X announced on 13 August 2026 that it has expanded the algorithmic source code published in its GitHub repository, adding detail on how the For You timeline ranks posts, including the content labels that can suppress a post's visibility. Alongside the code, X is testing a tool that lets a randomly selected group of accounts download the labels applied to their own posts over the previous month. Social Media Today reported the announcement on 14 August 2026.
The label export is the part that changes anything. Publishing ranking code tells you what the system can do to a post. Exporting your own labels tells you what the system did to your posts, which is the question every account holder who has watched their reach fall off a cliff has actually been asking.
What X published on 13 August 2026
X described the release as open-sourcing the code that affects a post's visibility in the For You timeline. The expansion covers the ranking path specifically, including the labels and tags a post can pick up that reduce how widely it is shown.
X stated its intent in its own words: "we want people to be able to answer for themselves whether a platform is limiting their reach, whether the system is fair, and why they see particular content." Keith Coleman, X's VP of Product, framed the release as users being able to independently analyse their own account performance and any restrictions on it.
The suggested workflow X put forward is unusual for a platform disclosure. Rather than reading the code, X proposes that a user feed the published source code plus their downloaded label data into a large language model and let the model work out what is affecting their reach. That is a real answer to the readability problem, since ranking code is not something a brand manager is going to parse by hand, and it also quietly moves the interpretation burden onto the user.
Why shadowbanning has been an unwinnable argument until now
This section is analysis rather than reporting. The reason the shadowbanning debate has run for a decade without resolution is that it has been structurally unfalsifiable. An account sees its reach drop. Two explanations fit the same evidence exactly: the platform suppressed the account, or the posts got worse and the audience responded accordingly. No data available to the account holder separated those two, so the argument was always about belief.
A downloadable list of labels applied to your own posts breaks that symmetry for the first time. If the export shows labels on the posts that underperformed, the suppression explanation gains evidence. If it shows a clean month against a reach collapse, the content explanation gains evidence. Neither outcome is comfortable, and both are more useful than the argument that came before.
The limit is worth stating plainly. A label export shows that a label was applied. It does not necessarily quantify how much reach that label cost, and the source did not say whether the export includes any magnitude figure alongside each label.
Who can get the tool, and who cannot
The label-checking tool is in testing, not general release. The table below sets out what X confirmed about access, based on the Social Media Today report of 14 August 2026.
| Question | What X confirmed |
|---|---|
| Who can access the label download | A randomised group of accounts at least one year old |
| How much history the export covers | Labels or tags applied over the past month |
| When it reaches everyone | No date given |
Two exclusions follow from that first row. Accounts younger than one year are out, which covers most brand accounts opened for a recent launch or a new market entry. And even among eligible accounts, selection is random, so an established brand handle has no route to request access. There is no waiting list mentioned and no eligibility check to run. The practical answer for most marketing teams is to check whether the option has appeared in their account and move on if it has not.
The credibility question sitting underneath this
Social Media Today notes the relevant history: X promised regular algorithm updates back in 2023 and did not deliver them, which is a reasonable basis for treating the current release as a first step rather than a settled commitment. The 2026 expansion is more detailed than what came before it, and there is still no published cadence for keeping it current.
That matters more than it sounds. Ranking code that is published once and then left to drift becomes a historical document while the live system moves on, and a user analysing month-old labels against year-old code would be reasoning from two things that no longer match. The source did not say how often X intends to refresh the repository.
What a marketer should do with this
- Check whether the label download has appeared for your account. If your handle is over a year old you are inside the eligible pool, though selection is still random.
- If you do get the export, line the labelled posts up against your own reach data for the same month before drawing any conclusion. A label on a post that performed fine is a useful data point too.
- Treat X's suggested workflow of feeding code plus labels into a language model as a starting point for a hypothesis, not as a verdict. The model is reading code it was not trained on and has no access to how the system actually behaved for you.
- Do not rebuild a content strategy on a single month of label data. One month across a normal posting cadence is a small sample.
What the source did not say
The report gives no rollout date for general availability. It does not state how many accounts are in the test group, which labels exist in full, whether the export shows a reach impact figure per label, or whether advertisers and verified organisations get separate treatment. It says nothing about how X will handle disputes when an account holder disagrees with a label. And it makes no reference to Thailand or to any regional availability, so nothing here can be read as confirming or denying local access.
What this means for Thai marketers
X is a smaller channel in most Thai media plans than LINE, Facebook or TikTok, and that shapes how much this is worth. For most Thai brands the direct value is low, because the accounts likely to be selected are older personal and publisher handles rather than the brand accounts running campaigns.
The indirect value is higher and worth watching. If label transparency at X produces usable evidence and no obvious harm to the platform, it becomes a reference point that Thai marketers can raise with every other platform they buy on. The argument stops being a complaint about opaque reach and starts being a request for something a competitor already ships. Teams running paid social across multiple platforms can start asking that question now.
There is also a content angle. If suppression labels become visible on any platform, the value of publishing where you control the rules goes up, because an owned channel has no label layer between the post and the reader. That is the same reasoning behind treating content published on your own site as the anchor and social as the distribution.
FAQ
Can I get the label download for my brand account?
Probably not yet. X is testing it with a randomly selected group of accounts at least one year old, so a brand account under a year old is excluded outright, and an older one still depends on random selection with no way to request access.
Is this available in Thailand?
The source did not state this. The Social Media Today report of 14 August 2026 described eligibility only by account age and random selection, and made no mention of any country or regional restriction either way.
Does this prove shadowbanning is real?
It does not prove it in general, but it gives individual accounts evidence they never had. X has published the code covering labels that can reduce a post's visibility and is testing an export of the labels applied to a given account, which turns a belief into something checkable case by case.
When will everyone get the label tool?
X gave no date. The announcement of 13 August 2026 described the feature as in testing with a randomised group and offered no timeline for a broader rollout.
Do I have to do anything today?
No. Nothing about this changes how you should post or advertise on X right now, and the reasonable action is to check whether the export has appeared for your account and to keep your own reach records so you have something to compare it against if it does.
If you want a clearer read on why reach is moving on any of your social channels rather than a theory about it, Relevant Audience can go through your posting and performance data with you and separate what the platform is doing from what the content is doing.
