The big idea: An engagement bait translator

A big idea to filter out the internet's most annoying content. Want to learn more?👇

The big idea: An engagement bait translator

When Elon Musk took over Twitter (“X”), he introduced creator payments in an attempt to get people to post more. For a few months, my timeline was full of screenshots of posters flexing their giant payouts.

The result was what I call homogenization from monetization. Some entrepreneurs started selling classes on making money through Twitter using engagement bait, and the techniques spread. Very quickly, content homogenized into thinly veiled sales pitches and attention hacks.¹

What if we could remove the noise but keep the content? Introducing the engagement bait translator: an extension that scans for engagement bait and replaces it with straightforward language.

What is engagement bait?

Let’s use the introduction as an example. If I were an engagement maximizer, it might look like this:

🚨 Most people will ignore this post, but that’s why they’re stuck. 🚨

A few years ago, Twitter completely changed — and most people STILL don’t get it.

Elon Musk’s new payouts? They created a secret formula for making thousands per tweet. But the game is rigged. If you don’t know the three engagement hacks that gurus don’t want you to know, you’ll never grow.

99% of people will ignore this. But the 1% who take action? They win. Are you one of them? 👇

With our translator, it might look something like this:

“I am trying to sell you a course.”

The unfortunate reality of these techniques is that they do sort of work in hacking your brain. I hate-read almost every one I get — even when I know it’s a waste of time. AI reinforces this trend; I asked ChatGPT how to write a tweet to cross-post one of my articles, and it told me to do more engagement bait (and used a 🚀 emoji, which was truly disgusting).

So how do we break the spell? The LLM-based translator reviews each piece of content for typical engagement bait patterns — too strong of a hook, unnecessary mystery, and aggressive calls to action. A score is assigned, with a cutoff that’s adjustable by the user. If the content comes over the cutoff, the translator restates it plainly with a toggle to switch back. Suddenly, your timeline is a mix of no-nonsense and human-written nonsense.

A mockup of the future we’re all waiting for

Why does engagement bait happen?

We live in an attention economy. Getting your eyeballs on something has value, and the internet has gotten progressively better at locking in your attention. Most of that value is captured by the platform, but some portion — whether payments or dopamine from a new subscriber — goes to the creator too.

For creators, there’s a strong incentive towards the most engagement bait-y templates. It’s non-trivial creative work to write content of any length. A writer’s style develops through lots of output and seeing how their audience reacts; this can lead to a unique voice (or audience capture if you’re not careful).

But that process is hard, and it’s tempting to mimic what works. And the hard way may not even be optimal; scrolling social media is not always intended to be a cognitively intense activity. Engagement bait provides a familiar structure that’s easy to passively read, which may actually be a better experience for most users. For a share of creators, that can be enough.

How do we drive adoption for a translator?

At least in principle, there’s demand for products to help reclaim attention and ability to focus. Digital detox programs are all over the place — books, organized programs, and even rehab-like programs to get your brain back. The digital detox app market is allegedly at $400 million dollars. It seems like the financial incentive is already there; the hard part is convincing someone to pay for it.

Today’s ad blockers are the closest analogue for this solution — you don’t quite abandon technology, but rather intercept content that hijacks your attention.² You could imagine a model running locally on your iPhone, working across all of your apps. As models get smaller and faster — see Llama — this will be possible pretty soon, if it isn’t already. You could even imagine a privacy focused company like Apple embracing this as a standard feature, like a detox layer for the entire internet. It would be the IDFA ad apocalypse on steroids.

But the platforms won’t like it

The problem is that platforms are going to hate this! LinkedIn has a huge amount of engagement bait; Imagine if your feed went from

"I took a chance on someone today.

They didn’t have a fancy degree. They didn’t check all the ‘required’ boxes. What DID they have? Grit. Passion. A hunger to learn.

Too many companies overlook incredible talent because of outdated hiring practices. Not me.

Today, we hired someone based on potential, not pedigree. And I couldn’t be prouder."

