X's ranking Algorithm is open source on [Github](https://github.com/twitter/the-algorithm). Based on that and millions of user experiments, it's possible to know a lot about what works and what doesn't when posting on X - or at least it's possible to avoid some easy-to-avoid mistakes (like point 1). 1. X penalizes posts with external links - they want to keep people on the platform 1. Multiple credible sources confirm X’s algorithm deprioritizes tweets that include links to external websites. Elon Musk referred to this practice as discouraging “lazy linking” — posting bare links without value — and recommended hiding the link in a reply instead. 2. Linking to other X accounts does not carry the same penalty, which is why many founders link to their startup's X profile instead of their website. 2. Engagement is king on X. It’s no longer follower count that matters, but reactions—likes, replies, reshares—especially replies. 3. Rich media (images, videos, polls, threads) consistently outperform link-only posts, especially for exposure on the “For You” feed According to [X's own website](https://blog.twitter.com/engineering/en_us/topics/open-source/2023/twitter-recommendation-algorithm), here’s how its [algorithm works](https://buffer.com/resources/how-to-grow-on-twitter/) in a nutshell: 1. **Candidate sourcing**: Determine 1,500 tweets to display based on the people you follow (in-network sources) and people whose tweets you might be interested in (out-of-network sources). 2. **Ranking tweets**: Rank the 1,500 tweets using X’s machine learning model. X tries to determine the tweets you’ll engage with and puts them at the top. 3. **Filtering tweets**: Filter the 1,500 tweets to remove people you’ve blocked or muted (called visibility filtering), increase author diversity (remove too many tweets from the same author), and do a content balance of in-network and out-of-network tweets. 4. **Mixing tweets**: Mix sponsored content with organic tweets. Twitter uses [RealGraph](https://www.ueo-workshop.com/wp-content/uploads/2014/04/sig-alternate.pdf) to predict how likely you are to engage with someone else’s tweet. **The higher the RealGraph score, the more likely the creator’s tweet to be included in your feed**. How does Twitter calculate the RealGraph score? It’s a logistic regression model dependent on various factors like how frequently you’ve interacted in the past, how you’ve engaged historically (like retweets, [DMs](https://buffer.com/social-media-terms/dm)), etc. ![[CleanShot 2025-08-07 at [email protected]]] --- **Created at:** [[2025-08-06]], [[2025-08]] ### References: - [Twitter's Recommendation Algorithm](https://blog.x.com/engineering/en_us/topics/open-source/2023/twitter-recommendation-algorithm) - [GitHub - twitter/the-algorithm: Source code for Twitter's Recommendation Algorithm](https://github.com/twitter/the-algorithm) - [How the X Algorithm Works in 2025 (And How to Beat It) – AVRAMIFY](https://www.avramify.com/blogs/news/x-algorithm-2025-guide) - [Is the X Algorithm Hurting Your Business's Link Visibility?](https://www.dkodetech.com/does-the-twitter-x-algorithm-downrank-your-links-what-small-businesses-should-know/) - [The Essential Guide to the X Algorithm: Understanding Its Impact \| Brandwatch](https://www.brandwatch.com/blog/x-algorithm/) - [How the Twitter (Now X) Algorithm Works in 2025 \| Buffer](https://buffer.com/resources/twitter-timeline-algorithm/) - [Steven Tey on X](https://x.com/steventey/status/1641892386564640768) ### Related: