Shadowbans: three different things wearing one word

6 August 2026

Almost every argument about shadowbans is really two people using one word for different things. The word is doing three separate jobs. Two of the three are real and documented. The third is not a mechanism at all. Once you separate them, most of the advice written on this subject stops making sense, including the parts that sound most confident.

Job one: enforcement that is not announced

This one exists, and a dating app has said so on the record. In a piece by Thomas Germain for Gizmodo on 21 August 2023, about how Hinge's algorithm works, a Hinge spokesperson said the company "sometimes handles bad actors by preventing other users from searching for them, seeing them in the feed of potential matches, or receiving their messages".

That is a shadowban by any reasonable definition. Note the two things that come with it. The same spokesperson also said "We do not withhold possible matches from users", and the mechanism as described is scoped to bad actors. So the company acknowledges the tool exists, and says it is pointed at people who broke the rules, not at people who are insufficiently attractive or insufficiently active.

This matters because the folklore version has the arrow pointing the wrong way. In the folklore, a shadowban is what happens when the algorithm decides you are not good enough. In the documented version, it is a moderation tool applied to conduct.

Job two: being deprioritised, which is official and is not a punishment

Tinder publishes a description of its matching in Powering Tinder: The Method Behind Our Matching (updated 11 July 2022). It says it factors in activity level and whether people are active at the same time, along with location and distance, age and gender preferences, profile interests, anonymised cues from photos, and Likes and Nopes. It also states that the algorithm "doesn't track social status, religion or ethnicity".

Read that honestly and a lot of "shadowban" experiences are described in it. If you stopped using the app for a fortnight, you are less likely to be surfaced, because showing an inactive account to an active one produces a match that goes nowhere. That is a design choice with an obvious rationale, it applies to everybody, and it reverses when you come back. It is not a sanction, there is nothing to appeal, and nobody decided anything about you.

The same goes for your own settings. A narrow distance radius and a tight age range will empty a queue quickly, and from the inside that is indistinguishable from being hidden.

Job three: an ordinary bad run

The third job the word does is explaining variance. Match counts are small numbers, and small numbers are noisy. Two good weeks followed by a dead one is what randomness looks like at that scale.

The pool is also getting smaller. Match Group's Q2 2026 prepared remarks, filed with the SEC on 4 August 2026, state that Tinder monthly active users declined 7% year over year in the quarter. The same filing says recommendation algorithm updates rolled out through 2026, including a version in mid-July being extended to further groups of users. A change in what you see, with no change on your side, has a documented explanation that has nothing to do with enforcement. That is the subject of Tinder matches suddenly stopped.

What Tinder actually documents

Tinder publishes its enforcement ladder in its remediation policy. The listed responses include removing the violating content, hiding violating content from others' view, requiring verification, suspension and banning.

Two things follow. First, content-level hiding is documented, so the concept is not imaginary. Second, and more usefully: account-wide silent throttling of a member who has broken no rules appears nowhere in that document. If it existed as a stated policy, this is where it would be. Absence from a policy page is not proof of absence in the code, and it should not be presented as such. But when a page tells you Tinder shadowbans people for swiping too fast or for having an incomplete profile, ask where that is written down.

Hinge

Hinge is the only one of the three with an on-the-record acknowledgement, quoted above. Its scope is the interesting part: the spokesperson framed it as handling bad actors, and separately denied withholding possible matches from users. Both statements come from the same company on the same day, and they are only contradictory if you assume moderation and recommendation are the same system. They are not.

If your Hinge results have gone quiet and you have not broken any rules, the format itself is the more likely culprit, and it is genuinely different from Tinder's. We covered that in No matches on Hinge.

Bumble

Bumble's published support material on content removal and account blocks states that a user who is warned or blocked is told, by email and by a notification when they next open the app, and that the decision can be appealed within six months.

The consequence is narrow and worth stating precisely. On Bumble, by their own documentation, a block comes with a notification. So a silent drop in matches, with no email and no in-app notice, is not the profile of an account block. That is a claim about blocks, and it does not extend to a claim about recommendation ordering, which is a different system that Bumble does not publish. Anybody telling you Bumble has officially confirmed that shadowbanning does not exist is repeating a line with no source behind it.

For the Bumble-specific version of the ordinary problem, see No matches on Bumble.

Why no checker tool can work

This is structural, not a matter of some checkers being better than others. No third party has read access to any of these platforms' ranking or moderation state, and none of them exposes account standing through a public or partner API. So a checker has exactly two mechanisms available, and both fail.

A questionnaire. It asks how many matches you used to get and how many you get now, then returns a verdict. The verdict is a restatement of your own answers. Nothing was measured.
A second-account probe. Make another account, see whether your profile appears. Your own profile is normally excluded from your own recommendations to begin with, so the test is confounded before it starts. On top of that, distance, age filters, gender preferences, recent activity and the ordering of the queue each independently decide whether you surface in any given session. Not appearing proves nothing at all.

The same problem sinks the popular "buy a Boost, and if it returns no likes you are shadowbanned" test. There is no control condition and no base rate. You do not know what a normal Boost returns for someone like you in your area on a Tuesday, so a disappointing one carries no information. It does reliably produce a purchase.

We have not built one, deliberately

A shadowban checker would be straightforward to build and would probably do well in search. We have not built one and will not, because there is no honest version of it. Any tool we shipped would either hand your answers back to you dressed as a diagnosis, or run a test whose negative result means nothing. Charging for that, or using it to collect an email address, would be selling a measurement that does not exist.

What you can establish from outside is short. Whether you were told you were banned or suspended, because bans and suspensions come with notification. What your own filters are set to. Whether you have had a gap in usage. Whether the drop is outside your normal range over weeks rather than days. Everything past that is unobservable from where you are standing, and a confident answer about it is a guess wearing a costume.

If the honest position is that things have been thin for a while rather than suddenly cut off, the ranked list of causes is in Why am I not getting matches.

Written by Fottory, which generates dating photos from photos of your own face. The first two are free, and none of the advice above depends on using it.