nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Introduce reply cooldowns to prevent AI-driven spam

Tomislav Tobijas

I've noticed an increasing number of replies that seem heavily AI-generated, often repetitive, generic, or completely missing the context of the original question and existing answers.

Some accounts are posting an almost inhuman number of replies daily. I'm not sure whether AI agents, MCP integrations, or other automation tools are involved, but it's becoming increasingly noticeable.

My suggestion would be to introduce temporary cooldowns or reasonable daily reply limits, at least as a short-term measure, while exploring better ways to address low(er)-quality, AI-generated contributions.

Coming from a gaming community, I've seen something similar happen before. ⚠️ Once bots and automated activity took over a reward-driven environment, it eventually became a zombie playground, with genuine engagement disappearing.

I'd hate to see the Atlassian Community head in that direction.

AI itself isn't the problem. Using it to replace genuine participation is. And lately, I feel we're getting dangerously close to that line.

I'd love to see this addressed before it becomes an even bigger issue. 🙂

One final note: I realize this might be an unpopular opinion, and some people probably won't appreciate me bringing it up. But I genuinely care about this Community and what makes it special, so I'd rather speak up now than watch it slowly lose the human element that brought us all here in the first place. 🙂

P.S. And yes, this post was written with a little help from AI! 😄 I explained my thoughts in my own words (I'm using voice-to-text more and more, hah), and AI helped me structure and polish them. That's exactly my point: AI is an amazing tool when used thoughtfully, but it shouldn't replace genuine human engagement.

Find more posts tagged with

Gathering Feedback

Status: Gathering Feedback

Comments

Benjamin Črnjak

Hi Tomislav,

I just want to give you full support on this!

I share this opinion my self and I'm also generally not against AI. But to be completely honest, I'm more and more leaning towards not loving AI, cause I think it's misused in more and more cases.

Instead of learning with AI, expanding knowledge and quality, I see more and more people using it just to dump input, copy and paste output, and no learning is done whatsoever.

So in that way, I support your request here, as it would be really, I will say it → Stupid ← if we allow that community is full of unverified AI spam.

In that case, I don't need the community anymore, I can find that on any page on the internet, or use an AI myself.

Regards,

Benjamin

Viswanathan Ramachandran

Hi @Tomislav Tobijas

I support the idea. 😎

Do you mean AI generated response/reply to the post/questions asked in the community? If so, yes, it can discourage from genuine participation.

Cooldowns or daily limits might help, but it is individuals who should make thoughtful contributions and follow guidelines.

I'm afraid, if the responses are for recognition/ranking/kudos points, the moderator review of suspicious and repetitive patterns is worth it.

Personally, I like AI, but seems to have lost interest now.

Tomislav Tobijas

Hehe, hvala na supportu kolega! 😄🙏

…In that case, I don't need the community anymore, I can find that on any page on the internet, or use an AI myself…

And exactly! 🤦 Not sure if we'd need forums anymore, and why not replace them with some kind of UI/AI chatbot instead 😅

Tomislav Tobijas

Do you mean AI generated response/reply to the post/questions asked in the community?

Mainly that, but also more and more content is being AI-generated and the thing is, it's hard to know who's using/doing what. For some folks, I could maybe say something as I know their 'style' before and after 'AI introduction' 😐

And the cooldowns are just a temporary solution/suggestion. 🤷

This is a much broader issue than this community/forums, but I believe it should be addressed sooner rather than later 👀

Arkadiusz Wroblewski

Upvoted. Yeah, I've noticed that too. 😅

I'm currently working on some migration projects, so I'm online most of the day and pretty active on both the forum and Slack.

Even then, I sometimes find it almost impossible to believe how quickly certain people spot a new question and somehow manage to post a whole wall of text within minutes. 😂

For example, my guilty pleasure with AI is text-to-speech and language correction. As a non-native speaker, I sometimes feel the pressure to write 100% gramatically, because other way people taking you as dumb 🤠😂

Rodrigo [Elevatic]

+1 to this idea.

One of the great ideas of the community was the real engagement, genuine interesting questions from news and experts causing interaction, discussion; a friendly environment.

I am unsure whats the goal of the bots here, but would love if we can stick to the roots.

Arkadiusz Wroblewski

@Tomislav Tobijas

I think that was implemented. I was testing it yesterday.

Tomislav Tobijas

@Monique vdB Have there been changes on this front?

Tomislav Tobijas

The more I think about it, the more I feel that cooldowns would solve maybe 10% of the problem, if even that. They might slow things down, but they wouldn't necessarily improve the quality of contributions.

The remaining 90% would probably require some kind of intelligent detection and monitoring system in the backend.

For example, such a system could analyze several signals together:

How likely a reply is to be AI-generated (although AI detection alone is far from reliable).
Whether the reply actually addresses the question or just repeats generic information.
The user's recent activity, posting frequency, and response times.
Similarities between their replies across different discussions.
Whether suspicious patterns continue over time, including replies to older threads.

If multiple indicators cross certain thresholds (from a math perspective, let's say threshold is 80%+), the system could flag the account or specific replies for Community moderator review, rather than automatically taking action.

Of course, this gets complicated very quickly. Someone using AI to improve grammar or structure shouldn't be treated the same as someone generating dozens of meaningless replies just to collect points.

And that's probably the biggest challenge here: how do we distinguish genuine AI-assisted participation from AI-driven spam or point farming? 🤔

It almost feels like we'd need AI to detect other AI-generated content, combined with behavioral analysis and human moderation. 🫠

I don't have a perfect solution, but I'm increasingly convinced that cooldowns would only be a temporary measure. (These seem more related to spam in general)
The actual problem requires something much broader, and I genuinely hope we can find a way to address it. 🙂

If you guys implement a backend engine for this, I'd really like to see it 👀😄

Arkadiusz Wroblewski

I agree 100% that this needs a broader approach.

We also need to be careful that we don’t discourage participation so much that we slowly kill the forum. AI is widely used today, for translations, grammar, structuring an explanation or checking an idea. That alone tells us very little about whether a contribution is useful.

The reward system needs attention too. Today I answered one Sourcetree question deliberately, then two others without noticing how old they were because people were still posting there. Around 2,100 Kudos for three replies? A newer member could put real effort into 20 answers and earn nowhere near that.

My worry is that we gradually become a circle of established members recognising each other, while newcomers struggle to find space and recognition.

Detection and human review can help, but we also need to review what the programme encourages. The focus should be on accuracy, relevance and whether a reply adds something useful. Otherwise, we risk discouraging helpful AI-assisted contributions while leaving the incentives for farming untouched.

But I don't worry. The community team has this, and if they need help, they have us.👀

Haha, regarding your Community Review.

👀