I want to talk about what I have notice on the platform and want to address what we all have been thinking about, but not really talked as much on ...
For further actions, you may consider blocking this person and/or reporting abuse
OMG, YES! I really do not understand this trend for generating comments... and I see a lot of this, especially here on DEV. I was always wondering about the reason for this, and could only think of not being a native and having some non-native complexes. As much as I can understand this (I always have a feeling I cannot express as much as I'd like in English), I'd still prefer to read a comment with typos and non-perfect grammar, rather than a repetitive AI-generated reply. We all have different styles, different vocabulary, and ways of expression; it's such a pity to lose it for AI.
RIGHT?! I don't know maybe the dead internet theory is actually true and the web is crawling with AI comment generating bots instead. But trust me, people know who is genuine and who is not, at least on my perspective. If you scroll down 5-10 articles, the rest are AI puke (and it is painfully obvious nowadays too).
Yeah, and instead of exchanging our own thoughts, we are drowning in em-dashes, verbose prose, and comments written with words no one would ever use in real life. I even caught myself appreciating typos in text and thinking, "oooh, this person didn't use AI" 😂
em-dashes also occur in google translate and others, so this is not an indicator
Em dashes aren't usual the indicator by its own, but the tone and structure of the comment gives it away. You could easily hide the structure, but people aren't going to spend time with that.
For @klaudiagrz, I have no idea why people AI generate comments either and that really do need to stop in my opinion. I can understand professionalism, but there are MANY ways to do that. I have a watch dog list of people who not only AI generate comments, but compare to the vast amount of comments they did, they are all VERY similar in terms of not only structure, but the theme such as promotion and such.
For @codingwithjiro, can't really do anything about it other than put it to low quality. Even that, I am outnumbered unless we all effortlessly push quality content that drag those articles down to the shadow realm.
Glad you guys are having a good thread here lol.
You know, there's been a recent trend of spam emails mentioning dev.to, like "I saw your profile." It's weird, but funny. The goal is to promote some AI crap. But it has something in common with AI comments on dev.to. Maybe it's (comments) a ploy to get to the top of dev.to rankings and attract the attention of popular sites (so your spam gets more views)? btw this is a good topic for an article. what do you think?
Thanks for the shout out @francistrdev! Been doing everything I can to prevent people from joining the "dark side" of just AI-prompting and blindly clicking "Accept all changes". Then ending up not batting an eye on the 500+ lines of AI code changes (because who in their sane mind would). Banger of a post as always!
Thanks Elmar and I am glad you are learning in this age since people can fall easily on using AI too much, but we also want to make sure we can utilize the tool that helps us while also still keeping us in check of our learning!
Lowkey, I was never into writing, so I had a friend run my blog. He wrote everything, hit publish, and I'm pretty sure it was all AI-generated. Looking back, it just had no soul.
Then I came across your articles... and now I actually want to write my own blog. Never thought I'd say that because I've literally never written one myself. Thanks for making me want to give it a real shot.
Thanks Divya. Writing is important, and it doesn't need to be professional at all! Everyone has different writing skills and everyone can connect to each other based on writing alone! I would prefer seeing many different writing styles than one monotonic style because it shows me that someone out there exist.
My hope for you, and for everyone on dev.to, is to be able to find someone they can connect with. I found @codingwithjiro based on my journey on dev.to and Virtual Coffee. Engagements comes alone naturally and there isn't a need to get "popular" right away. The best influencers, comes from the people who makes a positive difference.
Looking forward to it :D
Elmar seems like a really nice person. I think I've read a few of their articles, and I really liked them. I'm also happy that I got to meet both of you legends. Thanks! :D
We are grateful! Hopefully we will see you in Virtual Coffee!
Maybe I should join it tomorrow.
I'd like to join, but could you tell me who from dev.to is there? Please don't take it the wrong way. it's just My habit to ask these kind of questions.
I actually relate to this quite a bit. I've done something similar with AI, even for comments. Hell, I built a personal tool around it that I would've published by now if I could 😅
Personally, I don't really see AI helping with writing as a problem as long as the thought behind it is still yours. I'm pretty introverted by nature and communicating my thoughts clearly hasn't always come naturally to me, so one thing AI has genuinely helped me with is seeing how I could have expressed something better.
Over time, I've noticed some of those patterns actually becoming part of how I communicate without AI too.
That's why I really like your "quiz me then organize coherently" approach. The AI isn't inventing what you think; it's pulling the thoughts out of you and helping structure them.
I think the interesting question is where that line sits between AI expressing your thoughts better and AI gradually becoming the source of the thoughts themselves. As long as I can still explain and defend what I put my name on without needing AI beside me, I'm personally pretty comfortable using it.
Also, training custom skills around your own writing style sounds interesting. Has doing that changed the way you write or communicate without AI over time?
Hey Aryan! Thanks for sharing and appreciate you talking the time!
There is a couple of issues. How would you determine if the idea is "yours"? Is the idea overall came from your creativity or is it something you recognized? For example, I can tell AI to output me a summary on React. If the summary is something I already knew and changed up a bit, is it still mine or the AI?
Another thing is the expression. Yes, I am an introvert as well. It's hard for me to determine if someone takes a joke or not and sometimes it's difficult for me to put myself out there. However, I learned overtime based on my experiences with other people and learn from those experiences for my own. Sure, you are welcome to use AI to help you express in the right way, but I am wondering if you are learning how AI has helped you and see if you can do it on your own without AI. It's like riding a bike. At first, maybe you need a third wheel, but overtime at some point, you would not need it. I would follow this because it allows us to help us learn instead of relying. Hope this makes sense.
Can you elaborate further on what you mean by "custom skills around your own writing style"?
Thanks Aryan!
Thanks for the thoughtful reply, Francis! 😄 You brought up a couple of really good questions.
