There is no published duration
X does not publish how long a filtered-visibility state lasts, and it does not notify accounts when one starts or ends. Anyone quoting an exact number of days is repeating community observation, not documentation. Treat every figure below as reported experience, not a rule.
What is documented is the appeal path for enforcement actions — suspensions, lockouts, labelled posts. Quiet visibility filtering is not one of those, which is why most restricted accounts never receive a message at all.
Typical ranges people report
Across public creator discussions, the most commonly reported window for the mild, spam-filter style of filtering is a handful of days to about two weeks, with the midpoint around a week. Cases that persist longer are usually tied to repeated triggering behaviour rather than to the original incident.
The uncomfortable part: with no notification, people often cannot tell when it actually lifted. The timeline worth tracking is the one you measure — the same test, the same day, each week.
What shortens it, and what extends it
Shorter: deleting posts that plausibly triggered a filter, stopping aggressive behaviour such as bulk following or identical replies, and letting the account sit still for a few days. A quiet week resolves more cases than a week of frantic posting.
Longer: automation, follow churn, repeated link dumping in replies, and posting the same text across many threads. If the pattern continues, the filter is simply re-applied — duration is a symptom, not a decision.
How to spend the waiting time
Keep posting as you normally would. Going silent for two weeks costs you more reach than the restriction itself, and there is no evidence that silence lifts anything. If the account carries your business, publish what does not depend on discovery: conversations with your own community and your off-platform channels.
Measure once a week, write the result down, and only change strategy when two consecutive measurements agree. Daily testing produces anxiety, not signal.