When Elon Musk announced on October 19, 2025, that he wanted to replace the entire recommendation algorithm of a platform with hundreds of millions of users with a single AI model within four to six weeks, that was no footnote for social media managers. It's a structural turning point for every company that uses X (formerly Twitter) as a marketing channel. In the meantime, the announcement has turned into an ongoing, far-advanced process: since late November 2025, Grok has also been sorting the once-familiar "Following" feed by predicted relevance instead of chronologically, and in January 2026 xAI published the code of the new algorithm on GitHub under the name "Phoenix," or x-algorithm. For your business marketing, this means: the rules that determine whether your posts get visibility have changed fundamentally - and they may keep changing.

In this article, we break down what is actually known from the available reports, the disclosed code, and statements from X and xAI, what counts as established, and what should still be treated with caution. And above all: what does this mean, concretely, for the content strategy of a Mittelstand company on X?

What's fundamentally changing at X right now

Until now, feed sorting on X, as with most social networks, was based on a mix of hand-written rules - so-called heuristics - and individual machine-learning components. In his post from October 19, 2025, Musk announced he would fully replace this hybrid system:

"We are aiming for deletion of all heuristics within 4 to 6 weeks" – Elon Musk, X post from October 19, 2025

Concretely, this means: the entire feed algorithm is to be controlled exclusively by a Grok-based transformer model. All manually coded rules that previously determined how factors like account age, follower count, or topic area are weighted are set to disappear. Instead, an AI model decides which content is shown to which user - a model that, according to Musk, evaluates more than 100 million original posts daily, plus roughly 100 million replies and around 300 million quotes or reposts.

Important for context: the "4 to 6 weeks" announcement from October 2025 was the starting shot of a process that has since continued in several stages - not a one-time event. The switch of the Following feed at the end of November 2025 and the open-source release in January 2026 are the most visible milestones so far. Whether 100 percent of the old heuristics have actually been removed by now can't be conclusively verified from the outside - but the direction is unmistakable: X is becoming a thoroughly AI-curated platform, on which a single model decides what people get to see.

It's also worth noting that users can still manually switch to an unfiltered, chronological view. But the algorithmic feed is the default setting - and most users won't change it. For companies, that means: don't assume your followers see your posts chronologically. You're competing for algorithmic relevance anew with every single post.

The new Grok algorithm: how it distributes reach

With the open-source release in January 2026, outside developers and analysts got their first look at how the system actually works. The central building block is a transformer model that doesn't just predict a single signal - such as "will this user like it?" - but simultaneously forecasts around 15 different engagement signals. From these predictions, the system calculates a relevance score that determines how content is sorted in the feed.

This is a fundamental difference from earlier, rule-based systems: it's no longer about whether a post meets certain formal criteria, but about how likely it is that an individual user will react to that specific post with certain actions. X justifies the move, among other things, as an attempt to solve a long-standing problem: the "problem of small and new accounts," where good content simply gets buried because the account has no established reach. Grok is meant to anchor relevance more strongly in the content itself than in raw account history or follower count.

For the Mittelstand, this is fundamentally an opportunity: a small but content-strong business account theoretically has the same starting position as a large account - if the content itself triggers the right signals. At the same time, this shifts responsibility: it's no longer enough to "just post." Every single post has to be crafted so that it generates the signals the model weights heavily.

Signal weighting: what really counts now

Analyses of the disclosed code, published by outlets including Social Media Today and other trade media, allow for a rough ranking of signal weightings. These figures should be understood as approximations reported by third parties, not as documentation officially confirmed by X - but the underlying pattern is revealing:

This ranking overturns a lot of long-held social media wisdom. A like, for years the standard metric of social media success, is nearly meaningless in the new system. What counts is genuine, two-way conversation - one that the business account itself actively keeps going. A comment you simply leave sitting there gets you almost nothing. A comment you respond to yourself is, according to these analyses, the strongest lever in the entire system.

At the same time, according to reports, X has signaled that sheer reply volume no longer counts for reach and, in particular, for participation in the revenue-sharing program - the gist reportedly communicated was that "replies don't count anymore" in the sense of sheer quantity. This quote is attributed to a representative of X's product team but could not be verified against a clear primary quote and date, so it should be treated with caution. The underlying point, however, fits the overall picture: it's about the quality of individual dialogues, not the number of replies produced. Anyone hoping for quick numbers through mass replies or engagement bait is likely to be disappointed in the new system.

