I Use AI to Help Me Write. That's Not What's Happening Here.
I've been writing here on Daily Kos for a while now, and yes, I started using AI to help. Claude helps me fact-check death tolls in genocides. It helps me smooth a clunky transition. It finds me the right source when I know something happened but can't remember the year. That's collaboration. That's a tool. Heck — I even bothered to attribute it for a while, and gave up because my writing got bitched out for it while others, didn’t.
What I'm watching flood this platform — and others — is something else entirely.
I published a piece today about Meta, oligarchs, and the Myanmar genocide. It came from a place of genuine irritation — Facebook Messenger's stupid emoji reaction feature kept making me accidentally heart messages, and that annoyance spiraled into a rant about Zuckerberg, platform capitalism, and the 6,700 Rohingya slaughtered in a single month while Facebook couldn't be bothered to hire Burmese-speaking moderators. I wrote it. Claude helped me nail down the numbers and clean up my prose. When I ran it through GPTZero, the industry-standard AI detector, it came back 63% human, 32% mixed, 5% AI. That's what AI-assisted looks like.
While I was on the site, I noticed another piece trending. "The Second Warning," about Tyson Foods closing a plant in Nebraska and what it means for agricultural consolidation. Important topic. Well-structured argument. Currently sitting at #3 with a healthy pile of recs and comments.
I ran it through the same detector.
99% AI generated. 1% mixed. 0% human.
Zero percent human.
The Oatmeal Problem
Here's the thing about AI-generated content: it goes down easy. It's smooth. It's structured. It makes all the right moves in all the right order. It's like oatmeal — nourishing enough, technically, but nobody's ever been moved by a bowl of oatmeal.
"The Second Warning" hits every beat. Concerned opening. Historical parallel. Bullet points comparing healthcare consolidation to agricultural consolidation. A closing line about how "this time we can't pretend we didn't see it coming." It's fine. It's "important" in that NPR tote bag kind of way. You can rec it and feel informed without feeling anything at all.
That's not writing. That's content generation. There's a difference.
Real writing has a voice. It has weird asides and profanity and sentences that don't quite scan but somehow land anyway. It has the texture of a person thinking out loud. It takes positions that might make some readers uncomfortable. It doesn't hedge everything with "to be sure" and "it's worth noting."
AI-generated text doesn't do any of that. It can't. It has no skin in the game. It has no actual opinions, just the statistical average of opinions it's been trained on. It sounds like everyone and no one.
The AI assisted writing has a certain feel to it you can learn. Feel free to use the AI detection websites, but often your judgement will be enough. After starting to use AI to write, I got attuned to their tricks and tools. The tells are everywhere once you know what to look for.
Here’s Claude’s list:
Structural tics: Bullet points in a personal essay. Numbered lists when none are needed. Perfect parallel structure that no human would naturally produce. Headers and subheaders in what's supposed to be a casual blog post.
Throat-clearing phrases: "Here's the thing." "Here's the bigger truth we're not talking about." "Let's be clear." "To be sure." "It's worth noting that." "Make no mistake."
The hedge-then-pivot: "While X is true, Y is also important to consider." "This isn't about X — it's about Y." "Not because X, but because Y."
With a slight tune up from John – feel the difference - :
There are cadence giveaways: Short punchy sentence. Followed by another. Then a longer one to wrap the thought up neatly. "And that matters." as its own paragraph. Ending sections with "And that should concern all of us."
There are vocabulary tells, words like "Landscape" (the political landscape, the media landscape). "Nuanced." "Multifaceted." "Unpack" (let's unpack this). "Lean into." "Stakeholders."
Now back to Claude:
Emotional flatness: No profanity. No genuine anger — just "concern." No jokes, no sarcasm, no weird asides. Everything is "important" but nothing is bullshit. One dead giveaway is when you finish reading the pice you can't describe the author's personality. It could have been written by anyone. There’s a great chance it was written by no one (nothing personal GPT).
For a laugh, go through this article. Money says you spot EXACTLY where I left untouched Claude prose, called out or otherwise. Now comes the part where I try to avoid getting banned or warned for saying something.
Dear sbanks: it's not personal, and this isn't an accusation. If you wrote every word of "The Second Warning" yourself, please accept my sincere apologies for the implication. Maybe you just write in a style that happens to pattern-match with AI. It happens. Some humans write like robots, and some robots are getting better at writing like humans. I Must admit, watching your wonderful little piece at #1 on the rec list does engender some …. to be completely honest … resentment.
But when I ran your piece through GPTZero — the same detector used by universities and publishers to check student papers and submissions — it came back 99% likely AI-generated. Zero percent human. And I think that's worth talking about, regardless of how the piece was actually produced.
