Were you banned by a person, or by a machine?

If you were banned from a dating app and cannot work out what you did, there is a decent chance the answer is that nobody decided anything about you. A system matched a pattern, or counted something, and the account closed. The companies do not hide this. They write it down in regulatory filings, in privacy policies, and in one case in a published workflow description, and almost nobody reads any of it.

This page collects what the apps say about automated enforcement, in their own words and numbers, and then works out what it changes about your appeal. It changes the tone of what you write, it changes which lever you reach for, and in the EU and the UK it opens a named right that costs nothing to use.

What the platforms publish about automation

Start with the one number that settles the argument. In its EU transparency filing, Tinder reports the count of enforcement measures taken solely by automated means as 1,536,656, against 1,086,123 measures not taken by automated means. That is not a footnote about spam filters. That is the majority of Tinder’s enforcement, executed with no human in the loop, reported to a regulator by the company itself.

Hinge’s 2025 report puts it at 445,488 of 633,534 moderation measures taken solely by automated means, roughly 70 per cent. Plenty of Fish words it rather than counting it, conceding that at the account level “a majority of account actions were proactive and a substantial share were applied automatically”. Bumble’s own EU filing says roughly 636,827 of the approximately 739,365 accounts it limited in visibility in calendar 2025 were actioned by automated systems, and Badoo footnotes its equivalent table with the line “The highest number of accounts limited in visibility were detected and actioned by our automated systems.”

Grindr says it in ordinary English rather than in a table: it uses a mix of proactive, automated moderation including machine learning and human review, which means “profiles may be automatically banned immediately after violating our rules”. happn goes further. In the statements of reasons it files into the EU transparency database, the explanatory field was identical on all 1,000 rows sampled: “The breach was identified during our investigations, and the decision to take action was automated.”

Two caveats. Bumble’s number counts visibility restrictions rather than terminations, a quieter problem covered in the shadowban guide. And an app missing from this page is missing because it publishes nothing usable, not because it moderates by hand.

Feeld: the one company that published the mechanism

Every company on that list tells you automation happens. Only one tells you how. Feeld, in its own terms, describes a workflow it calls the Feeld Safety Service, and states its purpose in a single sentence: “Automatically blocking accounts when a specified number of community reports are received.” The document goes on to say what the workflow does, which is calculate the number of reports.

That is the clearest statement in the industry of something wrongly banned people say constantly and are never believed about. If a group reports you, or one person reports you repeatedly, or a row that started off the app followed you onto it, the threshold does not care why. It counts. Feeld also concedes in its help centre that bans of this kind get reversed on review, which converts “I did nothing wrong” from a plea into a category the company already recognises.

Feeld also names its four automated enforcement vendors and says what each one does, which no other app here does. The first, SEON, runs at registration and then throughout your use of the service, for blocking registrations from countries where Feeld is not available, fraud detection at signup, and fake account detection. Note the phrase “throughout Member usage”. It is not a gate you pass once.

Two Feeld warnings, both from its own documents. Its Digital Services Act help article promises prior warning before a suspension, while paragraph 15.3 of its UK terms authorises enforcement “with immediate effect, without prior notice, and without refund”. Both were live on the same day. And its help centre tells banned users they can create an account “as many times as you want”, while its terms require a continuing warranty that you have not been barred. Following the help article breaches the contract.

Breeze, and the right to a human review

The strongest thing found anywhere in this audit is not a number. It is a door. Breeze, the Dutch app, discloses that “we use large-language models to flag profiles for review on whether they adhere to our User Guidelines and to approve or reject profiles from our App”. It frames that expressly as automated decision-making under Article 22 of the GDPR, and it publishes a dedicated address for contesting an AI profile rejection and asking for a human to look instead: [email protected].

That is worth more than any accuracy statistic, because it is free, it is named, and it is a right rather than a favour. Breeze publishes a second address alongside it, [email protected], for a formal complaint against an enforcement decision within six months of the notification, which is the Digital Services Act internal complaint route wearing a company badge.

