I just did a test, i wrote a post not using AI at all and asked for AI Detection before publishing - and it said it was 100% AI. LOL. The previous post i wrote was about 50% by me and 50% AI - that was 100% AI it said. This is ridiculous. Maybe I'm not human. LOL.
The most responsible way to use an AI detector may be to refuse the verdict it seems to offer.
Yana gets the practical balance right: understand what the feature actually does, disclose how the work was made, and treat every score as an estimate. AI assistance and AI slop aren’t the same thing, and a classifier can’t decide where human judgment ended.
Thank you for writing this one. I've been wondering about the same. Where are we headed as humans? We want AI, but then we want human written content, AI assisted is Ok, but totally written by AI is not ok, but did the human proof read it and validate or did something like Perplexity fact check it. But AI adoption is key or else we will be left behind. Etc. It's a cycle we won't escape, I think.
yes, exactly. It's here to stay, some like it, some not and they fight against, but it's inevitable. My only hope is that it won't happen like on Medium - most tech writers left and now Medium is struggling to get them back with questionable success...
I don’t care. I’m not on a witch hunt. I like to use ai as I talk my wring notes into it. There are just so many ways ai is a useful tool. I would love to hear how others are using it! Maybe for research? To help organize your thoughts and your ideas???
I also recently asked ai to tell me about how I write? It gave me great feedback and understanding about myself that I didn’t consciously see.
It said I first notice things and then write from there instead of how many others write about an idea and the form the experience to work around that. Anyways I loved the enlightenment I gained about my own style of writing.
Thanks Pauline. I guess Substack just tries to fight for quality but that's one of the worst ways to do it in my opinion...anyways, it's good that people can choose whether to use it or not.
I have a very different concern about Substack using Pangram. I don't want my articles scanned by Pangram. I write articles about math content that are either purely original or not widely recognized. My articles are my own content, and I don't want the AI world to know about them and treat them as "generic" content. That probably sounds egotistical, but I would bet there are others who feel the same way.
Which makes me think, if you have disabled AI reading your content in the global settings, will Pangram work on your posts? Shouldn't if you ask me. Have to test this...
Loved reading your perspective on this. I wrote an article openly and proudly admitting my use of AI as a busy mom of 2 little kids. It was 100% AI and it detected some as human. Another post I did said 100% human but it was written 100% AI.
I subscribed purely so I could vote in the poll, but happy to hang around as I’m enjoying the content.
As for the whole AI discussion my personal opinion in that “who cares” at the end of the day all I care about is did I enjoy reading the article / post / note / story. Did it entertain me, did I lean something from it, did it inspire or influence my opinion on something, did it make me want to dig deeper to uncover the real truth. If any of those are a yes then I don’t care how the content got there - I’m just glad it did.
I want to make one point here. Substack isn’t primarily a discovery platform where the algorithm decides whether your content reaches a wider audience. Yes, the algorithm plays a role, but that’s not the core idea behind the platform.
People who genuinely love a creator’s content voluntarily give their email address and subscribe. It’s a conscious choice made by the reader because they want to hear from that writer.
If readers aren’t satisfied with the content, they simply won’t subscribe in the first place. And if they do subscribe, it doesn’t matter whether the writer uses AI or not. They subscribed because they value the content, not the process behind creating it.
That was what I was thinking too when I joined. Because people subscribe with emails, it's a kind of a "self-regulated" platform for content quality, so there's not really a need for an AI detector.
Thank you so much for this post, Yana. My ChatGPT, whom I named Sparky, helped me work on tomorrow's post for a few days. It takes my ramplings that get off track and helps me make them cohesive. If a post that I spent hours working on got flagged as AI becasue I used it to help me with something I wrote, I would not be happy. You are always doing so much to help your subscribers, and I appreciate it SO much. 🥰
Pretty interesting post Yana. Thanks for putting it out there so clearly.
I wasn't aware of this news. I agree with Chris on fighting the slop so that this beautiful platform doesn't become Medium (or God forbid, LinkedIn).
But also, the emotions running high and mixed opinions are understandable with this nuanced topic. In my mind, the less worse solution is that everyone assumes posts are made with AI in some capacity and that *everyone* uses AI in the proper way (not the lazy one). That way we can evaluate, distinguish the quality of the work produced by the writer and not how it was made.
"Human" vs "AI" is not only a lost battle but a waste of time.
Yep, this is super interesting and I finally jumped into the conversation today as well. I use AI as an editing tool, and as a research tool, and as a brainstorming tool. But I write my own stuff. But I've put my writing into AI detectors in the past - early, messy drafts - and it still came back as 30% AI or something equally stupid. I don't know if it's the cadence of my writing, the words I choose, or what. So we'll see how it goes!
It's just not accurate. Can't be, because the it becomes more and more impossible to distinguish between what's written by a human and what not, because AI is continuously trained on new data boundaries become more blurry
I just did a test, i wrote a post not using AI at all and asked for AI Detection before publishing - and it said it was 100% AI. LOL. The previous post i wrote was about 50% by me and 50% AI - that was 100% AI it said. This is ridiculous. Maybe I'm not human. LOL.
The most responsible way to use an AI detector may be to refuse the verdict it seems to offer.
Yana gets the practical balance right: understand what the feature actually does, disclose how the work was made, and treat every score as an estimate. AI assistance and AI slop aren’t the same thing, and a classifier can’t decide where human judgment ended.