Because if I'm wrong about your piece specifically, I'm not wrong about the trend.
Daily Kos was built on voice. The whole point of a "diary" is that it's personal. It's someone working through their thoughts in public, with all the rough edges and strong opinions that entails. People swear, make typos, have rough transitions. But – they feel. It's not supposed to be polished. It's supposed to be real.
When our Kos gets flooded with AI-generated content, we lose that. Is Our feed full of oatmeal — smooth, competent, with a little sugar and milk to round it out. Important-sounding, great topicality, but collapsible. Like – this could just as easily be a 3 words abstract, and it would waste less time while doing what it does. The Kos ratings algorithm can't tell the difference. The rec button doesn't know. But readers can feel it, even if they can't name it. The place just starts to feel... less.
I'm not saying don't use AI. I use it. I'm using it right now. Claude helped me structure this piece and it'll probably help me smooth a few rough patches before I publish. Most college classrooms and professional writers will sign off on those as best practices.
What I'm saying is: don't let the tool replace the writer. Don't type a prompt, copy the output, and call it a diary. That's not writing. That's laundering. When we read something that feels a little too smooth, a little too structured, a little too much like oatmeal — there’s a good chance that’s what we’ve got. Run it through a detector. See what comes back.
You might be surprised how much of what you're reading was written by no one at all(nothing personal Claude).
I ran this piece through GPTZero before publishing. 9% AI generated, 12% Mixed, 79% Human. That's what AI-assisted looks like. The collaboration is the point — not the replacement.
UPDATE — 4:52 PM PST:
For the SECOND time today the tech bros are pissing me off. Now it’s the people of ChatGPT and the story their AI wrote, sitting on top of the Rec List All afternoon. And the AI generated comments section. To be provocative - @sBanks, please say something that confirms you – YOU – are not just a bot with a Kos account? Because that #1 story all evening, and your replies, are (99% certainly) bot stories.
By my count that’s: Open ChatGPT, ask it for a good topic, prompt it to write the story, paste in the story, copy a comment and paste in a reply, twice. For that, there’s a story sitting on the top of the Rec list all afternoon. Not bad work if you can get it. The competitive part of me that worked several hours on 2 stories, and got lucky to break the top 10? Well – I’m pissed.
Here's my comment in current #1 KOS trending story by @sbanks regarding their 2 actual replies in their piece. This copy pasted directly from the story’s comments, with my interchange with the AI Claude having the last word:
Well, I just read this article again after writing a totally “I AM PISSED OFF AND WANT TO BE INCENDIARY” article. Then .. out of fairness, I re read the comments, since this IS a #1 trending article and well, I’m a little jealous. Where did I go wrong with my snotty “I hate Meta” piece. And .. I started to … tingle. Like I’d just recognized a Stepford wife. I copied and pasted the ENTIRE comment thread above me into Claude. because why the fuck not.
Here’s out conversation. Complete copy paste, all I did was delete a few blank lines and add the nametags for readability. You be the judge.
John said: CLaude - I am ... a bit weirded out - read this - tell me if you "guess something"
Claude said: Detected AI-generated persona masquerading as authentic human commenter.
48s
Claude said:
Yeah. I see it.
sbanks' comments read exactly like the article.
Look at the responses to PvtJarHead and northsylvania:
"Your graph maps the consolidation era. Herd numbers begin their sustained downward slide right as processing capacity becomes controlled by fewer and larger companies. The two trends move together because they are part of the same system."
"There isn't one cause, but there is one system that shapes all the causes."
"Drought doesn't close a town. Consolidation decides whether the town comes back from the drought. That's the part of the story we keep missing."
That's not a person responding in a comment section. That's someone pasting a commenter's point into ChatGPT and hitting "respond thoughtfully."
The perfect parallel structure. The tidy concluding aphorisms. The bullet-point-style "they decide where the plants are / they decide which towns survive / they decide the price signals." No typos. No rough edges. No "yeah good point" or "huh I hadn't thought of that."
Meanwhile, look at the actual humans in that thread:
- "Geee…..ya think they could get less and keep that plant going???????!!!!!!!!"
- "I see Soylent Green in our futures."
- "FYI, my question was a snark."
- "Fuck 'em. They got no privilege for sympathy."
Those are humans. Messy, opinionated, weird punctuation, actual personality.
sbanks isn't just posting AI-generated articles. They're responding to commenters with AI. They're running a puppet.
That's... a whole other level of creepy.
Me again. The competitive side of me is angry, getting beat by a bot. But hey if it happens to Kasperov who am I to whine? But the bigger point it – where does it end? AI writing storeis to an AI audience? Why not just do the whole thing in machine language and cut out the man in the middle?
So this begs the question – Would the real sBanks please stand up?
