Two weeks ago I published a piece explaining exactly why AI detectors are unreliable. Then Sloan flagged me.
My argument was simple: AI detectors ...
For further actions, you may consider blocking this person and/or reporting abuse
Update
Any further questions and inquiries on the moderation process you have, please comment here:
Insights on Sloan and the MLH acquisition
Ask a DEV Community Mod!
Hey Daniel. There seems to be misconceptions.
Counter point: You can still write well even without AI. I have proof read many articles in the past that weren't even AI generated or AI-assisted. It all comes down to your use of using AI and whether you are using AI as a tool or using it to mass produce articles, which creates unnecessary competition that overwhelms new users.
It's me lol. But for real: ClassifierAI has NOTHING to do with the decision making on sending the sloan messages. I have been using via GPTzero and also reading the article to determine the deciding factor and the quality I believe that meets the criteria.
Q/A time
1. Why do I need to disclose it?
Think of it this way. If you are writing a publication or a research paper, it is good practice to disclose the use of AI, even it is use as a tool. It is very very very important to be transparent of your work regardless. It is even treated as if you do not disclose the use of AI, you will get penalized.
2. It will look bad on my article. Why I need it?
Okay counter point: I left a disclaimer on my submission post for Gemma 4 that some parts are AI-assist and I am fine.
Gemma 4 Challenge: Write about Gemma 4 Submission
Should you use Gemma 4 for your Development? A Multiversal Analysis to Determine if Gemma 4 is Right for You!
If it left you a bad taste, maybe something happened before then or how your reputation were built up in the first place. I left a disclaimer because genuinely, I am an honest person and I want the reader to understand what they are reading.
I would love to hear from you and your thoughts about this. If you have any concerns, let me know! :)
Good write-up ♥️
My Take:
AI detection will soon become useless in many cases.
A more pragmatic approach for a platform such as DEV:
No need to change existing rules, but be more human when enforcing them.
Employ "innocent until proven guilty" policy.
Let's see how it all plays out!
"Detect quality not AI or human" is the only frame that survives the next 2 years. The detection arms race is already lost .
xulingfeng just documented identical content getting suppressed with and without disclosure. If the filter can't distinguish those two cases, it's not detecting AI. It's detecting something else and calling it AI. Quality is at least a signal that doesn't become useless as models improve.
Issue mentioned and resolved: dev.to/francistrdev/comment/39hdd
@dannwaneri
Thanks for writing both pieces and keeping this thread going — genuinely helped get this resolved. Appreciate you 🙌
Updated the essay. @francistrdev clarified that ClassifierAI wasn't used on DEV.to and the flagging decisions were based on GPTZero plus reading the articles. Corrected that in the piece. Appreciate the clarification.
Thanks Daniel. I will be making a post about this and take in question on moderation. Stay tuned.
A human being reading and judging, with the intention to keep the system clean and useful to other humans - that's valuable. ❤️
But rest is just noise, whether it's GPTZero or ClassifierAI 😏
I ran the same test. First with disclosure — gone, deleted it ([link]). Removed the disclosure, republished identical content — feed showed up, then vanished within a minute ([link]). Two posts, same result. It's not the disclosure. The content itself is getting filtered at publish time.😂
And now only my followers can see it — non-followers can't.
Both of my latest articles are the same — only followers can see them.😅
That's the most useful data point in this whole thread. Same content, disclosure or not, suppressed either way . That rules out the disclaimer as the variable entirely. Something is running at publish time that's separate from Sloan and separate from human moderation. Followers-only suppression isn't in the guidelines anywhere.
Worth tagging Ben directly with exactly what you documented — two posts, same result, timeline, screenshots
I've already emailed support@dev.to. No idea when they'll reply.🙃
Hey @xulingfeng
I want to let you know that the person behind the sloan messages were from me. Sloan doesn't detect articles automatically and is sent via human.
Based on the content you have published (mostly about stories which is fine), and how the content is written upon review, I believe it would be best practice to leave a disclaimer that you used AI to assist on your writing.
Let me know if you have any questions! (Usually @jess and @ben are busy and I am the only person using moderation to moderate the platform)! :D
My only question is — why are my two latest articles completely invisible in the feed and search, regardless of whether I added a disclosure or not? They can only be accessed through a direct link, and only my followers can see them. Why?
I just need an explanation. If there's something wrong with my articles, tell me what it is and I'll fix it. But right now, nobody is telling me why.