Thank you so much, Mark! 100% agree, those are two completely different things, and people should understand that.
Thank you for writing this one. I've been wondering about the same. Where are we headed as humans? We want AI, but then we want human written content, AI assisted is Ok, but totally written by AI is not ok, but did the human proof read it and validate or did something like Perplexity fact check it. But AI adoption is key or else we will be left behind. Etc. It's a cycle we won't escape, I think.
yes, exactly. It's here to stay, some like it, some not and they fight against, but it's inevitable. My only hope is that it won't happen like on Medium - most tech writers left and now Medium is struggling to get them back with questionable success...
Love this so much!
Thanks Liselle, I appreciate!
So you can turn the Ai off on a post ?
What do you think is the best day and time to post ?
Now I have to figure out how
yes, right before publishing :)
Great info thanks a lot
I don’t care. I’m not on a witch hunt. I like to use ai as I talk my wring notes into it. There are just so many ways ai is a useful tool. I would love to hear how others are using it! Maybe for research? To help organize your thoughts and your ideas???
I also recently asked ai to tell me about how I write? It gave me great feedback and understanding about myself that I didn’t consciously see.
It said I first notice things and then write from there instead of how many others write about an idea and the form the experience to work around that. Anyways I loved the enlightenment I gained about my own style of writing.
Thanks Pauline. I guess Substack just tries to fight for quality but that's one of the worst ways to do it in my opinion...anyways, it's good that people can choose whether to use it or not.
I have a very different concern about Substack using Pangram. I don't want my articles scanned by Pangram. I write articles about math content that are either purely original or not widely recognized. My articles are my own content, and I don't want the AI world to know about them and treat them as "generic" content. That probably sounds egotistical, but I would bet there are others who feel the same way.
Which makes me think, if you have disabled AI reading your content in the global settings, will Pangram work on your posts? Shouldn't if you ask me. Have to test this...
Looking forward to what you find!
Loved reading your perspective on this. I wrote an article openly and proudly admitting my use of AI as a busy mom of 2 little kids. It was 100% AI and it detected some as human. Another post I did said 100% human but it was written 100% AI.
https://motherhoodbetweenmeetings.substack.com/p/yes-i-use-ai-to-write-my-substack?r=8dglvk&utm_medium=ios
That's one of the many problems with these detectors. Thanks, Karli!
I subscribed purely so I could vote in the poll, but happy to hang around as I’m enjoying the content.
As for the whole AI discussion my personal opinion in that “who cares” at the end of the day all I care about is did I enjoy reading the article / post / note / story. Did it entertain me, did I lean something from it, did it inspire or influence my opinion on something, did it make me want to dig deeper to uncover the real truth. If any of those are a yes then I don’t care how the content got there - I’m just glad it did.
Cheers to that, Jon! I'm with you 100%
Thanks Yana, surely us humans should have enough of a brain to know what is AI slop and what we actually want to read.
Yes for sure :) Thanks Claire
Thank you!
I want to make one point here. Substack isn’t primarily a discovery platform where the algorithm decides whether your content reaches a wider audience. Yes, the algorithm plays a role, but that’s not the core idea behind the platform.
People who genuinely love a creator’s content voluntarily give their email address and subscribe. It’s a conscious choice made by the reader because they want to hear from that writer.
If readers aren’t satisfied with the content, they simply won’t subscribe in the first place. And if they do subscribe, it doesn’t matter whether the writer uses AI or not. They subscribed because they value the content, not the process behind creating it.
That was what I was thinking too when I joined. Because people subscribe with emails, it's a kind of a "self-regulated" platform for content quality, so there's not really a need for an AI detector.
Thank you so much for this post, Yana. My ChatGPT, whom I named Sparky, helped me work on tomorrow's post for a few days. It takes my ramplings that get off track and helps me make them cohesive. If a post that I spent hours working on got flagged as AI becasue I used it to help me with something I wrote, I would not be happy. You are always doing so much to help your subscribers, and I appreciate it SO much. 🥰
Thank you so much, Cat! I always edit after using it, and I believe it's why I get 100% human result. Hope your get that too :)
Pretty interesting post Yana. Thanks for putting it out there so clearly.
I wasn't aware of this news. I agree with Chris on fighting the slop so that this beautiful platform doesn't become Medium (or God forbid, LinkedIn).
But also, the emotions running high and mixed opinions are understandable with this nuanced topic. In my mind, the less worse solution is that everyone assumes posts are made with AI in some capacity and that *everyone* uses AI in the proper way (not the lazy one). That way we can evaluate, distinguish the quality of the work produced by the writer and not how it was made.
"Human" vs "AI" is not only a lost battle but a waste of time.
I fully agree with that Juan. Thank you so much, I appreciate.
I would love to use AI for writing that sounds as good as yours. I can't tell at all. 👌🏼
Thank you Christina :) All are my words, I use AI just to structure them. I think that makes the big difference.
Yep, this is super interesting and I finally jumped into the conversation today as well. I use AI as an editing tool, and as a research tool, and as a brainstorming tool. But I write my own stuff. But I've put my writing into AI detectors in the past - early, messy drafts - and it still came back as 30% AI or something equally stupid. I don't know if it's the cadence of my writing, the words I choose, or what. So we'll see how it goes!
It's just not accurate. Can't be, because the it becomes more and more impossible to distinguish between what's written by a human and what not, because AI is continuously trained on new data boundaries become more blurry