Breeze is also unusually candid about the trade-off it has chosen: “if there is any doubt about a breach of these Guidelines, we will take action while allowing the user to explain. In a dating context, protecting users’ safety and feelings thereof takes priority over fairness.” That is a company writing down, in the operative document, that it would rather ban you wrongly than miss a real threat. Most of its competitors run the same policy without saying so.

Two other apps publish something similar. happn states that its automated analysis “may result in the refusal of access to the Application” and then says, plainly, “You have the right to obtain human intervention.” Bumble discloses a human-review and contest route for its automated Deception Detector transaction blocking, in the privacy policy rather than the terms, and Badoo does the same for automated payment decisions. Note the shape of that: both grant the human-review right over money, not over moderation.

Who else admits it, and what they give you

Disclosure without a remedy is the industry norm. Boo publishes a dedicated automated decision-making page admitting bans can be fully automated, and grants no human-review right at all. Its FAQ adds that temporary bans “can arise as a result of our system automatically detecting content”, and that they restrict what you can do rather than removing the account, which is why some Boo users are convinced nothing happened.

• HER calls the systems invisible, in writing. Its own words: “Automated decision-making and profiling are invisible features that are at the core of the services you know and love”, and the processing can have a significant impact on affected users including suspension and account closure. Its CSAM detection runs on a third-party tool that isolates the account automatically, with no human ahead of the suspension.
• Hily discloses the most and gives you somewhere to go. It sets out its automated moderation in full, publishes an error rate for it, and names a channel to contest the decision. It also admits something rarer: access can be restricted on a prediction that a user “is likely to breach the Terms of Use”, before any breach, and affected users can contact Hily to contest that.
• BLK, The League, Match.com and OkCupid all disclose profiling that ends accounts. BLK’s wording is that the processing “may lead to suspension or account closure”. The League adds that messages and video calls are scanned automatically, with human agents reviewing what the automation flags. Match.com discloses that age bans start with automated face analysis.
• Coffee Meets Bagel denies it. Its position is that its automated processes “do not constitute solely automated decision-making that produces legal or similarly significant effects”, while its safety pages describe automated detection driving account blocks. Muzz sits better: it says accounts are flagged automatically and then reviewed by a person.

The accuracy numbers, and why they mislead

Three companies publish an accuracy figure for their automated moderation. Badoo claims 99.60 per cent accurate, 0.40 per cent error. Bumble claims 99.25 per cent and 0.75 per cent. eharmony, reporting for Australia for April to June 2025, claims 98.53 per cent and 1.47 per cent. Read cold, those numbers tell a banned person they are almost certainly guilty.

Now put the other half of the same filings next to them. Bumble’s EU report for calendar 2025 records 41,982 appeals and 9,065 granted, 21.6 per cent, at a median of 61 minutes. So one document says the automation is wrong 0.75 per cent of the time and that roughly one appeal in five succeeds. Both can be true, because they are not measuring the same thing.

eharmony is the most honest of the three, because it defines its accuracy figure as the reversal rate following user appeals, an express concession that automated bans are sometimes wrong and that appealing is how it finds out. Tinder and Hinge do something stranger. Both report a “precision” figure calculated, in their own published definition, as remediations where the user appealed and the appeal was accepted, divided by total remediations. Tinder’s comes out at 0.188 per cent and Hinge’s at 0.48 per cent, and both filings call the person appealing “the bad actor”. A figure that small is partly a measure of how few banned people appeal at all.

The same two filings carry the numbers that matter to you. Hinge received 86,650 internal complaints about account suspension or termination in 2025, upheld 55,399 and reversed 31,251, which is 36.07 per cent. Tinder received 66,877 and reversed 14,478. Plenty of Fish, on a far smaller base, logged 299 complaints about a decision to suspend or terminate an account, upheld 157 and reversed 119, at a median of five days.

