Bumble asked because its "Review before you send" feature thought something in your wording could come across as harmful, offensive or inappropriate. A machine-learning model checks messages as you type them, and when it flags one, it prompts you to review it, edit it, or rethink sending it at all. Whether you can still send it depends on how sure Bumble is: a clear break of its guidelines has to be edited first, while a borderline one can still go after a nudge to reconsider.

What the check is looking for

Bumble's help page describes the feature in broad terms. When you type a message, a machine-learning model checks it for language that could be harmful, offensive or inappropriate. The goal, Bumble says, is keeping conversations respectful and comfortable for everyone.

The page names the categories, not the words. So the useful question is not which word set it off. It is which line in your message a stranger could read as crude, aggressive or sexual. That is the line to look at first, because that is also the line she would have stopped on.

When you can send it anyway

Often you can. Bumble says that in cases where it is not fully sure, you still have the option to continue, although you will be encouraged to reconsider first.

In that case the prompt is not a block. It is a pause. If you read the message again and it genuinely says what you meant, in a way she would take the way you meant it, you can carry on.

Bumble also lets you give feedback directly from the prompt if you think the message was flagged by mistake. That is worth doing when it is true.

When you have to rewrite it

If the message clearly breaks Bumble's guidelines, Bumble says you need to edit it before it can be sent.

Take the hint. Bumble's safety page says hate speech, sexual harassment and fetishization are banned. Its content rules say that when those rules are broken, it may remove content, warn the member or ban them. The prompt is the cheapest version of that conversation you will ever have with the app.

A different nudge for a plain hi

Not every prompt is this one. Bumble's page on how conversations start says that if you type a very simple opener like "hi" or "hey," you may see a gentle suggestion to make it more personal. That message is not blocked. You can send it as it is or change it.

If that is the nudge you saw, nothing was wrong with your message except that it was weak, and the app is right about that. On Bumble, you and she each get one opening message, and you both have 72 hours from matching to send it. Spending yours on "hey" gives her nothing to answer.

Bumble's own tips for a first message include asking a specific question about something on her page, or mentioning a shared interest or one of her photos. Do that and you are no longer sending the kind of opener the nudge is aimed at.

Why the rewrite usually works better with her

Even when you could push the flagged message through, rewriting it is usually the stronger play, for a reason that has nothing to do with the app.

A line reads differently on a screen than it sounds in your head. Sarcasm loses its tone of voice. A tease that works across a bar table reads as an insult from a stranger with no face attached. A sexual comment that might land late in a good evening lands very differently in message three, from a man she has not met.

Early on, the job is simple. You want her comfortable enough to say yes to a drink. Every line that makes her less comfortable costs you some of that, and a line crude enough to trip a safety check usually costs a lot. If it goes badly, trying the same angle again only digs deeper. Back off, get the mood light again, and move toward the date.

Two rewrites that keep the edge

You do not have to make the message bland. You have to aim it at something she chose rather than something she is.

A comment about her body becomes a comment about what she put on her page. "That is a very confident hot sauce ranking. Defend number one."

A tease that sounds like an insult becomes a tease with a way out. "You look like someone with strong opinions about brunch. Am I wrong?"

Both are still playful. Both give her an easy reply. Neither would make her wonder about the man who sent it.

If you think the flag was wrong

Send the feedback from the prompt, and then read the message once more as if a stranger had sent it to your sister. If it still reads fine, and Bumble gave you the option to continue, send it.

If it does not read fine, you have your answer, and the app did you a favor before she had to.