Early access opens to the waitlist first. Join the waitlist

Scams and safety

How to Spot a Bot on Dating Apps: Is He Real or a Script?

Why instant replies and perfect English no longer tell you whether a chat is a bot, what Philippine officials found on the other end of scam chats, simple questions to ask inside the chat, why a video call is no longer proof, and what to do next.

Published November 7, 2026 · Updated November 14, 2026 · 27 min read
Contents · 11 sections

To tell if a chat is a bot, don’t go by how fast or how well he writes: instant replies and perfect English prove nothing now. Check whether he answers your exact words, remembers what you told him, and keeps steering the chat to one goal: another app, a link, or money. Passing these checks doesn’t prove he’s honest.

A bot is software that sends or writes messages automatically. In a dating chat the line is blurry, because a person can also copy a script or use AI to help with replies. Philippine officials describe love-scam offices where people work from written scripts, phones are set to send messages automatically, and AI helps keep up automated conversations and fake faces on video calls, so the same chat can be part machine and part person. The useful question is less “bot or person?” than “is he listening to me, and where is this going?” Checks like the ones below can catch a basic script, but passing them can’t tell you whether a person is typing, so your rules on money, links and private details stay the same either way.

This page is about the chat itself: how replies read, what to ask, and what to do. Whether the whole account is real, how to confirm who he is, and the wider signs of a romance scam each have their own page, linked where they come up. No dating app is named here: the checks work the same wherever you met.

How to spot a bot on dating apps: the signs in the chat

To spot a bot on dating apps, look for replies that fit any message instead of yours, a tone that never changes at any hour, details he forgets, and a chat that keeps bending back to one goal. Each sign alone has an innocent explanation. Two or three together are worth acting on.

His answers sound sweet but don’t respond to what I said

Replies that sound sweet but don’t respond to what you said are the sign most worth checking. They show the answer isn’t engaging with your words, whether a program, a copied script or a person who isn’t really reading is behind it.

Say you tell him your lola’s carinderia had to close this week. A reply like “You are so strong, my queen, I am always here for you” fits any sad message anyone could send. Someone who read it would usually ask something about it: what happened, how your lola is, what she cooked. Watch for replies that would still make sense if you swapped your message for someone else’s.

A reply that would fit anyone’s message is a reason to keep checking, whoever sent it.

Someone tired or texting in a second language can give a short, general reply now and then. The sign is a pattern: message after message where nothing you said leaves a mark.

He replies within seconds at any hour: could it be automated?

Replies within seconds at any hour could be automated, but they could just as well be a man in another time zone, or an office of people working in shifts. Speed on its own doesn’t tell you which.

One Philippine love-scam office raided in 2025 had phones set to exchange messages with victims automatically 1, and a team taking turns on one profile never sleeps either. A Filipina domestic worker whose online “prince” turned out to be a fake told AFP: “He kept on messaging me even when I was sleeping,” 2.

A real man abroad can also be awake when you’re asleep. What matters is not when he writes but whether he ever has a life: a day that goes wrong, a late reply with a real reason, plans that keep him away from his phone.

His English is perfect, like a textbook: does fluent writing rule out a bot?

No. Fluent, textbook English doesn’t rule out a bot, and it doesn’t prove one either. Generative AI, software that writes text on request, can write fluent, friendly English for anyone who asks it to.

Old advice treated instant, error-free replies as the mark of a bot. Neither polished nor awkward English now tells you whether a match is automated or honest. The Cybercrime Investigation and Coordinating Center (CICC), the government’s cybercrime agency, says AI lets scammers keep up emotional manipulation through fake identities, “automated conversations” and staged interactions 3. A scammer who writes poor English can now send polished messages, and an honest man who writes well is just a man who writes well.

A reply can feel personal without the person being genuine. Don’t let a convincing chat change your rules about money, links or private details.

I mentioned something earlier and he didn’t remember: can bots keep track of a chat?

Some can and some can’t, so a forgotten detail is a clue, not proof. Whether software can use an earlier detail depends on how much of the chat it is given, and a newer AI tool can keep track well inside one chat. A team sharing a profile forgets whatever the last person on shift didn’t read, and a busy man forgets too.