Badoo publishes 127,648 appeals with 9,422 granted, 7.4 per cent, at a median of two hours fifty-five minutes, and it is the one figure to handle carefully: Badoo’s own footnote says that pool mixes blocked members with people appealing a rejected report, so read it as a floor rather than your odds. Nobody here is lying. They are answering different questions and printing the answers side by side. None of it is a promise about your case, and no service, ours included, can make one.

What an automated ban means for your appeal

There may never have been a human judgement to argue with. People write furious appeals rebutting a decision that was never made in the sense they imagine. If a threshold fired, there is no reasoning to attack and no reviewer who weighed your side and rejected it. What there is instead is a first human reader, seeing your words and the internal record together. Write for that person. Give them the context the machine never had, in one calm and specific paragraph. The craft is in how to write a ban appeal that actually gets read.

A report count is not evidence, and the platforms know it. Feeld’s published workflow counts reports without weighing them. And every major app treats false reporting as a serious problem when it suits them: Tinder, Hinge, OkCupid, Match, Plenty of Fish, HER, The League, BLK, Bumble, Badoo, happn and Feeld all make manifestly unfounded reports or abuse of the appeals system a breach that can itself cost an account. A company that punishes false reports has conceded that false reports happen.

File once, and file well. That same rule cuts against you if you flood the channel. BLK publishes it outright: multiple appeal submissions will not be considered and may alter your position in the queue. Badoo and Bumble both reserve the right to suspend or close an account that frequently submits manifestly unfounded appeals. Volume is a hazard here, not a tactic.

In the EU and the UK there are two named levers, and they exist precisely for this. The first is the right to human review of an automated decision under Article 22 of the GDPR, which Breeze grants at a named address and happn states as a right to obtain human intervention. It is the correct question to ask of any company that has told a regulator its enforcement is automated. The second is the Digital Services Act internal complaint under Article 20, six months from the notification, the only deadline that recurs across the industry. Neither is magic, and out-of-court dispute bodies under Article 21 cannot order anyone reinstated, but both are free, named and built for this. The mechanics are in the DSA guide.

Three things not to do meanwhile. Do not dismiss the notification: on Badoo, Bumble and Breeze the appeal form arrives only as a link inside it, with no standalone URL published anywhere, so screenshot it first. Do not charge back: Plenty of Fish, Boo, HER and The League all reserve the right to terminate immediately if you reverse a payment, which converts an appealable ban into a contractually justified one. And do not file an erasure request hoping it lifts the ban: Hily says in advance that a data erasure request will not release you from the prohibition, and Badoo says it keeps the data specifically to enforce the block. The access request is the useful one, covered in the erasure guide.

If you would rather not do the drafting, that is the job. We draft every route your case has, you press send where only you can, and we file the rest in your name. We never ask for your password and never log into your account, because on most of these apps that is a fresh breach of the very terms you are appealing under. If we decline your case, you are refunded in full. Nobody can promise an outcome, because the platform always decides. Start with the free check, or see how it works.

Questions people actually ask

Was I banned by a person or by a machine?

On the published numbers, most likely by a machine. Tinder’s EU filing reports 1,536,656 enforcement measures taken solely by automated means against 1,086,123 that were not, and Hinge’s 2025 report puts its own share at 445,488 of 633,534, about 70 per cent. Neither company tells an individual user which bucket their ban fell into.

Which dating apps admit their bans are automated?

Tinder, Hinge, Plenty of Fish, Bumble, Badoo, Grindr, happn, Feeld, Breeze, Boo, HER, Hily, BLK, The League, Match.com and OkCupid all disclose automated enforcement or profiling that can end an account, in filings, terms or privacy policies. Coffee Meets Bagel denies that its automation makes significant automated decisions, while its safety pages describe automated detection driving blocks. Muzz says flagged accounts are then reviewed by a person.

Do I have a right to a human review?