For me, the dividing line is whether I could have arrived at the same conclusion if you took AI away. AI can help me structure, refine, or even challenge my thinking, but it shouldn't be the source of the core thought. If I ask it to summarize React, that's not really my idea. But if I've been working with React, formed my own opinions from experience, and use AI to organize or polish those ideas, I still consider the ownership mine.
I also really like your bike analogy because that's exactly how I see AI. The goal isn't to keep the training wheels forever; it's to learn from them. Whenever AI rewrites something I've written, I try to understand why it reads better instead of blindly accepting it. Over time I've noticed myself naturally writing that way without asking AI first, so in that sense it has genuinely improved my communication skills.
I think that's the balance I'm personally aiming for: AI should gradually make me less dependent on it, not more. If a year from now I still need AI to express every thought, I'd consider that a sign that I relied on it too much rather than learned from it.
As for the custom skills part, I actually wasn't referring to my own workflow there. I was curious about @anchildress1's setup because "custom skills" sounded like she'd built a personalized set of instructions around her own writing style. Her follow-up clarified it nicely, using an outline, having Claude interview her, and continuously updating it with her own vocabulary to reduce drift. I thought that was a pretty interesting approach.
Thanks @itsugo. I would probably go insane otherwise, because I can't let go of that much control of my writing most days. It definitely helps write faster, because I can follow a thread of thought through to the end and eliminate the scenic route along the way. 😆
@francistrdev You've given me an idea for a whole other post to dive into my workflow a bit more, which I hope to have ready before long. Mind if I link to yours there?
Yea for sure! Make sure to mention me since I will get notification. Sometimes it's 50/50 when notification pops up when someone makes a post. Looking forward to it :D
Love this post! I never use AI for comments, but I regularly use it for writing — anywhere from "write it for me" to "just finish it already". But I also invested heavily in writing custom skills so AI could emulate my own writing style. Imperfect, but better than generating nonsense at scale. The most helpful for me is the ADHD-style "quiz me then organize coherently" instruction I set up for Claude. I don't always write it myself anymore, but if my name is on it then its always my idea.
Hey Ashley. Thanks for sharing. I am curious on "invested heavily in writing custom skills so AI could emulate my own writing style." How often do you check yourself to see if you have a different writing style compare to what you tell the AI to do?
That's usually tricky because I can tell the AI that "I did this" that follows up with a topic I do not know. It's the same way as saying "I have ideas, but I don't want to write it. I will let my friend write it". Sure, it is YOUR idea, but it can also adds on to the idea that the friend has. What's interesting is that depending on your usage (telling the AI to refine over and over again), it tends to drift away from the idea you initially had, which leads to the idea you are not familiar with nor it's no longer the idea you have in mind.
Hope this makes sense. I appreciate you sharing your thoughts about this :D
I'm consistently updating that particular writing skill with words and phrases that are mine, especially. Also, I always start with an outline and in the case that AI is writing more than me, then I'll let it compile a list of questions to prompt me with as we go. Then it cleans up my answers and rewords some of it to make sense. If any refinement is done afterwards, then its done against that original outline to minimize drift. This is one of my much earlier posts on the topic that AI wrote for me using a much less refined version of that skill, which does tend to sand off the edges over time. Compared to not having one at all, though, it makes writing with AI much easier to deal with.
I agree with this, but it does come at a sacrifice of not sounding "professional". Like there's a difference to explaining something to your colleagues at work and then writing an article people are suppossed to derive value from. I think it's easy to get in your head about it. I just try to write anyway all the while balancing it with work life.
Don't sound like a robot guys.
i agree with you what you said on here.
It is why we practice. Yes, AI allows us to say the right things but it's good to note that as long as you are learning from why AI outputs the way it is that you want, you can start learning from that and apply to the real world. I learn how to speak based on how others speak in different settings without using AI. I do appreciate you sharing Daniel! Hope you are doing well on your YT channel :D
Yeah, the channel is improving, and I hope you have been well yourself.
I call things bad not because they are objectively bad, but because they cease to be interesting. A favorite game becomes boring, WinRar gets annoying with its 40-day period.
My situation is completely the opposite - CAN THIS DUDE EXPLAIN TO THE AI WHAT IT WANTS?! I think I fall into this category. I've seen a lot of memes about people politely talking to AI, but I'm the complete opposite - I constantly insult it, hoping it will somehow understand my meaning better (it didn't help).
For me, the big problem is that the AI is so stupid. Or maybe I have an extra chromo... ahem...
Assessing product quality is difficult because quality isn't just about functionality. At least in mobile games, thanks to advertising and rapid development, content can be considered good because it generates a lot of profit with minimal effort 🫰 ..Of course, you need to invest a few thousand in marketing, but how can you do that? A business without investment isn't a business.
Hey Ember. I think at this point people can explain to the AI without trying. It only takes them 5 mins to the point where they get the idea based on their experience using it. Sometimes, I just do all caps by stating "DO NOT CHANGE THE FUNCTIONALITY. KEEP AS IS" lol.
I found to be unforunate, especially in Y Combinator Startups where they get funding because their project revolves around AI. I have heard about this a lot and it does drift away about quality and more about hype. A video you may or may not seen before capture this:
Sometimes I don't realize content is good until I find myself sharing it with someone else.
i dont know what is going on my life is in stuck.
I mean people will have different opinions and perspective of a good article. I do appreciate you sharing though :)
Simple one rule for good content Write to express, not to impress.
Good simple rule! Without expression in writing, it's just simply informative (which is fine) but if you have that as a common writing practice on a blog site, it sets boring. Will love to see more blogs that are expressive in some way for human connection, especially for developers!
Indeed. From a technical writing perspective, the 7 Cs of effective communication (Clear, Concise, Concrete, Correct, Coherent, Complete, and Courteous) naturally encourage the use of simple, lightweight words, but they are only one part of producing high-quality content.
