We’re having some pretty nasty issues when it comes to testing and releasing our bot we’re making for a client.
Our client is planning on a Twitter Ad Buy to advertise the launch of the bot and they’re wondering if that’s a good idea given the state of Twitter Bots at the moment.
We were “load” testing our bot today with about 7 people using it at the same time testing random functionality.
90% of messages the bot send back include a RANDOM GIF along with a text message. We have a very large pool of GIFs to pick from.
Today almost half of the messages we sent back were marked as “This request looks like it might be automated. To protect our users from spam and other malicious activity, we can’t complete this action right now. Please try again later.”
How is this possible? Some users got this on their second or even first message!
The best percentage of messages successfully sent we’ve seen during testing is at most 90%. Today it was 50%.
We’re well within the 5 DMs per 24 hours limit on all use cases. Not a single link has been sent from this bot in at least 3 weeks.
Could this be due to not using a verified account to run the bot? Could we be white-listed to a beta environment when using the direct_messages/events/new?
It’s particularly frustrating to be told this message looks automated! Of course it’s automated, we’re using a endpoint with an adoptive rate limit DESIGNED for bots (5 per 24 hours per message received). This can’t be by design unless this endpoint was only designed for Tweet deck or something like that. How else could chat bots work on Twitter if the requirement is that every message needs to be unique?
We’re afraid to launch until this issue is sorted out but we’re not sure how much longer we can convince the client to delay without losing them completely.
What can we do?
Let me know if there is any more information I can provide.