An update for the delayed bot verification request process

Today we would like to provide an update for the delay in processing chatbot verification requests as mentioned in the Twitch chat documentation. As several developers have witnessed and inquired about, chatbot verification requests have gone unanswered for some time.

The verification process was paused a few months ago and was only intended to be in this state for a short period of time while necessary updates were made to the process and privileges regarding verified bots. The purpose of these updates are to improve expectations between developers and creators, and reduce abuse vectors. Guidelines will also be published to make it much clearer why a developer would require this status and exactly what is provided as verification should only be needed for high consumption usage. This effort was delayed and, unfortunately, was not communicated with more transparency or urgency. We apologize for the unexpected silence.

Unfortunately, we cannot provide a specific date for when the request queue will resume processing, but what we can promise is that the effort to complete the tasks mentioned above has begun, all received requests will be processed, and we will have a further update for these changes on July 24 26, 2021. That update will take place on this announcement below.

4 Likes

As promised, we are providing a further update regarding the chatbot verification process.

Pending Chatbot Verification Queue

We will begin processing the backlog of chatbot verification requests this week and they will be completed in the order in which they were received. We appreciate your patience during the time needed to do so. This post will receive a further update when the backlog has been completed.

Updates for Verified Chatbots

As of today, the documentation has been updated to reflect the following changes regarding verification.

1. Removing Known Bots Moving Forward

When the known bot status was released, it provided a way to identify Twitch chatbots with good intentions for the Twitch community and assist in our efforts to reduce malicious activity. With further advancements in the past several years and some confusion between known and verified chatbots, we have decided to simplify chatbot status into only two groups instead of three; non-verified and verified.

Any Twitch account that already has the known bot status will continue to have this designation and retain the Whisper limits of 10 per second, up to 200 per minute, for 500 accounts per day. However, no further requests can be made for this status moving forward.

2. Requirements For Requesting Verification

We’ve added wording to the documentation to explicitly state that verification requests should only be submitted for higher rate limits once the default rate limits have been reached. Submitting a request for other reasons (e.g. asking about a higher Whisper limit) or before the default rate limits are reached will not result in an approval.

3. Verified Bots Rate Limit Change

Verified bots will continue to have a higher rate limit for commands/messages, authentications, and joins. Verified bots as of today will no longer receive higher rate limits for Whispers. They will be subject to the same rate limits as a regular Twitch account.

Any Twitch account that already had verified bot status will continue to have the higher Whisper limit of 20 per second, up to 1200 per minute, for 100,000 accounts per day.

Thanks

Thank you again for your patience while making the updates above and continuing the verification request queue. If you have any further questions, please feel free to add them below.

1 Like

I’ve seen people have issues sending whispers at all from new accounts (TMI would reject whispers with a NOTICE saying “Your settings are preventing you from sending this whisper”). I’m talking from personal experience. Only answer to those was to get the account verified as a bot. Since you are removing the higher rate limits for verified bots. What is the new way of fixing this problem?

1 Like

Verified bot was never the answer. The previous documentation stated that it had some impact on Twitch’s anti-spam measures, but it was by no means a definitive way to allow bots to reliably use whispers. There have been many examples over the years of devs with bots that are normal users, known bots, or verified bots, that have all experienced issues with Whispers.

1 Like

I’m just wondering what does removing higher whisper rate limit from verified bots is supposed to do / fix?
If anything, higher rate limit helped bots not to run any issues while whispering users. Now this change is just making our lives harder and nothing else…

1 Like

Does your bot regularly exceed what would be the whisper rate limit? For the vast majority of developers with bot whisper issues it had nothing to do with rate limiting.