It's a good reminder to know your C's for me at least. I tend to just "ball" and go from there (which it's not efficient in writing, but I ensure to get my points across). However, I do love simplicity in a sense and I feel like if we were to express in a simple way AND being able to get our points across in a professional manner, I believe we would be a more welcoming community in all skill levels. Yes, it's important to know technical terms, but no need to over do it. A good balance is key.
Generally, content can't be categorized as good or bad, at least in my case. I'd say it's useful or interesting. There's a lot of good content here that falls outside my preferences, so... what does it matter?
Now for the controversial part: how long do you think it will be before publications or code written by humans are considered low quality? At the rate AI is advancing, I'd say not long. In any case, it's something we'll know soon.
I would say it matters, a LOT.
It's been reported that people are easily fall to misinformation, even before AI. With AI, it makes it worse because it's getting hard to differentiate between what is real, fake, AND true. This post's goal is to apply critical thinking in order for us to see what is actually good content. Sure, people have difference preferences to "good" content, but applying this knowledge will at least help instead of following the herd.
For the controversial part, who decides what is low quality is the question. AI has becoming unpopular based on people looking for jobs and the general public as well. I would say it would consider once we BELIEVE it is low quality but realize we are pointing to the wrong person. Hence, going back to my point, critical thinking!
Thanks for sharing Fran :D
You're confusing disinformation with AI-generated content, or you're establishing a false equivalence; AI-generated content isn't necessarily disinformation.
Disinformation can be spread without using AI, through the traditional methods we seem to have become accustomed to.
To be honest, the problem with AI isn't the quality, but the sheer quantity it generates, regardless of its quality. It's true that much of that content is irrelevant or redundant.
My stance on AI is pragmatic; I wish it were different, but I'm afraid it's inevitable, so...
Hey Fran. I would disagree to a degree (no pun intended). It can be both misinformative and disinformative.
It's been documented that content creators use AI to prank on other people. An example here is shown that this person use AI to generate a video that is fake, but looks real. youtube.com/watch?v=WmFu15JAazI
Other cases is if people are really smart, they can use AI to the max to scam other people. It will not just be the elderly people that get scammed, but people like us even though we may THINK we are safe.
It is problematic. Yes, disinformation can be spread without AI. BUT with AI, it has become almost too easy. Yes, maybe earlier times of AI is all fun and games but after seeing the improvements, I am not sure about that now.
I agree that we disagree.
Fair enough. Though I appreciate you sharing your thoughts again and keeping it formal!
Typing effort alone isn't the measure of good software. 🧐.
In threads like this, once I admit I rely heavily on AI for syntax, my guess is 99% of senior developers would take me for a fugazee.
That won't stop me, though. Every generation has used the best tools available. I wrote about that recently. I know how much time I've spent struggling with syntax, and I've accepted that it simply isn't where my strengths lie.
I have a great deal of respect for the senior developers here. Without people like them, AI assistants wouldn't even exist. In a way, they built the greatest junior developer ever created.
I'm not against learning syntax. AI isn't perfect; it misses things I know are possible, which pushes me to write better prompts and get the syntax right.
Just as society needs mechanics, artists, doctors, farmers, and countless other professions, software can also have people whose strength is solving problems rather than remembering every language rule.
I know we all want the best for the community and the future of software—but who really wants the worst? There should be room for people who enjoy mastering syntax and for people who use AI as a tool to build.
If everyone is expected to rely solely on syntax, then what exactly are AI coding assistants for?
Sorry I had to delete the comment for clarification. On my notification, I saw "Typing effort alone isn't the measure of good software. 🧐.". My fault lol.
Now seeing this, it seems you are confusing Syntax in writing. Using AI to correct Syntax makes sense and I do encourage using AI to clean up the codebase in a way as long as you initially built it yourself.
Syntax, in the way I see it, is just grammar. Nothing wrong with correcting grammar using AI since it tends to be the most common and more effective. The problem because it we go down the rabbit hole of using AI beyond the syntax. Maybe at first, you are fixing Syntax issues, then AI starts to ask you if they would love to "refractor the code". We have to be careful on how much we use AI because of those followups. If we fall into this spiral, we loose that ownership and we lose on what we have created so far. If anything, it will lead us to a forest we cannot recognize anymore.
Overall, Syntax in coding and "Syntax" in writing is different. If anything, it's best to learn the syntax rules at a MINIMUM. Learning is the best thing of anyone. I want to learn in the event I asked myself "What happens if AI doesn't exist? Can I still do what I do best?".
Obviously, people will judge you and can't you blame them? AI has been increasingly becoming unpopular and it's been reported that companies are back pedaling AI usage based on how much developers are using that cost more than a developer salary, which is expected but crazy. Not everyone acts like this though. I added a transparent note in one of my articles that I used AI to enhance the image quality and no one complain. If people are judging you about your use of AI, I would question further of how that became to be in the first place.
Another thing to point out, great syntax doesn't mean good software. Good software provides not only usefulness, but being able to know how it is built in the first place. If you Vibe Code a project and call it good software, I would back pedal that on many factors when considering what is "good" software. Is it Secure? Is my Data Safe? Is it accessible? If something breaks, will I be the one to fix it? Is it following best practices in terms of written code? There is so much variables that AI can't really cover and it is something I learned overtime.
Hope this makes sense! Let me know if anything.
When I mentioned syntax I wasn't referring to article writing as much as java() HTML</> & @media_css lol 🤣
Oh no lol I got it. I saw your edits and apologies for that. Again, I only saw the initial comment and not the edited version. This is what I saw at first lol:
Now rereading this thread, I confused myself. My bad again D:
The reply still stands minus the part where I thought you got confused on Syntax in Writing. TL;DR Use AI as a tool, not relying to the point where you asked the AI "How to write strings in python" or a similar sense if anything.
