If you've ever wondered why a harmless-looking word like “mascara” or “corn” keeps popping up in odd places on social media, you're in the right place. Today's subject is algospeak, the coded language people use to outwit social platforms' algorithms. We wrote this article in 2024 and updated it in October 2026.
01Algorithms as gatekeepers
On social platforms, algorithms are the gatekeepers. Facebook, Instagram and TikTok analyse the text, images and videos people post, and use AI to pick out keywords and patterns that break their community rules. The aim is to stop misinformation and offensive or inappropriate content.
The figures come from TikTok's Community Guidelines Enforcement Report for April–June 2026 and Meta's January 2025 announcement, updated in May 2025.
These systems are good at removing harmful content, but they can also censor important conversations by mistake. In January 2025, Meta announced that for less severe violations it would act only once someone reports the content, keeping its automated systems focused on illegal and high-severity violations.
Users, though, found a way round the filters long ago. That is how algospeak came about. Even emojis such as the aubergine and the peach have picked up new meanings along the way.
02Why we dodge the algorithms
Many people have good reason to talk about sensitive subjects online. Survivors of sexual assault, and people living with anxiety or depression, find support in online communities. Yet these are exactly the conversations algorithms often block.
In support groups, people share their experiences and offer each other advice. In mental health communities, talking about depression and anxiety is how people support one another. Automated censorship can do real damage here, cutting people off from the help they need.
03The code words
To say what they think without being penalised, users invented words that slip past filters which would automatically block the explicit terms. A few examples:
- “Mascara”
- Sex or, in some contexts, sexual assault.
- “Unalive”
- Death or suicide.
- “Accountant”
- A sex worker.
- “Leg booty”
- LGBTQ, an example The New York Times reported in 2022.
The code words change quickly. As platforms learn to recognise algospeak, users invent new terms to stay a step ahead. It's an endless game of cat and mouse, with each side adapting to the other's latest move.
04From leetspeak to algospeak
Algospeak looks new, but it isn't. It's an evolution of the leetspeak of the ’80s, when internet users swapped letters for numbers and symbols to get past moderation. On forums and in chats, “3” stood in for “E” and “$” for “S”. Ringing any bells? You've probably seen writing like this without knowing it had a name.
Leetspeak started in hacker communities and soon spread to other online subcultures. It was first used to talk about hacking and other illegal activity, then online gamers and other niche groups adopted it too.
Subscribe to the newsletter.
Like how we think? Get an article like this one twice a month.
05Protest and creativity
Coded language isn't only for everyday users. In countries where the internet is censored, protesters have used creative spellings to stay under the radar. During the 2019 Hong Kong protests, demonstrators used encrypted messaging apps and their own codes so the authorities couldn't intercept their messages. That linguistic creativity helped them organise despite the censorship.
06When the codes collide
The tug of war between platforms and users keeps producing new terms, and misunderstandings with them. In January 2023, actress Julia Fox made a seemingly insensitive comment about “mascara” on TikTok without realising the word was being used there to mean sexual assault. The backlash led her to apologise.
Episodes like this show how easy it is to get algospeak wrong. Users have to watch the context and the hidden meanings of words. Platforms, for their part, have to strike a balance between removing harmful content and letting legitimate conversations happen.
07In short
Algospeak shows how internet users adapt to automated censorship, and push back against it. It has grown big enough that in 2025 Adam Aleksic devoted a whole book to it: Algospeak: How Social Media Is Transforming the Future of Language, published by Knopf.
Next time you spot “mascara” or “unalive” in your feed, you'll know what's meant. And beyond the game with the algorithms, algospeak is a reminder of how much free speech matters online.
