Open Under the Hood on X and record the report period, label name, stated effect and either the number of affected posts or days. Those fields tell you what X reported and when it applied.
For example, SPAM_HIGH_RECALL describes exclusion from recommendations to non-followers. SPAM describes a post that is not shown on X. The similar names lead to different next checks.
The examples below come from X’s published report implementation and label map. The links point to fixed code versions so the conditions and label wording remain inspectable.
Contents
Keep the period beside every label
Read the period printed in the report rather than assuming it covers last month. The published implementation selects a prior UTC month after a calendar waiting period and can fall back another month when observation data is incomplete.
- Post label
- Record the label, its effect, the affected-post count and the total eligible posts used for the percentage. This is a count of posts carrying that label.
- Account label
- Record the label, its effect, the days it was active and the days included in the period. A percentage of days describes time, not a percentage of followers or lost impressions.
- Generated time and report period
- Keep both when available. A report generated today can still describe an older month.
A worked hypothetical report: 30 eligible posts and 20 covered days
This constructed example follows the published report functions. It is not a capture from an X account. Suppose the report period is September 1–30, 2026 UTC, generated on October 15. Its eligible-post aggregates contain 12 posts on one day and 18 on another: 30 in total.
The SPAM_HIGH_RECALL aggregates carry counts of three and two. A separate SPAM row carries two. For an account-level NsfwHighPrecision label, three dates appear in activeDays. Observation coverage runs from September 11 through September 30, inclusive.
The report builder sums the eligible-post counts, sums each post label’s carried counts, and counts an account label’s activeDays. Here is what those inputs produce.
- SPAM_HIGH_RECALL: 5 of 30 posts → 16.66%
- Three plus two gives five affected posts. The denominator is the 30 eligible posts, not the number of followers or impressions.
- SPAM: 2 of 30 posts → 6.66%
- This row is calculated independently. The report’s per-label aggregates contain no post-ID list for deduplicating overlap, so adding the two percentages does not establish a unique affected-post total.
- NsfwHighPrecision: 3 of 20 days → 15.00%
- September 11–30 contains 20 covered days. Three active dates divided by 20 produces 15.00%; dividing by the full calendar month would use the wrong denominator for these inputs.
The percentage formatter truncates to two decimal places: 5 ÷ 30 becomes 16.66%, not 16.67%. The day-count function uses a valid firstCoveredDay/completeThroughDay range when supplied, otherwise the month’s calendar-day count.
Read the effect beside each result. Five SPAM_HIGH_RECALL posts describes a recommendation restriction on those labelled posts. Three account-label days describes time with an account-level condition. Neither number is a measured loss of views.
Similar labels can change different surfaces
X supplies an explanation and an effect for each supported label in the published map. These examples show different surfaces a label can affect.
- SPAM_HIGH_RECALL
- The post is excluded from recommendations to people who do not follow its author. Check that recommendation boundary rather than expecting the direct URL to disappear.
- FOSNR_ABUSE
- Discoverability is restricted to the author’s profile, with a notice explaining limited visibility. A post found on that profile can therefore still carry this restriction. This is a different scope from excluding only recommendations to non-followers.
- SPAM
- The post is not shown on X. Start by checking the post itself and any notice X has supplied.
- PDNA
- The post is not shown while it awaits further review. Preserve the notice and review state attached to the post.
- NSFW_HIGH_PRECISION
- The effect includes a content warning, limits on recommendations to non-followers, and restrictions for underage, unknown-age and logged-out viewers. The map also describes a country-specific restriction in India. Viewer state and location belong in the observation.
Copy the effect shown in your own report when making a note. Account labels, post labels and country-specific takedown rows can describe different scopes.
Follow one label from its meaning to the monthly row
For SPAM_HIGH_RECALL, the label map supplies the spam explanation and exclusion from non-follower recommendations. The postLabelRow function combines that mapped effect with the aggregated count and eligible-post denominator.
The monthly row tells you the label’s reported scope and frequency. To investigate a post today, keep its original URL and check the surface named by that effect. A direct link can open while non-follower recommendations remain restricted.
Ranking is a separate question: a post can be eligible for a viewer’s feed yet lose to other candidates. The monthly label counts are not ranking scores. Compare a reported restriction with the particular symptom you observed rather than treating low views as a matching label.
An unavailable report has several possible causes
The linked serving code uses a 365-day account-age check and at least 10 eligible posts in the selected month. It also requires complete month-observation data. Your account’s displayed checks and messages determine access.
The month-selection function starts from the prior UTC month. For September, its calendar threshold is September 30 at 00:00 UTC plus ten days: October 10 at 00:00 UTC. The examples below assume the fallback month has complete user data and the account passes the age and post checks.
- October 5, no month cap → August
- September is the prior month, but its calendar threshold has not been reached. The primary candidate becomes August.
- October 15, complete September data → September
- The calendar threshold has passed. With a complete observation marker, September remains the selected report month.
- October 15, incomplete September data → August
- The primary candidate is September, but the data-readiness check fails. The serving path falls back once, to August. Incomplete fallback user data still leaves the report unavailable.
- October 15, configured maximum 202608 → August
- A positive maximum-month setting can cap the primary candidate at August before the data-readiness check. A newer generation date therefore does not guarantee a September report.
These dates and the worked percentages were calculated from the fixed code paths, using hypothetical inputs. The fallback branch moves back one month; it does not search indefinitely for any available older report.
If only eligibility information appears, preserve that message and its period. A public checker cannot supply the account’s missing internal report.
Read an empty label list within its period
An empty report says that no included labels were returned for the displayed period. X describes the report as best-effort. Ordinary ranking and a current search or reply problem still need their own observations.
If your only concern was whether the report contains a label, the check is finished. If a particular post is missing today, keep that post URL and test the surface where it is missing.
Use today’s post for today’s visibility question
Choose the exact post that prompted the question. Compare its direct URL with Latest search. For a reply, also open the parent conversation from another account or a logged-out window and look for that same reply.
A useful note can be short: “Direct URL opened; this post was absent from the same Latest query twice.” Add the query, time and viewer state. Keep any report label in a separate dated line so the two observations do not become one claim.
- Collect the public sample
- Open the visibility checkerCarryFeed compares public profile access, recent posts and a Latest-search sample. Its result includes the sources to inspect manually.
- Find the exact post first
- Search by account and clueUse the account, phrase and date you remember, then keep the final status URL.
Keep the code link with your interpretation
The report-serving source defines month selection, eligibility and post/day counts. The label map defines the reported meanings and effects.
For a notice attached to your account or post, follow the action X displays. X’s enforcement guide explains the different restriction types and available appeal routes.