Also also, forgot to answer your burning question:
AI Coding Assistants is mainly to help you navigate large codebases, which is great if you are contributing to Open Source because it tells you what parts of the feature is located. It will save manual work.
I tend to use AI Assistants on that and create a simple template based on what I am working on. For example, I use AI assist to create an example Unit test based on the codebase and what it's built upon. That way, it's easy to refer to and captures all the cases I need to build my own without having to rely on AI as much. In this case, I am still learning since I only use AI to help me navigate the codebase to find the files while also not relying to the point where it doesn't give me the "complete answer" but gives me the idea of how it is built.
Sorry for the longest thread. I'm tired and my brain is fried lol. Hopefully, this answers everything!
I agree, I think it's really important to question and understand what AI has written. Many times I'll ask someone a follow up question to their solution, and they'll respond with "AI told me".
I think AI seeming so confident in its answers is causing people to believe what it says without questioning it. With pressure from businesses to use AI for everything, and now being accustomed to the faster speeds at which we get work done, how do we take a step back and research the problem ourselves before consulting AI?
Hey Shayan! Thanks for the question!
Whenever I use AI, I make sure I understand what is being output to me such as "Do I know this specific topic already"? If so, no need to go further into it other than writing it in your own words. It is equivalent to reading something and paraphrasing it in your own words IF you know the topic. Either way, you will have to cite because it came from the source.
If an AI output something that we do not know, we learn. We don't CLAIM that we know this. We apply the knowledge in some way where we can fully grasp on the understanding. Even better, search it up the old fasion way because AI tends to hallucinate sometimes.
This gets tricky because it's based on you. No one is going to tell you this or that and make you do it. You need to have the will power to take initiative and have discipline to apply this. It's always hard, but so id the job market and people are still landing jobs either way. It's only a matter of if you care for yourself to learn and grow instead of following others. Thanks Shayan!
Hi Francis, thank you for your reply! Those are good points. It's tempting to use AI for everything, but it should be used as a research tool that we learn from.
I typically use AI in development by asking it to do the initial research on how things work in the app, and I'll have my terminal on the side going through the code at the same time. It helps me understand better, and catch hallucinations the AI has made.
Great article, I think it's important for people to know that AI is not a replacement for learning.
For sure and thanks for taking the time to read and asking good questions! :D
I usually judge content by whether it comes from experience. I'd rather read about one real problem someone solved than ten generic tips.
It is a good start on finding good content when it comes to experience. Thanks for sharing :D
YES YES YES 100% Agree with your article. I dont hate use of AI, in fact i use it daily to debug something. but we should've been use AI as a tool, not treat it as a person to do things.
I want to quotes tony stark from Spiderman Homecoming movie where he wants to take back the suit from peter. He said something like this "If you're nothing without the suit, then you don't deserve the suit", and the suit here is AI.
AI is great, it helps you do things, but you are the brain. you need to understand what you build / make first!
OMG I forgot about that quote from Homecoming!! I should have added that to my article since it made SOO much sense. I am glad you brought this up! We should treat AI as a tool and not as something to rely on, which I defentely agree with you. Thanks Khairunnisass!
I actually see it a different way. I've learnt so much from AI. It has allowed me to focus on research, architecture design, product design and marketing - taking away the time consuming manual coding and research.
Before AI I used generators, scaffolding, and a documentation platform. I still use all of those - it just has AI integrate into it. So if you have a good framework in place, AI is the glue.
You absolutely need to be the architect, understand what you're building or generating. AI should be the "tool".
AI is very bad at generating blog posts - no number of skills and examples makes it generate something even close to interesting like this blog post. But you can research your blog post ideas and content and verify your claims.
I think the big difference is that software is for machines. It's literally what machines run on. Blog posts are for humans. So it makes sense that AI should code for you but you should write your own blog posts.
I do see that Gabe. I know some people on the platform use AI to fix grammatical parts of the blog post. I believe that method is a good use of the tool, though I don't use that for this case for myself since I want to learn to write and speak. The problem is that some people don't see it that way and use AI to mass produce articles (more or less) to get popularity, which makes me disappointing to see.
Sure, but some people don't see it that way, which is interesting. I am wondering on @klaudiagrz take on this because it seems people just simply don't care. Maybe that's me or people just have given up. I do and definitely agree that AI should be use as a tool, but it's only a matter of us. Are we discipline enough where we can handle of our use of AI.
Thanks for sharing Gabe!
Thanks for the call-up, Francis! You know, I've seen an interesting tweet today:
I agree 💯! It's about documentation, but it applies to blog posts even more. I sometimes also run AI over my text to correct obvious typos and stylistic mistakes. However, I always make sure it doesn't interfere with my writing style. Because, in the end, this makes the article authentic.
Human brains do not comprehend well what LLMs generate – the generated content is very often too verbose, sentences too long, and vocabulary too non-natural, which makes the content hard to focus on.
In the end, you should always have in mind a person on the other side who will read the content you publish. Of course, you can quickly generate a slop that will look like real documentation or a post. The question is: will it be helpful to the person who wants to read and learn from it?
This line hit hard: 'Can they explain, in detail and in real time, what they did and why?' 🎯
As a developer, I’ve felt that exact trap—it’s so easy to let AI write a perfect solution or polish a post, but the moment you can’t defend or explain your own work, the learning stops.
I really like your point about asking questions that expand rather than just recite. It shifts the comment section from an echo chamber into a real learning space. Thanks for writing this reality check for the community!
