Clarification on Automatic Replies



I would like to make an auto-reply bot and I wanted to check that my understanding of the automation rules is correct. My bot is simple. I manually post a tweet that says something like “Reply to this tweet for x! (make sure to include the word ‘y’)” where x is some kind of information I provide, and y is a keyword. The bot will wait for a user to reply to that tweet, and then will reply to their reply with whatever “x” is.

For example, x might be a system to score your account, so I would post “Reply here (include the word ‘score’) for your score!” and then a user would reply (and the reply would be something like “get my score”), and the bot would reply with their score.

I read through the Automation rules and I see that automatic replies are only allowed if the following rules are followed.

  1. the recipient or mentioned user(s) have requested (i.e. opted in) to be contacted by you on Twitter or have clearly indicated an intent to be contacted in advance

  2. you provide a clear and easy way for such users to opt-out of receiving automated replies and mentions, and promptly honor all such opt-out requests

  3. you only send one automated reply or mention per user interaction

  4. the automated reply or mention is a reply to the user’s original Tweet (if your campaign is based on users posting a reply to your Tweet)

I think that my bot is compliant with these rules:

  1. The post clearly states that replying to the post will result in an automatic response, so I think that this constitutes a sufficient opt-in requirement for point #1.

  2. The user will not receive any reply if they do not reply to the post, and they will also not receive a reply if they reply to the post without the specific keyword. To me that seems to be sufficient for the opt-out requirement for point #2.

  3. I only send one reply for each reply to my post, so I think I have point #3 covered.

  4. The reply is a reply to the user’s original reply to my post, so I think that I have #4 covered.

I made a prototype of the bot, and it seemed to be working fine for a while (it got over 400 replies), but it eventually got write-restricted. I don’t yet know the reason for this (and I’m sure twitter support will get back to me eventually…) but I wanted to run my general idea by here to see if there is anything obviously wrong with it. I am pretty sure I am following the rules correctly, though I am slightly unsure if I need a stronger opt-out solution. What do you think?


It’s probably a case of too many people getting the exact same response and triggering an automated flag on Twitter’s system. I’d recommend maybe including the screen name of the person so “@UserA, your score is 1.” rather than having the possibility of 10 people all getting “Your score is 1” and looking like spam.


I actually did do that. I ran it twice: the first time the response looked
like this:

@userA userA, your score is n” where userA and n were different.

The second time it was similar, but I also uploaded two images in addition
to the text. Both times I was flagged and write-restricted. Also, a few
users responded more than once, which meant my bot did send out the same
exact response a couple times- could that have done it? Do I need to make
sure that I never send a duplicate response, even if it is to a separate


I can see why users responding more than once could trigger a flag. I can’t really give you a solid answer though. Just know that Twitter deals with a lot of spam and therefore they have some pretty sophisticated systems that sometimes result in false positives. While inconvenient, it’s understandable.


I appreciate the tips. I hope someone from Twitter comes along soon and verifies whether or not this is what is going on.


Is anyone from Twitter going to reply? It’s been several weeks.


Hi! We aren’t able to comment on specific applications on the forums, but we’d followed up via your support request. Please respond to that request if you’re still having issues with this app or use case.

In terms of the more general question you asked about our detection systems: We appreciate your diligence in trying to comply with the Automation Rules, but I’m not able to comment on the specific signal that might have triggered your app being write-restricted. We use a wide range of automated systems to detect spammy or policy-violating activity via the APIs. One of the signals we look at is duplication (since posting duplicate Tweets is prohibited under the Twitter Rules and Automation Rules) — so yes, it’s generally a good idea to avoid repeatedly tweeting the same thing. But beyond that, we’re not able to reveal the specifics of our algorithms and systems.


I was asked to provide additional details about the app about 3 weeks ago. I provided those, and then I never received any follow-up on the support request (case 61050072) in my email. I eventually gave up hope of a response and attempted to start over, implementing all the changes I assumed needed to be made to make the bot compliant. Unfortunately, it was flagged again, so I am obviously still missing something here.

If I actually did receive some sort of follow-up, and just missed it, then Is there a way I can look up my case number and see the follow-up?


I’ve tried to do the same thing: make a bot which acts like a text adventure game as a little christmas present for my friends on Twitter, but its responses are being hidden because text adventures always reply with similar content :confused: It was a fun little project, but even though I thought I was complying with the rules, making only one reply per command, and never sending unsolicited messages, oh well…