Yes, it tends to happen that bot sends more than 3 whispers in the span of 1 second (not very often, but I’ve seen it happen on one of the bots one of my friends host when streamer’s chat was very active and bot had to reply to many people at once).
I don’t have a very big community myself, but in some bigger ones bots whisper over 50-60 users in the span of 24 hours with plenty different messages, such as: warnings for violating channel rules with timeout/ban explanation to the offender, updates about moderation reports that were made by trusted non-moderators or even replying to moderators, who execute commands such as “nuke” within whispers.
So these rate limits currently present are not sufficient and it makes it impossible to set up and host another instances of these bots in any bigger communities with active communities without risking bot banned from whispers for 24 hours for no apparent reason.

All of this sounds like you need/should use a product more “reliable” than Whispers.

And for most of those use cases I’ve done exactrly that with my own tooling.

Edit: thats not me saying don’t use whispers, thats me saying I’d never of considered whispers for this in the first place

If bursts are your issue, implement a message queue.

Whispers are not designed for this sort of purpose, so if you’re trying to force a solution together using it then you’ll run into the limitations of the Whisper system. Rather than open up the whisper system for potential abuse by expanding rate lmiits, and having to have the devs apply for and wait for rate limit increases, it makes sense to use/develop a solution that’s more fit for purpose as many of the larger streams/events or even small streamers with structured moderation teams already do.

Will I really have to force people, who came to the stream without any knowing any background of what moderation “tooling” is used, prompt to authenticate on some dashboard or extension and authorize the application just so they can get a single message regarding some moderation message regarding a warning they received? Realistically speaking, people won’t ever bother opening up chatbot’s website and learning things there ever.

Using basic twitch’s features is more convenient than having to open up a browser but fair enough.

Or you just have the bot say the warning in chat, the medium in which the offense took place, so that the offending user is warned, and everyone else is reminded of the rules. Even in high traffic channels this has proven effectiveness.

For moderators yes. They will be educated to use the tools that the channel uses. And moderators may keep the tool open for the duration of the stream, even more so if that tool embeds the stream and chat. A “custom” third party mod view, which Twitch itself now has. See → Twitch Twitch Moderator dashboard/modview

An Example:

For viewers/users no, I wouldn’t whisper I’d use ban/timeout reasons. Since I don’t expect good deliverbility rate on whispers, especially to new people whom are new to the channel. And/or @ messages in chat.

I’d expect having even worse rate of reaching the user with timeout reasons (especially for short timeouts or purges) because users either use the mobile app (which doesn’t support ban reasons, i just tested it :󠀀) or can’t read fast enough. Whispers are just plain better because they are older, thus more supported and they stay permanently so there is no problem with reading them fast.

Sounds like a bug that should be reported so it can be resolved.

And then miss the whisper if they even are on a whisper supported platform and/or have whispers enabled anyway.

fixed it :stuck_out_tongue:

Often find repeat offenders didn’t read the channel rules un the first place or any message that gets sent to advise them of the rules. But it’s a mixed bag

Except that whispers have proven unreliability for bots, as they are not designed for this use case. Regardless of verified/known status a significant portion of bots can’t send whispers with the degree of reliability that would be suitable for sending moderation related messages.

What do you consider the designed use case for whispers to be?

If thats an open question to the forum.

I consider whispers to be used by two people whom are not bots/automated entities to send a private message to each other, where there is no channel context.

Since whispers don’t “belong” to a channel. When I’m off watching some streamer and someone whispers me going “can I get an unban” I first have to ask them which channel they refer to, for example.

IE: for channel agnostic messaging.

First party usage, powering the functionality of the twitch.tv site itself and other Twitch apps.

Much like PubSub, the design is primarily a first party product, which developers are allowed to have access to as a secondary concern but not the primary function. Unlikely products like EventSub, or Helix, which are products developed for the 3rd party community as Twitch uses separate services internally.

All 3rd parties attempting to use whispers have to do so within the limitations of whatever anti-spam measures, restrictions, rate limits, and other limitations that Twitch don’t disclose for obvious reasons. Because of this whispers are not a reliable communications medium for what a lot of people are wrongly attempting to use it for.