No problem. Glad it was helpful Hosein :)
Just want to say thank you for this piece! I've done professional editing in academia for many years, and that last sentence in your article pretty much expresses how I also feel about the topic on AI content generation. As for whether the content is good or bad, I tend to view tech writing from an editor's point of view still. If the text has a logical flow, reads easily with ideas clearly expressed in well-constructed paragraphs, and the supporting evidence is solid, that pretty much ticks most of my boxes, regardless how that content is generated. Some of us may be super knowledgeable about a topic, but we might struggle getting those ideas out in written form, and so the assistance that AI affords is truly valuable for them. The trade off, of course, is that the personal touch could be missing, and so I genuinely appreciate articles like this one where the author's voice is at the forefront and unmistakably unique. Whether writing conventions (like the em dash) still matter as much would depend on who actually cares about it, but for me, if the target platform isn't a publishing house with strict style guidelines and if I as a reader can fully understand your argument already despite the occasional typo, then that's more than good enough for me.
Hey Elaine!
It would be good enough for me as well. It's just seeing a lot of articles on dev.to that sounds monotonic, which questions a lot of people if the person is even real. With that said, as a result, AI has becoming unpopular in a way. As a Mod, I see comments that they despise people using AI because of the way it's structured, and tone of the article read. My hope on dev.to is people applying critical thinking whenever reading articles and being mindful of how to use AI the right way. Thanks for the share Elaine!
Building a translation pipeline for Korean/Japanese/English DOCX files, the biggest surprise was that most LLMs drop merged cells and table nesting even when the text is perfect. We ended up parsing OOXML directly with python-docx and reconstructing the structure rather than trusting any library's high-level API. For terminology consistency we cache a per-project glossary and do a post-pass substitution rather than forcing the LLM to hold it in context.
?
I agree with your take. AI-generated text often needs careful fact-checking, since models can confidently hallucinate details — wrong prices, made-up features, non-existent APIs, you name it. If you don't verify, you risk publishing misinformation without even realizing it.
That's why I think the better approach is feeding the model your own accurate content/context first, then letting it help with phrasing or structure, rather than asking it to generate facts from scratch. It shifts AI's role from "source of truth" to "editing assistant," which matches your point about still doing the thinking yourself.
It's a good strategy, though I would be mindful of what you are feeding into the model specifically. You have to ensure you are still learning while AI is helping you in some way that is not fully relying on it. Thanks for sharing Tomasz :D
One angle that's easy to miss is that content quality isn't measured at the moment it's published, it's measured by what readers can actually do after reading it. If a post helps someone make a better decision, solve a real problem, or understand a complex topic more clearly, that's a far stronger indicator of quality than writing style alone.
Right, but at the end of the day, that all depends on the reader itself.
Let's say there is a AI-generated article. Reader 1 reads it, then applies the knowledge. Reader 2 reads it, but does not.
It is soly based on the reader's knowledge and expertise that decides on if they found it helpful. The point of the post is to ensure the reader not only understand the topic, but encourage to ask questions to the author.
Writing style plays a factor because the goal is to explain something as simple as possible. If I were to explain "How to use React" using words 90% do not know and the style is messy, the reader will not be engage in the topic. However, having it as simple as possible while also having a casual style helps the reader to be more engage, knowing how to use React, and being able to ask questions after the fact. Hope this makes sense and appreciate the thoughts Glen!
I care less about whether a draft touched AI and more about whether the author can answer the second question. If they can explain the tradeoff, name what they checked, and say where they are unsure, there is a person behind the piece. If they can't, the polish is mostly camouflage.
Agree and kind of disagree. Yes, it is important that they can be able to explain what you mention which is the "tradeoff, name what they checked, and say where they are unsure". However, it is important to not use AI too much to the point where it doesn't sound like your voice at all.
For example, you can explain React in MANY different ways. There isn't one "true" way to expressing what React is. Some people explain in simple terms while others explain in professional terms. Either way, the way you write is the way you express.
I agree, which is why I encourage for everyone, including myself, to put more thought in comments and in post especially. I think of it as if someone gave me a piece of work, I start to question not only to challenge the author, but to also be able to learn about what they have written. What makes content good is if there is meaningful engagement between both parties, not just an author getting 1 million likes with comments that is the synonym of "Good work".
Thanks for sharing Reid!
I think the focus shouldn’t only be on whether AI was involved, but on the value, originality, and personal experience behind the content. AI can help with ideas, structure, and productivity, but the human perspective is what makes an article worth reading.
I would agree and kind of disagree. Yes, knowing who is behind the content helps but we also need to know how much AI they are using and to what degree. If it's too much, it doesn't matter if it is their idea or not because we don't know for sure. Thanks Vikash!
Hey @francistrdev Nice article. I agree with your first point to "Ask questions that expand," and this is one of the areas where I struggle, especially while growing my personal brand on X, LinkedIn, and this platform, Dev.to. I noticed that I don't ask meaningful questions that expand the conversation and tend to spam comments on posts with "Nice share," "Amazing read," and so on and so forth. Asking good questions is an art that comes from practice, thinking, and reading. Although I read a lot, I never practice thinking.
One point where I partially agree with you is that an individual who is just at the learning phase should avoid using AI as much as possible. It depends on where an individual is in their learning phase because some of them aren't beginners, and as long as they know what AI is doing, then I don't see any problem with it. Even I use AI a lot in my workflow and while learning new stuff, and 80% of the time I don't write code by myself, but I know what AI is doing.
Hey Tanay. As long as you understand what the AI is giving you and being able to explain without the help of AI, then I am all ears lol.
There is nothing wrong with saying "Good work" and such, but like I mention, that should be an occasion. Having the habit to ask questions and apply thinking is good practice, especially in this age where we blindly believe in almost everything without having to stop and question for ourselves. Thanks Tanay!
As a student who has started to rely on AI for too many things, this blog has given me a change of perspective. Instead of just producing AI-generated content mindlessly for views, I think I should focus on having a purpose for each project and actually understanding my work. Btw, I appreciate your creativity with the GIFs; the JJK ones really made this blog fun to read!