The negative side of the weighting is just as important. According to reports, offensive language can reduce a single post's reach by as much as 80 percent. All-caps text, external links in the main body, reported content, and negative user reactions also have a dampening effect. Especially relevant: these negative signals apparently feed into a cumulative reputation score for the entire account, which also affects future posts. A single misstep can therefore have longer-lasting consequences than it used to - one more reason to consistently keep an eye on tone and compliance in your company's presence.

Verification as a cost factor for reach

One aspect especially relevant for business decision-makers concerns the difference between verified and unverified accounts. According to the analyses examined, verified Premium or Organization accounts get a significantly higher reach cap of up to +100, while unverified accounts are capped at around +55. From this, it's inferred that unverified business accounts have to generate four to eight times as much organic engagement as a verified account to achieve the same reach.

According to reports, the price for X Premium in the organization segment ("Verified Organizations") is around $200 a month, while the basic Premium tier for individuals costs about $3 a month. These prices come from secondary sources and may vary by region or pricing changes - it's worth checking X's current pricing page before making a decision. But the fundamental point remains: verification is no longer an optional extra in the new Grok system - it's effectively a cost factor for reach. Anyone who wants to work seriously with X as a marketing channel should budget verification as part of their infrastructure costs, similar to an advertising budget or a content tool.

For many Mittelstand companies, this becomes a genuine calculation question: $200 a month is manageable if X is a central channel for reach, customer communication, or recruiting. If the channel is only run on the side, the benefit shrinks accordingly. The decision shouldn't be made "on principle," then, but based on the actual strategic importance of X for your own marketing.

What this concretely means for your business marketing on X

From the facts at hand, a practical checklist for the coming weeks can be derived. It doesn't replace an individual strategy, but it provides a solid starting point:

These points reveal a larger pattern: the Grok algorithm demands less "broadcasting" and more genuine interaction. For many business accounts that have used X like a digital notice board so far, this means a strategy change. Content needs to be designed to raise questions, provoke reactions, and factor in your own active replies to comments - not as an afterthought, but as a fixed part of the editorial plan.

If you'd rather not manually track feed signals, posting times, and reply rates yourself: our AI agents take care of research, content creation, and timely replies to comments - tuned to exactly these kinds of algorithmic changes. Schedule a no-obligation conversation.

Beyond X: Grokipedia and AI visibility

The changes at X are part of a larger trend: more and more visibility for companies is being generated not by human editors or classic search-engine rankings, but by AI systems that curate, summarize, and process content themselves. One example is Grokipedia, the AI encyclopedia run by Musk and xAI. Since December 2025, over 75 percent of all suggested edits there have come directly from Grok itself - largely without human involvement. At the same time, several major AI systems, including GPT-5.2, Google AI Overviews or Gemini, Microsoft Copilot, and Perplexity, are already citing entries from Grokipedia.

For companies, this means: how a brand shows up in AI answer systems is increasingly shaped by platforms like Grokipedia too - not just by your own content on X. A factually accurate, well-maintained entry about your own brand can therefore indirectly influence visibility in AI-powered search results beyond X as well. That's another reason why business marketing should no longer orient itself solely around individual platform algorithms, but around the question of how AI systems as a whole "talk about" your brand.

This development fits into a pattern currently running through almost every digital channel: AI models are increasingly taking over content curation - the feed at X, the encyclopedia entries at Grokipedia, the AI-generated summaries in search engines. Companies that understand which signals these models decide by gain a structural advantage over those still communicating by the old rules.

Want to know how your brand currently shows up in AI systems like Grok, ChatGPT, or Gemini - and how that can be deliberately improved? Let's analyze this together.

The shift to a fully Grok-driven algorithm marks one of the most far-reaching changes to how X works since Musk's takeover. For business marketing, this means: reach is no longer a static state you earn once and keep, but an ongoing task that depends on genuine conversation, high content quality, a clean reputation score, and - depending on the channel's strategic importance - a verification decision as well.

Anyone who uses the coming weeks to adapt posting formats, response times, and their own tone to the new weightings gains an advantage over competitors still communicating by the old rules. At the same time: the exact weightings come from secondary analyses of the disclosed code and may change as the system is further adjusted. Regularly checking your own metrics therefore remains essential.

Want to consistently adapt your social media presence on X and beyond to the new AI algorithms, without manually tracking every change yourself? Book a free initial consultation with Amazing Agents.