Google's Jigsaw unit has released an enhanced version of its AI tool Perspective, which will rank how “toxic” a person’s social media posts are.

This Google AI Will Rate How Toxic You Are On The Internet And This Should End Well

3 min read

This article is for general information purposes only and isn’t intended to be financial product advice. You should always obtain your own independent advice before making any financial decisions. The Chainsaw and its contributors aren’t liable for any decisions based on this content.



The internet can be a dark, nasty place. With the arrival of AI, misinformation, as well as deepfake photos and videos have also made surfing the internet a more disturbing experience than ever.

Who’s to blame for the amount of AI-generated pollution online? It’s hard to point fingers at a single Big Tech player, but perhaps that doesn’t really matter now. This is because Google’s Jigsaw unit has released an enhanced version of its AI tool Perspective, which will rank how “toxic” a person’s social media posts are — it’ll even measure how others would be affected by a “toxic” post.

Perspective is taking on the roles of social media researcher, analyst, literature student and counsellor all in one.


Google’s Jigsaw AI

Google’s Jigsaw says the aim of Perspective is to counter misinformation, censorship, toxicity and violent extremism on the internet. Perspective was launched in 2017, years before the AI race that’s taking place today. This means a dramatic update is due, and Jigsaw has done just that.

Perspective is being offered as an API — akin to a third-party website plugin. You first have to access your Google Account, create a Google Cloud account with it, and then you can download the API.

How does it work?

The tool mainly analyses text comments, be it comments on a news website, forum, blog post or chat message. Because, let’s be real, the comments section is where people go unhinged and leave the rudest, foulest opinions nobody asked for. The AI tool will then predict the “nature” of the comment and assign a label.

Is your comment downright rude or disrespectful? Congratulations, you earn a “TOXICITY” or “SEVERE_TOXICITY” label.

Is your comment targeting someone on the basis of their identity? Congratulations, you earn an “IDENTITY_ATTACK” label.

Is your comment chock-full of vulgar swear words? Congratulations, your expansive vocabulary of curses earn you a “PROFANITY” label.


What if you’re a responsible social media punter who doesn’t leave rude comments online? Perspective will still assign you a label: “AFFINITY”, “COMPASSION”, “CURIOSITY”, “RESPECT”, and so on. Congratulations!

Anyway, for posts deemed to be toxic, the AI tool will also assign it a score between zero and one. Zero means “not likely to be toxic”, whereas one is “very likely to be toxic.” A score of 0.8, for example, would be toxic, and 0.1 would be mildly toxic.

Jigsaw AI

As impressive as Perspective appears to be, Jigsaw cautions that limitations and errors still exist. The tool may not be able to analyse comments that are too long, and some languages are not supported. Nonetheless, Perspective is a handy tool for academics and researchers to study our relationship with social media content and algorithms.