Is Your X Feed Stuck on Repeat? Here’s What Went Wrong

Grok Logo

If you’ve opened X recently and felt like you’re watching the same movie on loop, you’re not alone. That unsettling sense of déjà vu isn’t just in your head but the result of one of the platform’s most ambitious algorithm changes to date.

The Big Announcement: Grok Takes the Wheel

In mid-October 2025, Elon Musk made a bold promise: X’s recommendation system would go fully AI-powered within four to six weeks. The culprit? Grok, X’s in-house AI that was supposed to revolutionize how we experience our feeds.

“The X recommendation system is evolving very rapidly,” Musk declared. “Grok will literally read every post and watch every video (100M+ per day) to match users with content they’re most likely to find interesting.”

Elon Musk quote tweet on the X algorithm

For years, smaller accounts complained about posting quality content into the void, with nobody seeing it. Grok was positioned as the great equalizer and AI that would surface great content regardless of follower count, finally giving new creators a fighting chance.

Users would even be able to adjust their feeds just by asking Grok directly. It was supposed to be personalization perfection, delivered by AI that understood you better than you understood yourself.

The Grand Vision: What X Expected

X’s leadership painted an optimistic picture of the Grok-powered future. An AI that would eliminate all hand-crafted rules and heuristics, replacing them with pure machine learning that processed over 100 million posts daily.

The goal was noble enough: help users escape the mainstream algorithm’s battlegrounds and find their niche communities. Whether you were into Formula One, molecular gastronomy, or competitive dog grooming, Grok would connect you with your people.

For content creators, the promise was even sweeter. The algorithm would focus on content quality rather than follower count, theoretically democratizing reach. Small accounts with great content could finally break through the noise.

The Rollout: A Timeline of Transformation

The change didn’t happen overnight. By late November 2025, Grok’s influence began appearing across the platform. First, it took over the “For You” feed which is an algorithmic timeline that already competed with the chronological “Following” feed.

Then came the controversial part: in late November, X made Grok the default for both feeds. Even your carefully curated “Following” feed, which had always been chronological, was now being ranked by AI. The platform started defaulting users to “For You” even if they’d closed the app viewing the“Following” tab.

The chronological option wasn’t eliminated entirely (you could still access it through settings), but the message was clear: X wanted everyone on the AI-powered experience, whether they liked it or not.

The Reality Check: When AIGets Stuck in a Loop

Here’s where things got weird. And by weird, I mean repetitive.

Users quickly noticed their feeds developing an unsettling uniformity. The same types of posts appeared over and over, creating an echo chamber effect reminiscent TikTok’s algorithm at its most aggressive.

The pattern was unmistakable. Feeds became dominated by specific tweet formats that seemed to trigger Grok’s recommendation engine:

“Tell me the lore behind your…” (profile picture, username, bio)

“How do I fit this into my header… Nvm” (usually accompanied by an image)

“My pinned tweet says otherwise” (the engagement bait heard round the platform)

These weren’t isolated incidents. Scroll through X for five minutes and you’d see the same post structures repeated by different accounts, creating a bizarre sense of algorithmic homogeneity. It was as if Grok had identified certain content templates as universally engaging and decided to flood everyone’s feed with variations on the same theme.

The complaints rolled in fast. Users compared the experience to TikTok’s notorious filter bubbles, where the algorithm can trap you in increasingly narrow content silos. Except on TikTok, at least the content looked different. On X, even the phrasing felt copy-pasted.

The Double-Edged Sword: Positives and Pitfalls

To be fair, the Grok algorithm wasn’t all bad. Some users reported genuine improvements:

The Wins:

• Smaller accounts did see increased reach when their content resonated
• Niche communities became easier to discover for those with specific interests
• The algorithm successfully reduced some rage-bait issues that had dominated feeds
• For users who engaged deeply with specific topics, content relevance improved

The Losses:

• Real-time information flow suffered as timeliness took a backseat to predicted engagement
• The “Following” feed lost its chronological nature, frustrating users who wanted to see posts from accounts they deliberately chose to follow
• Engagement farming posts dominated feeds as creators learned to game the algorithm

Perhaps most troublingly, there was the technical glitch Musk himself acknowledged on the “All-In Podcast.”

The algorithm’s interaction weight had been set too high, meaning any engagement; a like, a reply, even a save would flood your feed with similar content. Musk claimed this was fixed, but users continued reporting the repetitive content problem weeks later.

What Now?

X continues to iterate on Grok, promising improvements and adjustments. The platform has added semantic search and claims to be fine-tuning the balance between engagement and content diversity. Whether these adjustments will address the repetitive content problem remains to be seen.

For users, the chronological option still exists, and you can theoretically train the algorithm by aggressively engaging with content you want to see.

The Bottom Line

The Grok algorithm saga is a reminder that even the most sophisticated AI can’t replace thoughtful product design and user choice. X took a big swing at improving content discovery, and while some users benefited, the execution revealed the risks of prioritizing engagement over everything else. As social platforms continue their AI arms race, the real winners will be those who remember that algorithms should serve users and not train them to serve the algorithm. Until then, maybe keep that chronological feed option bookmarked. You may need it from time to time.