"We can confirm that liking based on keyword search is aggressive."


#1

I don’t know if anyone here can actually provide some assistance here. We’ve had a few suspensions due to ‘spam complaints’ which, after several emails back and forth with the Twitter Platform Operation were resolved. We explained how the application works and there were no issues with the way it works.

Today, we received a suspension again for ‘spam complaints’ (what this means is a mystery along with a lot of Twitter’s responses). Again, a few emails back and forth and suddenly the application breaks their ‘aggressive activity’ rules.

Apparently: “We can confirm that liking based on keyword search is aggressive.”

So, if I’m understanding this correctly, if a user was to login to Twitter, type in a search term in the search box (as that’s why it’s there) and see a tweet they like, then liking it would be aggressive?

We are not liking anything in bulk. We are not liking anything automatically. We are not liking multiple tweets a second. As far as we can work out, nothing we’re doing is aggressive.

Sorry to rant, but there are so many apps out there that allow for automated liking of tweets (some liking 7000+ a month) and they get away with it. But we try to build an app that’s within Twitter’s rules and its suspended continuously?!

I just wondered if anyone else has experienced anything similar where they’ve received completely different answers/responses from the team at Twitter Platform Operations?


#2

I imagine “spam complaints” means other users are reporting your accounts or tweets as “Spam”

It’s hard to say what could be seen as “aggressive” without knowing exactly what you’re doing.

I don’t know how exactly the anti spam system works - but I’m pretty sure the “BotMaker” system changes over time and adapts, so you can’t really get a clear limit of what’s “aggressive” or not or a clear set of reasons explaining a suspension - that could be why you’re getting different responses.


#3

Sure we get that and have worked with Twitter to understand where the spam complaints come from and are working to fix our app.

But we were thrown when they said “We can confirm that liking based on keyword search is aggressive.”

That makes absolutely no sense to us or anyone we’ve mentioned it to. What this person is effectively saying is that searching on Twitter for tweets and then reading a tweet you agree with and then liking it is aggressive?

I’m just not buying that at all. If we were liking 100 tweets and hour, even if we were liking 20 tweets an hour then I can understand this being aggressive but liking 1 tweet over the course of an hour or two, aggressive??


#4

I’m guessing that there are different rules for tweeting, following, liking tweets, and other “write actions” made through the API as opposed to the web or mobile - I don’t think it’s the volume that’s important, it’s more where the request comes from, and probably a bunch of other features.

I’m sure the system remembers patterns of actions and resulting spam reports, and is much more likely to classify something as spam in the future if it was flagged before. Also, you might be getting classified as spam if other, clearly spammy accounts perform some actions similar to something you’re doing, because the system is twitter wide, and it learnt that a certain pattern leads to spam.


#5

It’s just the word of saying searching for tweets using a keyword and then liking that tweet is aggressive.

Just makes no sense to me.


#6

How did you tackle it at the end? I am in a similar situation right now: kept being suspended although none of the actions are automated.

Twitter support is non-existent and helpless, so far I have been dealing with automated & vague replies from their system.


#7

I can assure you that there is a team of humans behind the automated emails. We have to be able to deal with support queries at scale, so many of the response flows start off with what might appear to be an automated reply, and after a couple of exchanges may result in manual review of a case. I’m sorry if you are currently frustrated by the situation you find yourself in.

We often cannot be 100% specific on the algorithm behind our antispam systems because doing so would just lead to more efforts by bad actors to game the system - plus, the systems use machine learning to detect unusual patterns of behaviour.


#8

Gave up! Twitter support was very poor. They kept moving the goal posts so we thought we were compliant and then they’d change their mind. Different support people were dealing with our requests and each were giving different answers so it’s as if they’re making it up as they’re going along.

I’d avoid working with Twitter if you can and look at the other social networks who provide more support for developers/API.


#9

Hello Andy,

Thanks for your message - indeed, you are still one of the good human beings I recognize on this forum, wish you could be dealing with those email supports so many of us sending and feeling like they are going nowhere.