Thanks Kulsum! Glad you found this post helpful :D
AI is a useful assistant, but it shouldn't replace the author's thinking. Using it for grammar, translation, or brainstorming makes sense because the ideas are still yours. The problem starts when entire articles are generated with little or no human input. Readers come to dev.to for real experiences, lessons learned, and honest opinions. AI can help polish the message, but authenticity is what keeps a technical community valuable. Moderation that encourages transparency and quality over quantity would be a positive step.
Exactly. It should be use as a tool. It's unfortunate we see a lot of content being sound really monotonic and structured in a way that doesn't sound unique. I always want to stand out on dev.to, so that's why I have a writing style that sounds casual, but straightforward for others to understand. If not, always good to ask to clarify! Thanks Edward :D
I saw a comment that reads "Write to express, not to impress;" and this hits the exact core of this debate.
When AI handles the "impressing" - polishing grammar, structuring bullet points, and summarizing specs—it strips away the personal friction that makes human writing relatable. A developer sharing raw, messy learnings from a real debugging session will always offer more value than a flawless, surface-level summary generated in seconds.
Your call to ask expanding questions rather than just repeating key points is the exact mindset shift dev.to needs as AI tools become ubiquitous. Great reflection!
You can use AI to smooth out prose or optimize code snippets, but if you can't walk someone through your architectural choices off the cuff, did you actually learn anything?
Pair that with "Write to express, not to impress" and the standard for good content becomes obvious: good content communicates real-world experience and clear problem-solving, not just a clean surface. Using AI as a learning scaffold makes sense, but substituting it for actual understanding leaves you stranded when real technical discussions happen.
Thanks! Glad you found the comment that @vmodal_ai wrote! Good content always come from one's experience because it allows us to enforce their POV and seeing from their perspective. Better than saying one point without backing up in a meaningful way.
AI itself isn’t the issue, it’s how we use it. Using AI for support is helpful, but original thinking, experience, and personal insights are what make content truly valuable.
Hey Allen. I do agree. AI itself is not the issue. It's just how we use it in a way where we start to rely on it. I hope this article will help others to apply more critical thinking than just relying on AI. Thanks Allen!
The part I keep running into is that "good" isn't one axis. A piece can be factually clean but tonally off, or engaging but subtly wrong, and a single quality score collapses all of that into one number that hides the disagreement. I've had better luck scoring a few named dimensions separately and only then combining them, because it makes the failure legible. How are you handling cases where two reviewers, human or model, split on the same piece?
Thanks Karitk and great question!
I would love to have some clarification, but based on my understanding from your comment, here is what I think:
I do agree that it's not in a single axis. Everyone has different preferences on what makes the content "good" such as your current skill level, understanding of the posts, etc. A person who is a professional will see a professional article as Good while the beginner person will see it as "not good" because it is simply to difficult to understand.
When it comes to AI writing, the way I see it is repetition in terms of tone. You can write any topic as you like with AI, but the key difference isn't em dashes, its the tone and how it is structured.
I have work with Claude in my development and I notice that it tends to have this "AI tone" to it when I read it. Additionally, it has the AI structure. Sometimes I always believe that "Maybe that's their voice" but after investigating further, it's usually never the case.
I see people on dev.to use AI in a good way because of Grammar fixes and such. In this way, it doesn't remove the "tone" of the original author and it is still their work unless the reiterate 100 times to the point it's no longer their voice. It's just AI-generated text with extra steps, which is why we need to be careful of how we use AI to not rely.
I hope this makes sense. Do let me know if there are pieces that I am missing or parts that confused you! Thanks again Kartik for the question :D
Working with AI is a daily thing for me, tbh I use it everyday across a ton different use cases. But I keep pretty tight guardrails so I don't lose my own voice. I make sure to check myself, is this still my vision, or did the AI's recommendation start influencing me to do something else? The second it starts drifting from the core focus of what I’m trying to build, I pull it back. Real talk, staying honest with yourself about that is harder than it sounds when you build lol
Hey! Thanks for sharing!
It is hard. It's meant to be because it requires discipline, but I am glad that you acknowledge it and take action to ensure you are using AI but still learning!
I agree and am both a victim and enemy of blindly using AI. But also I don't mind AI as long as the content comes from real experience and original thought. Typically, what I do is write a really bad error-ridden giant block of text or notes and then prompt AI to just make it look better lol. Anyway, nice read!
Thanks Dallas and appreciate the share! One thing to note is that original thought can be tricky since we don't know. This is why we question things and apply critical thinking to see if it IS original or not. Regardless, thanks :D
Great post. This really resonates with me.
I work on tactile data standardization for embodied AI — essentially building an open annotation format standard (like ROS, but for touch sensors). It's deeply technical work where every design decision has to be defensible.
I use AI daily: for grammar fixes, structuring docs. But when someone asks me why we defined 14 semantic dimensions instead of 18, or why a specific sensor adapter works a certain way — AI can't answer that. I have to.
For me, the litmus test is simple: if someone challenged me on every claim in my article at a conference Q&A, could I defend all of it? If the answer is yes, the AI was a tool. If not, it was a crutch.
The tricky part with standards work is that "good content" isn't just about clarity — it's about whether the spec actually works when someone implements it. You can't vibe-code a protocol. The community either adopts it or they don't, and that judgment is brutally honest.
Thanks for sharing Liesliy!
In today's world it's all ai generated from the projects to caption and comment. I started my journey with freeCodeCamp. And i still love that platform because it really feels like you're learning something, making efforts as you have to write every single word by typing. But now I feel like if I try to continue to learn from there, i will fall behind. I feel like if I try to understand everything of my project which i want to do, others will make 10 projects by that time by vibe coding.
I get what you mean. I use to go on freeCodeCamp and it's still a good resource. I know it's difficult seeing a lot of people are vibe coded and it feels intimidated on paper. However, it's a matter of "is there a story behind it?"
