What would happen if you could talk directly to the recommendation algorithm? X is about to find out

If you’re online long enough, you’ll hear it whispered: The algorithm. The algorithm decides whose content is recommended to viewers. Whose content is surfaced in searches. It’s almost treated like a mercurial beast–one you have to placate, and hope you never fall afoul of.

YouTube has made some efforts to demystify the algorithm (its key piece of advice? Please the audience and the algorithm will follow), but across the web, creators are still wary of the systems that dish out their content to viewers, worrying their videos are suppressed or not shown to those who’d be interested.

On the flip side, viewers have their own rough experiences. We here at Tubefilter have seen YouTube’s Shorts algorithm recommend a fresh account pure slop irrelevant to any normal human interests. We’ve also seen it recommend right-wing content like Andrew Tate right out of the gate. Despite that, algorithms at YouTube, TikTok, Instagram, and more are in general pretty good at catching on to what viewers like and giving them more of it.

But there’s one platform having a particularly difficult time with its viewer-facing algorithm.

X.

Since being bought by Elon Musk, X–formerly Twitter–has seen its main feed algorithm deteriorate. It’s been (allegedly) tweaked to push posts from Musk and amplify right-wing views, and many users report a deluge of hate speech and violent videos showing gruesome fights and animal deaths.

Now, nearly three years after Musk’s purchase, X seems to have finally put two and two together and realized some users may not like seeing this content.

To fix that problem, it’s leaning hard into another arm of Musk’s business: generative AI. According to Musk X’s recommendation algorithm will soon be “purely AI.” (What does that mean and how is it different from the machine learning systems most recommendation algorithms utilize? No explanation.) It appears that between now and then, X plans to roll out tweaks to the feed that will make it actually reflect what individual users want to see.

“The goal for your X timeline is to get out of the mainstream algo and the political crusades and find your niche,” Nikita Bier, X’s Head of Product, said in a tweet. “You should be able to post about your interests and have friendly, relevant people chime in.”

He also added a knowing nod about the violent content, saying, “If you’re seeing gas station fight videos, your account is not ramped up yet. We are working everyday [sic] to fix this.”

Musk chimed in saying X is making “significant progress along the way” to full AI-dom. “We will open source the algorithm every two weeks or so,” he said.

The bit people should be paying the most attention to, however, is his final note: “By November or certainly December, you will be able to adjust your feed dynamically just by asking Grok.”

Grok, of course, is X’s inbuilt chatbot, which like X’s algorithm has a history of being (allegedly) tweaked for Musk’s purposes. If X implements this change, it’ll be a first-of-its-kind system where users can ask a platform in plain language to change what kind of content they’re being served. 99% of major platforms offer “don’t recommend content like this/this channel/etc” buttons, but those have limited effect, and must be implemented one at a time. If X’s algorithm works the way Bier and Musk describe, users will be able to make large-scale, personalized changes.

Will this work as intended? Maybe! Musk’s various tech companies don’t have the best track records with pulling things off at first go, but this concept is sort of a natural evolution from the world’s current obsession with generative AI–and X isn’t the only platform investigating it. Social Media Today reports Meta‘s Threads is experimenting with something similar, a tool that would let users change their feed by tagging @threads.algo in a post. Meta hasn’t confirmed this potential change, but it is similarly on the gen AI train, so we wouldn’t be surprised if plain-language, chatbot-based algorithm changing is on its roadmap, too.

Either way, we’ll have to wait till November or certainly December to see how this tool plays out in real time. If it’s successful, will we see YouTube and other video-based sites roll it out? Would that have an impact on creators’ views? Much to wonder…