I don’t expect you guys to give me detailed insights to how your algorithm works :slightly_smiling_face: All I want is to get a clear and concise answers which part of my app caused the bot to trigger so I can finally fix it (if there is anything to fix actually).
Cause right now is frustrating because I am wasting money and trust of my customers, your support is wasting time dealing with people like me and none of the sides can come up with an action plan which is all I need to move forward.

I understand that you are dealing with hundreds similar cases everyday but Twitter support could decrease an amount of tickets and follow ups if someone actually read the initial message you let the user include in “Details” field while submitting the form. A better customer service never hurts.

The initial message I sent from your official support form included 4 points I would need from Twitter in order to quickly resolve this problem on my end and guess what? I got 2nd automated message 3min after asking me to provide them few details.

So just now I saw someone actually opened my email (as I get notified everytime someone opens my emails) but then… no answer. In meanwhile I keep losing money and users and seems like no one understands that this is business case and I need solutions here.

If there is an option I could get a dedicated premium support from an agent at Twitter and pay monthly fee for it, that would be brilliant and your customer service could skyrocket automatically.


#10

This sounds to me like a huge waste of time then - not just for you but also for them. Seems like any trials to reach them out and actually get a real help are pointless.

So how did you manage at the end since you are still running Twoostly?


#11

Each user much register their own application to use the app.


#12

But isn’t that against ToS from what I read? (someone asked about this before if I am not mistaken)


#13

Hi twoostly

I can confirm a similar situation too with our company account.
We have our account excluded from notifications 3 days ago without any warning or explanation.
We tried to contact support throughout all forms, we try to verify our account, we even tried to contact sales to see if it helped.
No one answers anything.
I get the idea that support is just statistics and data.
If an account is tagged as spam or somekind of beaviour is considered spammy by a twitter algo we are in some kind of kafka trial.
Twitter doesn’t validate that assumption and support just doesn’t care or answer anything.
I tried to see if this behaviour was similar in api and all I got was a flag of off topic and some kind words from twitter staff without any real consequence.

I understand that there is a human team behind twitter but it would be great if twitter considered that there are also humans behind there user accounts!

So, I can only say that, we, as a company and as professionals are very disappointed with twitter support.


#14

Hi,

Update:
Today I noticed that twitter penaly was removed from our company’s account.
The account is normal right now.
I don’t have any feedback from support that explains me exactly what was the main cause.
Favourite count is messed up with 5k likes that we can’t access.

But, as a rule I would advice everyone not to favourite tweets based on keyword search.


#15

Favourites / Likes are limited to the last 3200 in the API, in the same way as a user timeline btw


#16

Thanks IgorBrigadir.

So once we overcome 3200 likes we can no longer delete them or view them, right?


#17

I’m not really sure - i think they’re still there, but you’ll need to know the exact favourited / liked tweet ids to find them.


#18

Just started running into this problem. Not performing automated likes, but have seen my account suddenly excluded from notifications without notice. That it even happens by performing likes manually is troubling. I understand curbing automated likes, but where is the line on regular manual use? If there is a limit that leads to likes being flagged I would love to know so I can avoid it.

Seeing developers offering apps that curate tweets to manually like, but the account will then be flagged & ghosted from notifications. Extremely troubling as a business interacts with users daily via twitter.


#19

If any of Twitter’s automated antispam and anti-abuse tools notice unusual account or application activity (often caused by API automation, but not always) then an account or application may be flagged. The precise rules that the systems use are difficult to document since they are a) intentionally designed to combat bad behaviour on the platform, so explicitly listing them out would be counter-productive and enable attackers to find ways around them; and b) partly based on machine learning models which adapt to different scenarios.

More broadly - please use support.twitter.com for queries about accounts, notifications etc. This forum exists to help developers with API questions.

We cannot offer any assistance with issues related to accounts (suspension, verification, usernames etc), or usage of the twitter.com website, analytics or ads dashboards, and the iOS and Android mobile apps in this forum

I’m going to close this topic out, as the original API policy questions were answered previously.


#20