X just open-sourced more of its algorithm. Here are 5 things worth knowing if you manage X accounts
X just open-sourced more of its For You recommendation system, so we spent some time digging through the code.
A lot of people are focusing on how likes, replies, and shares are weighted. But if you manage X accounts for a business, creator, or client, a few other details stood out to us:
1. Negative feedback can be expensive
Some of the current ranking coefficients include:
- Like: **+0.5**
- Reply: **+5**
- Copy link: **+20**
- Block: **-31.2**
- Mute: **-58.8**
- Report: **-234**
These are not points added or subtracted every time someone takes an action. They are coefficients applied to the model’s predicted probability of each action.
Still, the takeaway is clear: chasing engagement at any cost may backfire. A post that attracts replies but also makes people mute, block, or report the account may not be a good trade.
2. Standard out-of-network posts get a default 0.75× factor
For standard posts competing in the feeds of people who do not follow the author, X currently applies a default **0.75 out-of-network weight factor.
In other words, an otherwise similar post from an account the viewer does not follow can be scored at a disadvantage compared with in-network content.
Getting discovered by a new person is only part of the job. Turning that discovery into a follow can matter for future distribution too.
3. The For You pipeline has a 48-hour age filter
Posts older than 48 hours can be filtered before scoring in this pipeline.
That does not mean a post disappears from X after two days, but it does suggest that the first 48 hours are especially important for discovery through this part of the For You system.
4. The code includes a cold-start path for smaller authors
The released code contains thresholds such as:
- Author: **≤1,000 followers**
- Post view count: **<1,000**
- Replies excluded
- Reposts excluded
- A **24-hour** freshness limit in one experimental path
This does not guarantee every small account a boost. Eligibility also depends on additional ranking and experiment conditions.
But it does show that X has specific logic for giving some eligible, low-impression posts from smaller authors a better chance of being surfaced.
5. Multiple posts from the same author do not always get equal treatment
When several posts from the same author appear in the same candidate set, X can apply an author-diversity adjustment.
With the current defaults, the multiplier roughly decays like this:
1.00× → 0.625× → 0.438× → 0.344×
This is not a blanket penalty for posting frequently. It means repeated posts from the same author can receive progressively lower multipliers when they compete within the same feed request.
So publishing more does not automatically give every post an equal chance in someone’s For You feed.
The bigger takeaway for us:
Do not optimize only for individual posts. Think about the signals your content generates and how the account grows over time.
If you manage multiple X accounts, AdsPower profiles can keep each account’s cookies, login session, and workflow separate. That helps with the operational side, but it cannot replace good content or healthy account behavior.