Timeline of what happened:
On Jun 15, I published a new article (dev.to/xulingfeng/i-spent-3-months...) with an AI-assisted disclosure at the end. Some of my earlier posts had already been flagged by Sloan, so I figured I'd just be upfront this time. Result: it never appeared in the feed. Not in search either. But my followers could still like and comment on it.
On Jun 16, I published my latest one (dev.to/xulingfeng/a-company-ai-fla...) — same thing. I deleted it and republished without the disclosure. It showed up in the feed for about a minute, then disappeared again. Now both articles are only visible to my followers. Non-followers can't see them at all.🙃
Issue resolved based on thread: dev.to/francistrdev/comment/39hdd
I'm not your follower, I can see those articles.
You're making mistakes somewhere.
You can only find my articles through the link I gave you — they don't show up in the feed or search results.
@francistrdev Now I'm even more confused than before!
This is my original comment that @dannwaneri references in his post.
Could you clarify, please? Do I have to mark all my posts with a disclaimer? That's not a problem for me. I just want to know. It seems a bit absurd, but I'll do it if it's mandatory to prevent my posts from going unread. I joined this community with the intention of sharing my experience with young developers and discussing with experienced developers like myself. Not that my way of building things is the best, it's my proven method after 40 years, and maybe someone might be interested in trying it. Why do I use AI to help me refine my posts? Because sometimes a simple translation loses the real meaning of what I'm trying to express in Italian, and I don't want to run the risk of appearing pedantic or arrogant. AI helps me with this because it understands the nuances of technical and spoken language. I believe that, as I wrote in the comment, a post should be judged based on its value (not so much technical), but rather its human value and its contribution to the community. Thanks for your attention.
Marco
I will add another consideration if it can be helpful.
Why not to decide a default predefined banner by DevTo to insert in the post so the bot checker can ignore it?
So it leave the decision to judge to the readers. They know it the post is done with the assistance of AI and and can judge by themself.
Another situation is for posts entirely written as spam by bots.
I add this a separate comment because is another topic to consider for me.
Thank you.
That's actually a cleaner solution than the current text disclaimer. A platform-native badge that signals AI assistance without requiring the author to write their own disclosure — standardised, consistent, machine-readable and visible to readers without being buried at the bottom of the post. @fm 's "detect quality not AI" and your banner idea are pointing at the same thing: structural signals rather than text enforcement. Worth putting directly in Francis's post as a suggestion since Ben and Jess are watching that thread.
Marco your situation is precisely the one that exposes the policy's weakest point. Using AI to bridge a language barrier — to make sure your 40 years of experience reads the way you intend it to in a language that isn't your first - is not what anyone meant by "AI-generated content."
But the disclosure requirement doesn't distinguish between those cases. For now: yes, add a one-line disclaimer. It's a low-cost way to protect your visibility until the guidelines catch up to the nuance.
The structure of this story is almost too perfect: an essay arguing that AI detectors are unreliable, written by a human, gets flagged by an AI detector.
That's not just irony. It's an empirical demonstration embedded in the narrative itself. The essay's thesis became its own evidence.
What strikes me is the asymmetry in the incentive structure. From the detector's side, a false positive costs almost nothing — the author gets inconvenienced, adds a disclaimer, moves on. But from the author's side, it erodes trust in the platform: if structured, precise writing reliably trips the detector, then improving your writing quality becomes a liability. The detector optimizes against the exact thing it should be encouraging.
The part about the DEV.to founder liking the piece before Sloan flagged it in the same hour — that detail deserves more attention. The human reviewer and the automated system came to opposite conclusions about the same artifact within the same hour. One of them was pattern-matching; one was actually reading.
The question I'd add: what do platforms do when their quality signals start to conflict? Reaction counts, human editorial attention, and automated detection can all point different directions on the same post. At some point a platform has to decide which signal it trusts more to define 'quality.'
The asymmetry point is the one that doesn't get named enough. A false positive costs the detector nothing and costs the author trust in the platform. That's not a calibration problem . it's a structural problem. The detector has no feedback loop for being wrong, so it has no reason to get better at the margin where human writing and AI writing have converged.
"Which signal does the platform trust more" is exactly the question the new moderation guidelines will have to answer explicitly. Right now all three signals — reactions, editorial attention, automated detection are running in parallel with no declared hierarchy. The same post can score well on two and fail on one, and nobody's said which one wins. That's not a policy gap. That's the policy question.