Sure, you can vibe code projects all you like. It's only a matter of purpose. For example, is there a reason why you build this project in the first place? If your answer is "just because" or something along the lines, than it's not a good answer, especially in the interview.
As a person who does interviews, I would love to see a project that has a story behind it regardless of if it is unique idea or not. For example, I built a project that detects AI generated image automatically than scans 100 images at a time. Does a project like that exist? I bet.
BUT, the reason is that my friends doesn't like AI images and they have to use a separate tab on their browser to manually drag and drop the image, which takes a lot of time. I built this project to save manual work and understand the importance of using AI the right way.
That's a good story! Overall, if you have a good reason and a good solution to why you built it, then I will hire you :D
Thanks for sharing Mahbuba!
Francis, I get what you mean. I think for me it comes down to whether you actually understand what you're putting out there. Using AI is one thing, but if someone asks you about your own article or project and you can't explain it, then yeah, that's a problem.
Also liked what you said about comments. Not every comment needs to be a deep discussion of course 😄 but sometimes one question can take the conversation somewhere completely different and you end up learning something you didn't even think about while writing the post.
Exactly. Thanks Hema!
how to put gif in it
Basically drag and drop the image on a comment or article!
Great insights here 💡 Definitely bookmarking this for later! 🔖
Appreciate it Elsie. Let me know your thoughts once you get around to it :D
Solid writeup. The practical rule I keep is least privilege on any tool an agent can call, especially write paths.
Thanks Edu! Good rule to have!
I think good.
k
the part about being able to explain your work is the real test. i use ai a lot, but if i cant explain why a choice was made, i probably skipped the learning part.
I would skip it too. If AI tells me something I don't know, I don't bother claiming that "I know this". Appreciate you sharing Julian :D
The interview version of this test conflates two different things: can you recite the mechanism, and did you make the calls that mattered. Those come apart constantly outside of AI too. An architect who never wrote the query layer can still read a diff and catch a bad join. A PM who never touched the auth code can still tell you why the session timeout is fifteen minutes and not five. The split between who owns a decision and who owns its implementation is old. AI just moved more of the implementer role onto a tool instead of a second person.
Where the live-explanation test earns its keep is one level up from where it usually gets aimed. Not "explain the mechanism," but "would you catch it if the mechanism were wrong." Someone who owns the decisions but has never actually checked whether the output matches them is functionally the same as someone who cannot explain anything, they just have a more convincing story for why that is fine. The tell is not fluency under pressure. It is whether there is a real feedback loop between what you intended and what shipped, something that would surface a mismatch before a stranger on a call does.
I think this is a question that existed even before AI times. What makes a content good? I guess everyone writes for different purpose in this or other platforms. For me it is to validate my own knowledge, i am a strong believer that if you can't explain something (simple) you dont understand it.
AI-generated or AI-assisted content can happen for many reasons:
So for me, the interesting question isn’t the Was AI used to write this? but rather the Was there a human thought worth writing about in the first place?
I think one distinction that should be made is that you are referring to Generative AI not AI, the two terms get intermingled but it causes confusion as to where the line is.
That said, in coding, its a very blurry line, someone who types "make be a todo app" ... and go! that ends up being poor quality, no understanding of the code.
vs
Someone who builds out a series of "tickets" and goes into a planning mode to build out what is actually to be built, spends time reviewing the code after, building MD files to ensure it follows standards and preferences the developer has, this ends up with good code, that the dev understands, because they spent the time with it.
They both get lumped in as AI slop, but really one is and one isn't. But how do you know which it is? There is definitely tell tale signs for sure.
When it comes to writing, while I do believe writings should be human written. I dont have any issue with putting it through Gen AI to point out faults in it (not for writing it).
One thing I do think is that people spend so much time trying to decide if something was done with Gen AI or not, that they forget about the content in it. If you are reading an article with good quality, accurate information about how indexes in databases work, should it matter if it was built using Gen AI? If you miss out on learning opportunities because you spent time wondering if it was built using Gen AI or not, you have wasted an opportunity.
There is good and bad in Gen AI, but I think it comes down to was the article easy to read, was it accurate and informative.
The irony of article writing is that there is so much to be learned by writing articles, its an opportunity to go and check your understandings, maybe learn a little know thing related to what you are writing about. Getting Gen AI to write an entire article on a subject, you gain nothing out of. I think that matters.
I think I should probably do a 'day in the life of' post at some point. I heavily use AI, not to be lazy, or not understand what I'm doing, I learn something new everyday and I can push out 10s of thousands of lines of code a day, while still being able to describe any aspect of it, not because I wrote it, but because I designed it and I validated it.
Take a restaurant kitchen as an example, head chef gives instructions 'make me a cheese omelet', chef goes and makes it. Both head chef and chef have read 'cheese omelet for dummies' and know how to make it. Chef brings back an omelet, Gordon Ramsey throws it across the kitchen, yelling 'you used the wrong cheese'... Sometimes, it looks right, feels right, but it takes experience to know it wont taste right. That experience cant come from creating, it comes from validating. Do you think Gordon Ramsey can produce every single dish his restaurants serve, off by memory? No. But he can look at it, taste it and immediately throw a tantrum, because it's not right and can tell you why it's not right and what needs to be done differently. That being said, he's not the best at it, food critics are, because their have finer tastes and spot the difference between a pinch of salt and a quarter pinch too much. The difference, is that Gordon Ramsey has far more experience reviewing, while the critics have far more experience validating.
