X Publicizes Full Recommendation Algorithm, How to Write Posts for High Exposure

X publicly revealed its entire recommendation algorithm yesterday, which has some nuances worth exploring beyond the well-known high-interaction bonus.

Instead of assigning a fixed total score to a post and then pushing it to everyone, the algorithm considers each individual reader and evaluates "what is the most likely action this person will take after seeing this post": like, comment, quote, repost, share via direct message, copy link, follow the author, or indicate "not interested," mute, block, or report. The model predicts the probability of these actions, which are then combined into a ranking score with different weights.

How Exactly Does This Algorithm Score Posts?

This can be simplified into a formula:

Post's Base Score for a specific reader = ∑ (Predicted Probability of a reader taking a certain action × Weight of that action)

Furthermore, the system addresses issues such as excessive content from the same author, whether a post is from an account the reader follows, and whether the content is suitable for display. In other words, a post may be very appealing to Reader A but not necessarily receive the same high score from Reader B.

In the default public parameters:
The weight for copying and sharing a link is 20,
much higher than the weights for commenting and quoting at 5,
following the author at 4,
regular sharing at 2,
reposting at 1,
liking at 0.5,
clicking on the post at 0.4,
opening an external link at 0.2,
and expanding an image or video at 0.05.

This set of numbers can be easily misinterpreted. Adding an image does not automatically add 0.05 points; the model first has to believe that the reader is likely to expand the image before including "image expansion probability × 0.05" in the total score. Similarly, copying and sharing a link is not equivalent to twenty likes in one go; it is a higher-weighted behavior that typically occurs less frequently. The algorithm always considers the product of probability and weight.

Therefore, simply aiming for likes is not the optimal strategy. Content that a reader would find worthy of sharing with colleagues, friends, or peers, or content that gives the reader ample reason to provide additional comments, rebuttals, and references, aligns better with this ranking objective. The determination of what is "worthy" is a content strategy judgment, but it directly stems from the weights assigned to sharing, commenting, and quoting by the system.

In addition to the base score, the algorithm also reveals several overlooked realities. For the same candidate ranking, the first post by the same author usually competes normally, while subsequent posts from the same author experience author diversity attenuation: multiplying by 0.5 for each additional post, down to a minimum of 0.25. Content from accounts the reader does not follow has a default external recommendation coefficient of 0.75; replies and reposts from unfollowed accounts are even more disadvantaged, with some being filtered out even before the ranking.

This explains a very common phenomenon: posting five similar threads in a short period of time will not simply multiply exposure by five. What is really worth betting on is a complete, independent, and instantly understandable original main post. This conclusion is an inference of the content layer of the public rules, rather than an additional system parameter.

Why “Usefulness” Is More Important Than “Buzz”

If the goal is to get readers to copy a link or share via private message, then the most effective content is usually not a general attitude, but something tangible: a judgment, a case study, a framework, a checklist, or a clear decision recommendation.

This is not because the algorithm gives extra points for the word “checklist.” It's simply because a good checklist is easier to pass on to a specific person than a vague sentiment; a evidence-backed counterintuitive conclusion is also easier to quote and discuss than vague inspirational content. In other words, content format is not the score, reader behavior is the score; content format only increases the likelihood of those positive behaviors occurring.

For example, the following two sentences express almost the same idea:

AI will change the content industry.

I reviewed 30 AI content teams, and model capability is not important; breaking down topic selection, source material, and review into three independent stages is what matters.

The second sentence does not receive any “data reward” from the algorithm. But it provides context, examples, and actionable judgments. Those who have actually managed content teams can add cases, present counterexamples, or pass it on to the person in charge of the process; this is the condition where replies, quotes, and shares are more likely to occur.

A high-exposure thread is best positioned in the first sentence: it tells who it's for, presents a specific judgment, and explains why readers should keep reading. The body does not need to be long, but should deliver enough information for readers to gain value without leaving X. Links can serve as evidence or supplementary material, but the default public parameters do not show “linking out will directly reduce authority”; a more prudent approach is to first deliver the core answer in the body, and then let the link provide the source and details.

The conclusion should also serve real discussion, not solicit engagement. “What does everyone think?” easily leads to generic replies; “For those who have developed B2B AI products, what is the most effective second-step activation action you have seen?” gives the right people a clear entry to respond. It is a public fact that replies and quotes carry equal weight of 5; that narrow questions are more likely to bring high-quality interaction than broad questions is a writing judgment.