To

“I gave someone a job offer; they said yes”

Engagement would plummet, even if you would personally waste less time. It’s not just LinkedIn; platforms in general are pivoting towards engagement bait and AI-generated content. Facebook in particular has embraced AI, with some of their top posts being what could generously be called low-quality engagement bait.

The famous Mark Zuckerberg endorsed bread horse

Platforms could react to a bait translator by hampering your experience — Google allegedly does this with ad blockers on Youtube. There’s a big incentive for platforms to control your experience by punishing user-end modification; maybe LinkedIn starts only showing job postings you’re unqualified for.

These companies could even start using their troves of data to train a generative AI that tells creators if their content will trigger the engagement bait translator, allowing them to do the minimum to circumvent it. And that could be the start of the real problem.

The engagement bait arms race

Content pirates aren’t just going to give up if this got adoption — they’re going to adapt. Let’s think through the evolution of engagement bait:

1. Basic Bait

Emojis and an insinuation that you’re inferior if you don’t click. “If you’re serious about growth, keep reading 🔥”

2. Stealth Bait

We are giving you some secret, arcane knowledge: click to learn more. “Nobody talks about this, but it’s the key to success.”

—

We are here, and this what we’re solving for. But if a translator got wide adoption, there would be a reaction:

3. AI-Assisted Bait

AI models learn to circumvent engagement bait filtering, reopening the door to easy content. We get into a cognitive arms race as new and more insidious forms of engagement bait are discovered and spread, until they too become clichés.

4. Total AI vs. AI Content Filtering

AI generates engagement bait; the bait is filtered by the translator. Nobody is creating or consuming the actual content. The ad revenue keeps flowing. And deep in a forgotten server farm, a generative model posts a new thread:
"10 productivity hacks that changed my life. Number 7 will SHOCK you!"

Another AI model reads it, translates it from engagement bait, and send it on to the reader (who is also an AI model).

Where it all goes

I’d hypothesize there’s two potential end points for content. The first is that AI learns how to write better content than humans themselves — developing ideas and writing so compelling that people begin to prefer it to human-written material. There’s some evidence that this could be the future; people are already falling in love with ChatGPT. I don't think it's too crazy to imagine that AI will eventually surpass 90% of human writers in quality. Considering how much of what’s on social media falls below that threshold, maybe it wouldn't be so bad?

Alternatively, signaling you’re not an AI could become increasingly important in the content market. Validation that you’re a human is a signaling problem. Surveillance could be an answer — all content must be written on camera, with active monitoring to ensure it’s really an original human thought. But the most likely solution is probably leaning fully into social capital. I wouldn’t be surprised to see a resurgence of invite-only forums, where you specifically validate the humanity of each person you invite. You may be creating little tyrants by empowering these gatekeepers, but they’ll at least be human tyrants.

Of course, proving humanity doesn’t guarantee genuine human interaction. In the not so distant future, a grizzled old B2B lead generation marketer pulls out their laptop and goes to their human-only forum:

“AIs don't want you to know this, but the secret to growth is authenticity. A 🧵👇”

Official idea rating:

4.8/5. I think we really will go through a phase of LLMs filtering LLM generated content in the future. But even without LLMs, the financial incentives between platforms and creators will continue to encourage tactics to hijack your attention. Until those incentives change, it’s up to the user to find a way to filter the bait out — or give in to the hijacking of their attention span.

Coda: after I finished writing this article, I found a guy on Hacker News that already had a similar idea and prototyped it. I promise I had the idea independently Daniel — consider me an early customer if you go all in on this. If you don’t, my only choice is to optimize this article for engagement and find someone who will. A 🧵👇

---

1. You can start to see some of the same impulses on Substack; I see Notes on how to build a big newsletter following to monetize (key words: find a niche! Show value! Cross-promote! You too can escape the 9-5). That said, still mostly better than other platforms after some curation.

2. I personally put my phone in black and white mode, it breaks a lot of engagement loops