You likely know the term dog-food? Generally, it's used to describe someone writing software for themselves as the primary customer. Now though, it's applicable to did you eat your own AI-generated code? If you want it used in production, did you use that code to validate it? It doesnt help it compiles, did you run it, test it, verify that the UI is stable and editing a field, edits the total? You're the critic of your AI, you're also the Gordon Ramsey that throws the dish when it's slop before even compiling it. You cant sit in the kitchen all day... Unfortunately, AI is the chef, you're the QC and QA pipeline at once. You are meant to validate it's formulas, you are meant to test it live to see if it works, because you are the one who ordered it and if it doesnt taste right to you, throw the plate and yell what's wrong with it, so the AI fixes it to match your order.
I fundamentally disagree with the notion that you need to write code, to understand code. I much more believe you have to review code, to understand code. Just how a chef can make a dish that looks and smells good, it doesnt mean it tastes good, AI generates code that looks good and compiles good, but does it run good? If you're going to delegate the work (and you should, because it's a force multiplier), you cant stay in the kitchen with them, because you wont have time to review and validate. That's why your job is to place an order, correct the outcome to fit the scope and test it yourself, so you know it's good.
If your mom made the nicest lasagna and you absolutely love it, that doesnt mean you can make it... But it does mean whenever you eat lasagna at a restaurant, you can tell the difference and while a restaurant isnt your place to tell the chef 'make it how my mom makes it', if it's your restaurant, you absolutely should... Though dont bring up your mom, give critique on the flavor and spice choice, till it tastes right. Same goes with AI, critique it's work, so it produces the kind of work you like. Whether that's yelling corrections, or typing out a step by step guide with an ingredients list, whatever produces the right flavor, because that's all that matters.
It doesnt matter how you wrote your app, infact 99.99999% of users really dont care. All they care about is that the app does what it's meant to, the way they expect it to. That's what I've learned writing production code for a general codebase and making custom modules for specific clients. Not everyone's tastebuds are the same, they dont care how you make it, they just care how they'd break it and that you've engineered it to make sure it doesnt.
As for writing, I guess it comes down to what you're trying to convey? Take my series on AI concepts, I could have spent 2h writing articles for each, or as I did, I had the AI write it, based on what we built together. It understands the order well enough to create it, but it's my job to validate it. If it's formatting is wrong, I dont go and manually replace ** with **, I tell it '* should have been *** so it's more pronounced', that way it learns (atleast mine does) and stores it to the memory, which means next time I order an article, it writes it to fit my palate better. It never wrote it wrong, it just didnt write it the way I like it and your palate is different, you'd likely want it a different way.
Personally, I like dense technical documentation, no fluff, just a single block of text that reads like a masters thesis, though that's my flavor, hence for my flavor, AI generated articles are too fluffy most of the time, unless it's generated straight from my AI agent that wrote the code, because they write it like a technical document or a Repo Wikis. But that's because I dehumanize text I read, because stripping out the human factor, leaves cleaner logic, like using the AOM instead of the DOM for scraping, it strips the fluffiness, leaving cold, hard, clean facts. I'd take a wall of granular benchmark results over a 5 page article on trials and errors, because the benchmarks tell me where they spent their time and where they didnt, because that's what I need to know as a developer whether the system is feasible for production, or not. Not because they can explain how they fixed a null-pointer error in Rust, but by seeing their results on 1m transactions never hitting GC at all and the rule engine running faster than the theoretical max transactions per second, because that means it wont bloat and it wont ever bog down.
But that's just me. What about you? Do you have experience reviewing code and fixing production errors from someone else's work? If you do, do you also find you learn more from validating and fixing their code, than you do writing a module from scratch by hand?
Book marking to read this later. Just letting you know lol
So to answer the question, "How would you decide whether the content is good or bad", you validate it. Honestly there's no different way about it, if it's an article you read it, if it's code, you test it E2E, if it's a business process you trial it, the only way to be certain and to judge it fairly, is thoroughly using it for the intended purpose and see if it breaks under pressure or edge-cases. Eg. AI doesnt make grammatical errors like over time written as overtime, but it can also go into too much depth for average readers, eg. giving formula breakdowns and lots of big words in a row, but that's tunable, by just telling it 'in layman terms please' and it rewrites it to be human readable, without needing in-depth knowledge of a dozen programming books. Though I feel like that subtracts from the content in many ways, by making it accessible, it's less targeted and is half the reason why people dont understand complex words like idempotency anymore. Because those were exact terms used for exact purposes, which translates to 'written once, never edited again', which doesnt quite convey the fact that it's a continuation system, like a merkle root, where the state is locked and is iterated on each edit, as opposed to edited in place. A long sentence, just to explain what got lost in using layman terms, instead of the right jargon.
So I guess, it's the target audience that has the final say, as it's always been. You cant please everyone, but generally speaking, atleast half the audience likes vanilla ice-cream enough to not complain, even if it's not as extravagant as chili-chocolate or as much depth as bourbon vanilla. AI chooses to write in those full-depth extravagant flavors, which many people dont have a taste for, atleast not in large servings. So it depends on your audience. Best convo I had on Dev.to was with someone using AI to generate every response. Best responses I've had to an article was entirely AI generated game + article, because people understood it better than if I had to explain hallucinations and context windows in my own words. So to each their own, if an article doesnt taste right, pick a different restaurant, only send it back to the kitchen, if you're willing to do so with suggestions to make it better.
I find generating comments with AI not helpful as much especially in learning and understanding the writers article. What of writing an article using AI ? For instance I pose a well descriptive idea or a concept I would like to share, keeping in mind I'm not a good writer, is refining it with AI a bad practice?
AI를 썼는지 여부와 콘텐츠 품질을 같은 축으로 판단하면 놓치는 부분이 많다는 질문에 공감합니다. 제 기준은 독자가 읽은 뒤 더 나은 결정을 내리거나 문제를 재현·검증할 수 있는가입니다. 특히 바이브코딩 프로젝트라면 결과물보다 왜 그 선택을 했고 무엇을 버렸는지 설명할 수 있어야 작성자의 판단이 드러난다고 봅니다.