Is Linkedin an early look at our future ai hellscape?

First released:

Have you been on LinkedIn recently? It’s horrible.

Not in the usual way that LinkedIn is horrible, either. Horrible because suddenly, everywhere you look, generative ai tools are writing posts for people, replying on behalf of humans, giving prompts to encourage engagement, creating ideas for “collaborative articles”, contributing to collaborative articles, and generally filling the whole place up with gen ai ooze.

Here’s a perfect example.

An example of meaningless ai content on LinkedIn
Linkedin Ai Garbage2
Linkedin Ai Garbage3
Linkedin Ai Garbage4

I’ve obscured the identity of the person leaving these answers because… well because it’s quite embarrassing and quite possibly they didn’t even know those answers were being left (more on that in a second).

To me, this a perfect example of ai generated nonsense. It’s grammatically correct, professional, uses advanced but not too advanced language. You can skim it and come away feeling nothing at all. But take a moment to examine it carefully and you notice holes all over the place.

First, the topic of the article. It makes no sense. What, precisely, is a “feedback conversation in the field of Search Engines?” It’s exactly the kind of meaning-free slush LLMs come up with if you let them run wild.

Second, when viewed next to each other it’s plain as day that these “contributions” are ChatGPT responses:

  • They acknowledge the question in the first line of the response
  • They use awkward language like “elucidate”
  • They throw together semantically similar concepts in ways that make no sense

So, a gen ai article idea which makes no sense with automated contributions which also make no sense. What value is being created here?

We’ve written about the various quality issues with ChatGPT and other large language models, but if one place could be the right setting for gen ai ooze, it’s LinkedIn. The business-casual tone of most chatbots is the perfect fit, and their prohibitions on saying anything rude mean you don’t risk calling your boss a dickhead.

Also consider that LinkedIn is all-in on generative ai. Which makes sense; they are owned by Microsoft, which has its own ai platform in Copilot, and has invested big in firms like Mistral building similar systems. Currently LinkedIn is encouraging users to re-write posts with ai, and respond to posts using ai prompts. But that’s the tip of the iceberg when it comes to ai content on LinkedIn.

Having a big following on LinkedIn is a great way to grow your business, so naturally people want to fast-track their way to a big following. So if you can get an ai chatbot to reply to posts on your behalf and get more eyeballs on your profile, why not?

I created a quick bot based on GPT4 with these instructions:

When you receive a LinkedIn url, read the post and generate a short, polite reply expressing a lukewarm opinion on the subject of the post. Keep it short and professional.

And fed it this post by Brad Smith:

It responded:

Interesting to see Microsoft making such a significant investment in AI. It’s definitely a move that could shape the industry’s future. Thanks for sharing this update!

So far, so banal.

Many people have taken this idea and turned it into a product in its own right (although, like so many other ai products right now, they’re just wrappers for other people’s models). A quick search turns up Taplio, Hootsuite, Hyperwrite, MagicPost, EngageAI, CommenterAI, Hushl, Replai, Brand engine, Commentbuddy, Mentis, Linkedin Post Generator (inspired), Postfluencer, ContentIn, Buzzli, Luna, GrowMe, FeedBird, CopyTruck and many, many others.

These products promise to supercharge your LinkedIn engagements by not only coming up with things for you post, but also auto-replying to replies on your posts with more banality. Because more replies on a post is more juice to the algorithm, and begets yet more replies. But always remember, dear reader, that LLMs do not understand nuance, sarcasm, humour, context or a thousand other elements of human interaction and can make you look, generously, like a moron.

An example of a LinkedIn post designed to create engagement
An example of an ai-generated reply on LinkedIn

We’re cruising towards a LinkedIn future of ai posts with ai comments with zero human eyeballs involved. We might already be there. Thankfully, people are already calling it out and I would expect LinkedIn to act to remove this kind of content because, despite their full embrace of ai, it is making their platform undeniably worse.

Side note: Many of these system want to log into LinkedIn on your behalf without using proper authorisation. This is a massive security issue, don’t give them access to your account.

For me, this is an entirely backwards use of ai. Current ai systems do an average job of generalism, but can do a great job at very specific tasks, when set up correctly (see our tools). The promise of ai is to free us from busywork, but interacting with other people does not fall into that category. You should be using ai to file your taxes and manage your calendar, specifically so you can spend more time interacting with other people. Don’t cheapen human interaction for the sake of a few internet points.


Posted by

About the author