usually that's minority groups (or) fringe communities, because the mainstream already has mainstream language representing them perfectly. It's people who don't feel represented by language as a whole that need to come up with these new words. And, historically, some very prolific producers of language have been minority communities or fringe communities, like incels. The way we adopt slang words usually follows the conduits of what we see as funny or cool.

Who owns algospeak?

Owning a word is a very interesting thing. You go back to 2014, the phrase "on fleek," which was popularized through (the short-form video app) Vine, was the first time we have video popularising phrases. It was coined by this user Peaches Monroe, whose real name is Kayla Newman. Kayla never got a cent of royalties for this. She tried, repeatedly, to trademark the phrase, but couldn't until three years later and by then the trend was dead. In the meantime, [on fleek] was repurposed by brands. I consulted on a lawsuit a few months back where one influencer sued another influencer for stealing her vibe and part of that was the linguistic intonations of her manner of speaking. My recommendation there was that they're kind of both copying other people because you have to; this is how language works. We draw our idea of what language is from previous things and then we replicate that.

Do you see algospeak as a form of digital protest?

Absolutely. All language has multiple uses. I think this is definitely one of the uses. Also, there are two definitions of algospeak: the old definition is the "unalive" kind of stuff, the censorship avoidance. And then I tried to expand that definition, arguing that everything is algospeak. The genre of "brain rot" itself is a cry for help from the algorithmic oversaturation of trending slang. It's poking fun at the fact that these words are overrepresented, which already is a meta critique of the algorithm and (the) panopticon that we're in. You look at individual surveillance-avoidant language because you do have to recognize every single word is being surveilled, which is a crazy thing that we haven't really had in the past.

Is algospeak mostly an English-language phenomenon, or are you seeing similar patterns globally?

Totally globally. In Spanish, some people say "desvivirse," which had a previous meaning of "do your utmost.” Now it can also mean “unalive” in the same way (it does in) English. In Chinese, the word for censorship is censored, so people use a word like "harmony" in allusion to the Chinese government's goal of making a harmonious society. But then that started being censored, so people started saying “rivercrab" because it sounded similar; it was one tone off. Then that started being censored, so people started saying "aquatic product.” I find that it's alarming that the censorship exists, but it's beautiful that humans are always able to find ways to say what they want to say.

Do you see algospeak evolving into a kind of digital pidgin or could it vanish as content moderation tools get smarter?

We're always going to be one step ahead of AI. I strongly believe that because AI has a biased representation of what language is. Language is constantly being changed and updated, and you can't possibly have the AI catch on to that. There's a lot of stuff with context that only humans can get. I think humans are extremely resilient at finding ways to say what they want to say.