The useful version of this sign is a mismatch: he asks again where you work after you told him twice, or he mixes up your name, your city or your children with someone else’s. A busy man forgets small things. Forgetting the big things you told him, again and again, is a reason to slow down: he may not be reading, or may not be one man. It doesn’t tell you which.

His messages feel copied from a romance novel: how do I spot scripted flattery?

Scripted flattery is praise that could be sent to any woman: about your beautiful soul, your smile, destiny and forever, with nothing in it that only you could have earned. You spot it by asking whether he’s praising something you actually said or did.

Praise tied to something you said, like the joke you made, the photo of your dog or the way you described your hometown, gives you more to go on. A general compliment gives you less. Specific praise can still be copied, tailored or written with AI, so it’s one clue, not a pass. Big love words very early, before he knows much about you, are part of the same pattern. The CICC describes love-scam offices that use scripted conversations to “emotionally groom multiple victims at once” 4, and lines written for many women tend to stay general.

The same opening message came from two different profiles: what does that tell me?

The same opening message from two different profiles tells you a line is being reused. It doesn’t tell you who copied it, whether the two accounts are connected, or whether a program or a person sent it.

Some real men do reuse a good opener. But when the same unusual wording, the same story or the same questions arrive from different names, the accounts may be working from one script. Treat both chats as unconfirmed, and don’t take either one off the site you met on.

Bot, script or a real person: what is on the other end of these chats?

On the other end of a scam chat there can be a person working from a script, phones sending messages automatically, AI help, or all of these at once. Philippine reports describe these mixed setups. Nobody has counted how often a chat is software alone.

The clearest picture comes from the NBI’s 2025 raid on a love-scam office in Makati. Investigators found scripts on the desktops and phones automated to message victims 1. They also seized AI software that “can generate realistic female faces and superimpose them on scammers’ faces” during video calls 1.

The 167 Filipinos arrested there were the operators; their targets were in Europe and the Middle East. So the raid shows how one such office ran, not what is behind your chat. In February 2026 the CICC described scripted conversations and staged video calls used to groom many victims at once 4.

That mix is why a chat can feel like a machine one day and a person the next. It can be exactly that. A program or a junior worker may send the routine messages, and someone else may take over when you ask something hard, cry, or start to pull away. A change of tone, spelling or memory between morning and night can be a change of who is typing.

A basic script may fail your questions and a well-run operation may pass them. Neither result tells you he can be trusted.

Questions to ask to tell if someone is a bot

The questions that tell you most are ones that need your conversation: something you said earlier, or a playful twist on a normal message. Ask one or two inside the chat you already have, in a normal friendly way, and watch whether the answer engages with your exact detail. They check whether the replies use your words. They can expose a rigid script, but they aren’t a bot test: a person, or AI with the same chat in front of it, can answer them well, and a real man can forget.

  1. Go back to something you said days ago. “Remember what I told you about my sister’s wedding? Guess what happened.” Notice whether the reply remembers the wedding.
  2. Mix something small and local into a normal message. Write “Kumain ka na? Haha, that’s how we say hello here” or tell him you’re having halo-halo for merienda. Notice whether the answer picks up the local detail. He may already know the phrase, and AI can understand it too, so this shows attention, not whether he’s a real foreigner.
  3. Ask him to use your words. “Tell me about your day, but in three words.” “What would you name my cat?” Answers can be silly, and that’s fine: notice whether he plays along with the exact request.
  4. Ask about his ordinary life, then ask again a few days later. “What did you eat for lunch?” “Who’s your annoying coworker?” Details that change or never come are worth noting.
  5. Change the subject on purpose and see whether he follows you or pulls the talk back to his topic.

These may trip a rigid script or a busy operator juggling many chats. They don’t catch everything. A good AI tool can answer a playful question smoothly, and a scammer chatting by hand passes every one, because he’s a real person. A relevant answer shows the reply used your detail. It doesn’t show who is typing, and it doesn’t mean he’s honest.

Some old tricks are worth dropping. On Ask MetaFilter, an English-lang