Why AI Generated Comments Weaken Real Writing

A collaborative guide to what makes comments feel human.

This post began in a Note and a long conversation that followed. The issue surfaced quickly. Most uses of AI sit comfortably within our writing. Drafting, revising, translating, clarifying. All fine.

There is one place where this breaks. Comments.

In this post we will:

• Explore why automated comments feel hollow and erode trust.

• Share examples from writers who see this happening across the platform.

• Gather practical advice on what makes comments feel human and worth replying to.

Slow AI is built on reflection and deliberate use of tools. Comments are one of the clearest places where that deliberation matters.

---

---

Why people react strongly to automated comments

Writers open their inbox and find a pattern. A polite statement that could appear under any post. A template tone. A surface warmth with no presence behind it. The goal is not to respond. The goal is to be seen responding. It makes the space feel thinner for everyone.

Mia Kiraki 🎭 points out that comments are not about volume. They are about care. A comment that aims for reach instead of connection stops being a comment.

Dallas Payne highlights the confusion that follows. Many AI generated comments read like translations of translations. They contain sentences no one would ever say. The reader ends up analysing the comment instead of engaging with it.

Goodnex sees this as a structural risk. When communities fill with shallow, automated replies, genuine conversation collapses. People stop showing up.

There are also legitimate uses of AI in comments.

Marcela Distefano uses AI to translate her own thoughts. The intention is hers. AI handles the phrasing. That expands participation. It does not replace it.

Others use AI to check grammar or adjust clarity. This is fine. The problem is outsourcing the whole act of response to a system that has no stake in the conversation.

---

If you are new to Slow AI, here is our first invitation.

Slow AI #1 – Teach It Something It Cannot Know

---

What writers value in comments

Marcela Distefano My native language is Spanish, a language so rich in varieties (especially in Argentina, where we have tons of unique phrases and sayings that other Spanish-speaking countries don’t even know) that sometimes even Spanish speakers struggle to understand each other. Because of this, I found that using AI to write my comments on Substack helps me express what I truly mean. Traditional translators often produce literal translations that miss the nuances of idiomatic expressions or cultural context, leading to confusing or awkward results. For example, they might translate pescar en la pecera as “fishing in the fish tank,” which doesn’t make sense in English, whereas AI suggests idiomatic equivalents like “preaching to the choir.” Another funny example is when I want to say no pegar un ojo to mean I couldn’t sleep, but a literal translator might say “not hit an eye,” making readers think I’m mentally disturbed rather than just sleepy! This ability to grasp meaning beyond words is why AI has become an invaluable tool for me to communicate naturally and accurately in a language that is not my own. But I understand the point. No one likes shallow comments meaning nothing. We are here to learn from each other.

Dallas Payne: The act of writing and publishing can be a very vulnerable experience for the author. When I comment on a post or note, I see it as a moment that I can take to interact with the work and the author in a meaningful way. I have the opportunity to say that I see them, I understand what they are sharing, that I appreciate their contribution. Whether I highlight a statement I liked, riff off a joke they made, or just let them know that I learned something valuable and enjoyed the read, it matters. If I’m feeling particularly brave, I test my understanding and ask a deeper question. And receiving the same back on my own work? Well, it really makes my day! Comments don’t need to be the most articulate, intelligent words you have ever written, they just need to be genuine and sincere. This is how we create community.

Mia Kiraki 🎭 : I write a lot about robots and how they eat your homework. My entire newsletter is dedicated to engineering AI workflows that help you be more exceptional, not less. But there is a very hard line between automation and impersonation. Substack has been one of the few places on the internet lately that really feels like a community - a lovely dinner party with family and friends rather than a content farm. AI comments are the drunk uncles that ramble on about something totally irrelevant to the topic. They have the “texture” of engagement but without the substance of attention. Thing is, I want your typos, your specific questions, your messy human thoughts, your personality. If I wanted a hollow observation from a machine, I’d just prompt one myself.

AI Meets Girlboss: I remember my first week on Substack clearly. My inbox filled with comments that looked friendly at first glance but felt hollow the moment I read them. They were the kind of lines you could paste under any post. I even caught myself wanting to call it out, but I was too new to risk sounding dramatic, so I swallowed the disappointment and assumed this might be the culture here. Comments to me, mean that I’m actually seen on this platform, that my thoughts landed somewhere and were understood. Fortunately enough, I soon found writers whose comments sounded unmistakably human. They had original thought and felt more alive. Real comments make the Substack space feel warm, like we’re at an actual networking event and we’re cheering each other on. Automated ones make it feel empty. I learned the difference and it now shapes how I show up.

