🤖 Using AI on the forums: A guide to our community rules

[#1]

Atlassian AI content guidelines allow the use of AI for formatting, language, and polishing, provided the core contribution is human-written, accurate, and helpful. Wholly AI-generated content without original thought is not permitted.

---

Hey community 👋

Earlier this year, @Monique vdB wrote an article entitled Keepin' it real: forums that are AI-friendly, but human-first, where she talked about the importance of keeping our community human-centred. Human expertise, real experience, and genuine connection are still what make this community valuable. That hasn't changed.

But since AI use has evolved so rapidly and continues to evolve at a speed that could give you whiplash, I wanted to give an update on how we in the community management and moderation teams view AI-generated content.

What is actually fine, and what should you report

First up, we recently tweaked the AI part of our main rules of engagement to now be:

AI usage for the purpose of formatting, language, spelling, etc. is allowed if utilized as a tool, but the core of your contributions should be written by you. AI should not be used as a substitute for providing accurate answers and knowledge to our users, and you are responsible for ensuring that any information you provide is accurate.

We removed the line that asked members to disclose wholly AI-generated content. Why? Because if content is wholly AI-generated, it shouldn't be posted as-is in the first place. Disclosure made sense when we were still figuring out where the line was; now that we're clearer, the rest of the guidelines cover it.

I want to also call out a key phrase here: "the core of your contributions should be written by you." AI is a tool — like spell check, like Grammarly, like a thesaurus. (I like to ask Rovo to check over what I’ve written, suggest any tweaks and point out any gaps in information or flow!) Using it to polish, format, or clarify your own knowledge is fine. Replacing your own knowledge entirely with AI output is not.

✅ With this in mind, the following examples of AI content are fine:

• It's well-structured or uses formal language - some people just write clearly! We encourage our members to write in their ‘voice’ and would never want that to be stripped out. A well-organised answer with headings and bullet points isn't automatically AI slop. Our guidelines explicitly allow using AI to format and structure your own knowledge, and a clearer answer is a better answer for everyone.

• The person used AI to help format or translate their answer - our community is global, and not everyone writes in their first language, and AI can help bridge that gap. If the answer is helpful, assume positive intent and focus on the content rather than the tone.

• It uses common phrasing - phrases like "I hope this helps" or "Here are some steps you can try" are community staples, not AI red flags.

• You have a gut feeling but no specific concern about quality or accuracy - if the answer is helpful, accurate, and solves someone's problem, it's doing its job regardless of how it was written.

🚩 Some AI use does cross the line. Click "Report to Moderators" if you see content that:

• Is wholly or mostly AI-generated with no original thought - a wall of text that reads like a ChatGPT response with no personal experience, no product-specific knowledge, and no indication the author actually knows the answer.

• Contains inaccurate or hallucinated information - especially if it could lead someone down the wrong path. To be clear: we don't moderate people for getting an answer wrong (we've all been there! 😅). But AI-generated content that confidently presents false information as fact (without the author having checked it) erodes trust in the community in a different way.

• Is promotional or spammy content dressed up in AI-generated language - for example, AI-written articles that exist primarily to promote a product, service, or link. (We have a whole separate Partner Rules of Engagement that goes into promotional content in more detail, if you’re interested!)

• Is undisclosed AI content posted at scale - a pattern of posting frequent, clearly AI-generated answers across multiple threads without disclosure.

So, we're not here to play AI detective, and we're not asking you to be one either! For now, the TL;DR is: focus on quality, not on tools, and report content that is harmful, inaccurate, or dishonest, not content that simply sounds like AI.

We'll keep revisiting our approach as AI evolves, and we'll keep you in the loop. Thanks for being part of what makes this community great, and for keeping community members' best interests at heart. 💙

Disclaimer: I used Rovo to help herd my scattered thoughts into something actually useful 😜

[Tim Martin]

@Mubashir Iqbal , how ironic...  - glad to see the policy in action already!

@Kate C_ , yeah, it's going to be an ongoing battle. Maybe we should introduce a similar concept to the Enhanced Games? Make an alternative forum where AI can have at it.

On a serious note, it's a good stance and I continue to back Atlassian on this. LLMs are great for clarity, accessibility, translation, etc. Super useful, but some people will take usage too far - for a variety of reasons and incentives. I believe the average person is still quite wary. I know I constantly get that uncanny valley feeling (the "AI ick" as the young'uns probably say). 

I think our most powerful tool is to hold each other accountable. *shrug*

[Kris Klima _K15t_]

Thanks @Kate C_ for this.

I'm a journalist turned tech writer turned content architect.

When I write, it's structured. I use words like 'plethora' and 'realm'. Technical writing is structured by nature. Structure carries the meaning. If you're helping someone in the tech realm, you speak tech, you organize, you structure. 

AI was trained on writing that peole like me have been producing for decades.

Do I use AI? Yes. The same way I'd use an intern, spell checker/language editor, and an editor as such. 

[Kate C_]

@Tim Martin LOL I think if we had a forum solely made of AI generated content, no one would hang out there... 😂

Thanks for your comment! Holding each other accountable (while assuming positive intent) is the way to go.

[Kate C_]

@Kris Klima _K15t_ thank you for your insight and background, as always! 

Your point that AI was trained on the kind of purposeful writing that people like you have been producing for years is something maybe not enough people stop to consider.

And honestly, "AI as an intern/spell checker/editor" is great framing for using it with this type of writing. I might steal that! I actually joked with a colleague this week that I now have her and Rovo checking my work over. Twice the insights 😄 and same energy as what you're describing.

[Kris Klima _K15t_]

@Kate C_ 

Feel free, that's the gospel I preach ;) 

It's the best journo practice. Even if you don't have a dedicated editor, always have a second pair of eyes checking whatever of significance you wrote. 

And AI is good at that if instructed properly. Just as it's in doing the research or drafting one would assign to an intern. But it's important that assignment be made by a person who understands the assignment.