In the EU and the UK, Article 22 of the GDPR gives you a route to contest a decision based solely on automated processing and to ask for human involvement. Breeze grants it expressly, at [email protected], for profiles its language models rejected. happn states it as a right to obtain human intervention. Most other apps disclose the automation without publishing a human-review address, which does not remove the right but does mean you have to name it yourself.

What is GDPR Article 22 in plain words?

It is the part of European data law about decisions made about you with no meaningful human involvement, where the effect on you is significant. Losing an account is a decision with an effect. The practical use is not a lawsuit, it is a sentence in your complaint: you are asking a human being to make the decision, rather than asking a company to admit its system failed.

If an app says its automation is 99 per cent accurate, why was I banned?

Because that number is measured across every automated action, nearly all of which nobody contested. Bumble publishes a 0.75 per cent error rate and, in the same report, 9,065 of 41,982 appeals granted, which is 21.6 per cent. Badoo publishes 99.60 per cent accuracy, eharmony 98.53 per cent for its Australian quarter. The accuracy figure describes the whole population. The grant rate describes people who appealed, which is the group you are now in.

How many reports does it take to get banned?

No company publishes the threshold. Feeld is the only one that publishes the mechanism, describing a workflow whose purpose is “Automatically blocking accounts when a specified number of community reports are received”, without saying what the number is. Badoo and Bumble both confirm in their EU reports that a strike threshold exists and neither publishes it either.

Does an automated ban make the appeal easier?

It makes it different, not easier. There is no reviewer’s reasoning to demolish, which removes the temptation to argue with a decision nobody made. What is left is the thing that actually works: a short, dated, specific account of what happened, aimed at the first human who will read your file. The platform still decides, and nobody can promise you a result.

Can I find out which system flagged me?

Partly. A data subject access request under the GDPR or UK GDPR gets you the personal data a company holds, and it is the strongest tool for seeing what the record says. Be realistic about the limit: Hinge states plainly that ban reason information is not included in data export requests, and Tinder, Match, Plenty of Fish, BLK, The League and Grindr all reserve the right not to tell you why you were banned at all.

Should I say “I was banned by an algorithm” in the appeal?

Not as an accusation, and not as the opening line. It works as a quiet framing in one sentence, usually near the end: you understand the decision may have been automated, and you are asking for a human review of the facts below. Then give the facts. An appeal that spends three paragraphs attacking the company’s moderation systems reads as a grievance, and grievances get closed.

Do dating apps use AI to read my messages?

Several say so. The League discloses that messages and video calls are scanned automatically, with human agents reviewing what gets flagged and using it as training data. EliteSingles discloses automated review of messages for keywords, phrases and patterns, with the right to interrupt or limit chat functions. Breeze uses large language models on profiles. Where an app is not named here, it is because it does not publish the answer.

Can I be restricted before I have broken any rule?

Hily publishes exactly that: if a profile meets criteria showing the user “is likely to breach the Terms of Use”, access can be restricted or blocked, and it says affected users can contact it to contest. Separately, Bumble and Badoo both report large volumes of accounts whose visibility was limited rather than removed, and mostly by automated systems, which is a state that comes with no notification and often no appeal form.

Should I appeal more than once if the first one was ignored?

Be careful. BLK publishes that multiple appeal submissions will not be considered and may alter your position in the queue. Badoo and Bumble both reserve the right to suspend or close an account for frequently submitting manifestly unfounded appeals, and Tinder, Hinge, OkCupid, Match, Plenty of Fish, HER and The League all list abuse of the appeals system as a prohibited action. One good filing beats five.

What if I am not in the EU or the UK?

The Article 22 and Article 20 levers are European, and there is no equivalent in the United States. What survives everywhere is the platform’s own appeal, written properly and sent once, plus the money track if you were paying, since a ban does not always stop the billing. That sequencing is in the refund guide, and the whole ladder is in the free guide.