Carolina Wilke: I use social media as my online garden. Here on Substack, I comment on accounts with content and perspectives I appreciate… or who share things I love seeing, like watercolor, cute drawings, and travel. If I like what someone does and I feel called to comment, I want them to feel me the same way I felt them when I read their piece. I’m all in for using AI to help me express something clearly or to correct my English mistakes (English is not my first language). But you can feel the emptiness of a question or comment… the almost perfect formulation that comes from AI only, without the heart of the commenter… and that doesn’t amplify connection. If you can’t comment from your heart (whether agreeing or disagreeing) … and since most of us are here to truly connect with one another … why bother?

Goodnex: LinkedIn used to be my favorite corner of the internet… until LLMs went mainstream. And as someone who helps consultants keep their pipelines warm through outreach, I’ve watched how fast distrust spreads when comments and DMs start sounding hollow or templated. Substack became my escape - real writers, real conversations — until lately. Now I’m seeing comments that just summarize the author’s post and spit it back without an ounce of personal context. I get using AI for typos, grammar, clarity (especially when English isn’t someone’s first language). But automating your comments? That’s where I draw the line. Commenting is community building - the heartbeat of our social engagement, the first point of contact (look how I ended up here), and see and use it as an extension to my outreaches. I love leaving jokes, random POVs, and bringing a friend to others’ comment threads. Tell me how I could do that if I summarily use AI to automate the process? And I don’t want my post summarized back to me. I want real voices, jokes, unrelated questions, and all the stuff that makes engagement feel alive instead of AI-airbrushed. A single line of reframe in your own words is all it takes to keep a thread alive.

Laura O'Driscoll: There’s a very specific kind of exchange that happens in comment sections, which goes far beyond words. You, a writer, exchange your time, energy, and – often – little pieces of your heart and soul with a wider community. Those offerings are not small, and neither is their value. So, when somebody makes the gesture of returning you a genuine slice of their own time, energy, and maybe even heart or soul in the comment section, it’s quite a beautiful thing. I, for one, was pretty much convinced you couldn’t find that kind of genuine human exchange on the internet any more, if ever you could…until I stumbled into the marvellous existential cocktail party that is Substack. And that’s why I’m all for constructive, thoughtful AI use, including in comment sections. If you know what you want to say, but for one reason or another (and there are many good ones) you can’t find the language for it, go for gold. Use the tools you have to find the words you need. There’s still a beating heart inside that comment. But when somebody walks into that party and starts handing out word salad without taking the time to put some of their own thought into the recipe, it serves no-one. At the end of the day, a comment is a small but vital brick in the foundation of human connection. You don’t need to think up a masterpiece: you just need to think.

---

For a selection of free resources to help slow down and reflect with AI visit The Slow AI Library.

The Slow AI Library

---

What we are learning from this discussion

A simple approach for commenting with care is emerging.

• Say one thing that shows you read it.

• Say one thing it made you think about.

• Ask one question or offer one thread of connection.

This is enough. It signals presence. It invites conversation. It keeps the space human.

---

Let’s Collaborate

If you enjoy Slow AI and would like to create something together, I would love to collaborate. To find out how, click here.

Collaborating with Slow AI

---

Why this matters

Slow AI is about deliberate practice. Comments are one of the last unautomated corners of the internet. They work because they require presence. When that presence is simulated, the signal weakens.

A genuine sentence that shows attention carries more weight than a perfect paragraph with no human behind it. This piece aims to protect the kind of community that Substack can still support.

---

From Teaching AI to Notice Neurodiverse Perception

Susan Gordon Byron and I grateful for all of your diverse and creative responses to this prompt.

Mia Kiraki 🎭 described how the Wednesday colour prompt helped her name a sensory experience she has always felt but struggled to express, showing how AI can support neurodiverse perception instead of flattening it.

Dancing Snail Chronicles highlighted how these questions feel poetic as well as technical and welcomed an approach to AI that leans toward creativity rather than pure productivity.

John Brewton put it simply: slowing down helps you notice where design leaves people out.

Taken together, these responses show why Slow AI depends on specific human experience and attention, not templated engagement.

---

If you have experienced automated comments yourself or have any tips for how to respond to writers, then please let us know below the line.

We read and respond to all your comments.

Go slow.