Images, Videos, Links: To Use or Not to Use

Images, videos, and links are all means of expression, not algorithmic talismans.

Both image expansion and video opening have a default weight of only 0.05. If a chart can make a growth trend, experimental results, or before-and-after comparison easily understandable at a glance, it is worth using; if it is just adding an unrelated visual to accompany text, it will not naturally receive a higher score. The same logic applies to videos: demonstrations, processes, comparisons, and on-site experiences can help content be understood, but there is no evidence to suggest that "just posting a video will increase exposure."

The weight for external links opening is 0.2, and there is also no "external link penalty" for open-source code. Therefore, there is no need to hide links to evade any supposed link demotion. More importantly, do not leave all the value of a post behind a link. A reader should be able to get conclusions, be willing to share or discuss within the main text for the link to serve as a trust-building supplement rather than a reading threshold.

Behaviors That Will Lower Your Score

The default weight for public feedback is quite severe: not interested is -43.2, blocking the author is -31.2, muting the author is -58.8, and reporting is -234. These values are also probabilities, not fixed deductions per occurrence; however, they indicate that the algorithm would rather recommend fewer posts than push a significant amount of content that readers clearly dislike.

This leads to several practical writing consequences. Promising an amazing conclusion in the title but not providing evidence or answers in the main text can easily make readers feel misled; attacking individuals to evoke emotion, riding on unrelated hot topics, may spark brief discussions but is more likely to trigger disinterest, muting, and reporting. Breaking down a piece of content into a series of posts without independent value will also be affected by author diversity decay.

These effects of "clickbait reduction" and "controversy for the sake of controversy harming long-term exposure" are not verbatim rules in the source code but reasoned inferences based on negative feedback weight and diversity mechanisms. Their value lies in helping you decide what is not worth the risk, rather than turning writing into a mechanical minefield.

Some Situations Result in Lost Exposure to Strangers Instead of Point Deduction

Beyond ranking, there is visibility filtering. Public rules show that posts with spam, malicious links, pornography, explicit content, graphic violence, hate speech, or severe insults as safety labels may be directly removed when recommended to non-followers; accounts marked as impersonation, compromised, restricted, or high risk may also be removed. Content may also be unable to enter a user's recommended feed if the reader has blocked or muted the author.

This is not the same as "losing a few likes." The former removes the post's candidacy directly. The repository also clearly states that some anti-spam rules are not public, so do not treat the open-source code as a manual for evading detection. Practices like inflating engagement, purchasing followers, engagement pods, or automated interaction manipulation should not be attempted; the open-source code neither does nor should be used to deduce what actions might trigger certain labels.

There is another type of action that is not a security penalty but will also limit the spread: the candidate process will remove posts older than 48 hours. Just because a piece of content didn't find enough initial matching doesn't mean it's worthless; but it's not suitable to expect it to re-enter widespread For You recommendations a few days later. If necessary, write about new developments, data, or insights as a new original post instead of reposting the old one verbatim.

So, How to Write a Good Post

The best approach is not to fit into a "viral formula," but to accomplish four things in a standalone main post: start with a specific conclusion, provide enough reasoning, deliver a actionable insight for the reader to take away, and end with a question worth answering by specific individuals.

For example:

I analyzed 30 AI product launch posts and found that what brings the most engagement is not "powerful features" but "which step it eliminates for whom."

Saying "We support Agent workflow" is hard to reshare.

Saying "Our sales team used it to cut down customer research from 2 hours to 12 minutes" is more likely to be shared with colleagues.

Good product copy does not describe capabilities but describes a role's pain points.

What product have you seen that excels in translating abstract capabilities into tangible benefits?

This example does not guarantee exposure because the algorithm still depends on the specific reader, account history, candidate competition, and experimental configuration. However, it possesses content features aligned with the public goal: it is an independently original post; the conclusion is given in the first sentence; a shareable framework is provided in the middle; and the invitation at the end is for specific experiences rather than generic endorsements.

A truly sustainable strategy is not to chase a one-time viral hit but to repeatedly provide a judgment worth resharing to the same type of people. The public algorithm predicts behavior based on reader personalization; consistent themes, genuine value, and healthy interactions will give the system more opportunities to place the next piece of content in front of those who are truly willing to respond, quote, and share.