{"id":103289,"date":"2026-06-28T05:00:15","date_gmt":"2026-06-28T09:00:15","guid":{"rendered":"https:\/\/quotulatiousness.ca\/blog\/?p=103289"},"modified":"2026-06-27T12:30:49","modified_gmt":"2026-06-27T16:30:49","slug":"human-writing-has-a-unique-shape-and-the-the-end-of-social-media","status":"publish","type":"post","link":"https:\/\/quotulatiousness.ca\/blog\/2026\/06\/28\/human-writing-has-a-unique-shape-and-the-the-end-of-social-media\/","title":{"rendered":"“Human writing has a unique shape” and the the end of social media"},"content":{"rendered":"

On Substack<\/em>, Ryan Levesque<\/a> explains the major differences between human writing and AI-trained-on-human-writing:<\/p>\n

\"\"<\/a>

Graphic from The Digital Contrarian<\/em><\/p><\/div>\n

It turns out, slop has a shape.<\/strong><\/p>\n

And it’s the reason why AI generated writing sounds the way it does.<\/p>\n

In a new study<\/a>, a team of researchers at the University of Maryland<\/strong> and Google DeepMind<\/strong> ran an experiment.<\/p>\n

They took 10,272 writing prompts<\/strong> and gave each one to a human author and to five AI models: Claude, GPT, Gemini, DeepSeek, and Kimi.<\/p>\n

They generated 61,608 stories, at around 5,000 words each.<\/p>\n

Then, they looked at the underlying structure of each story: how the plot progresses, where the tension and conflict is placed, etc. etc.<\/p>\n

And from that structure, they could identify a human-written story from AI-generated slop nearly 93% of the time.<\/p>\n

\"\"<\/a>

Graphic from The Digital Contrarian<\/em><\/p><\/div>\n

What you’re seeing here in that image is the shape of AI Slop vs. Human Writing.<\/strong><\/p>\n

And there are five distinct ways that the shape of human<\/em> writing is decidedly different from the so-called slop generated by today’s AI models:<\/p>\n

    \n
  1. AI over-explains its themes<\/strong>. (instead of letting readers infer)<\/li>\n
  2. Human writing is less linear<\/strong>. (more time-jumps and flashbacks.)<\/li>\n
  3. AI relies on bodily metaphors to explain emotion<\/strong>. (81% vs. 38% human)<\/li>\n
  4. Humans reference specific texts, brands, places<\/strong>. (nearly 2x the AI rate)<\/li>\n
  5. AI narrative is less diverse<\/strong>. (fewer subplots and scenes, less dialogue)<\/li>\n<\/ol>\n

    […]<\/p>\n

    The Beginning of the End of Social Media?<\/strong><\/p>\n

    The clearest place to watch this shape materalize?<\/p>\n

    Social media.<\/p>\n

    This week, Farah Cormack<\/strong> mapped the predictable sequence, in a piece called “The Beginning of the End of Organic LinkedIn<\/a>“.<\/p>\n

    Her argument is that every platform moves through the same five stages:<\/p>\n

      \n
    1. Early adoption<\/strong>. A small group forms around something they love. It feels like a secret.<\/li>\n
    2. Scaling<\/strong>. The crowds show up, and so does the money.<\/li>\n
    3. Critical mass<\/strong>. Everyone’s here now. Organic and paid are both running hot.<\/li>\n
    4. Enshittification<\/strong>. The business model takes over the product. The feed fills with ads, and the place starts to feel like every other place.<\/li>\n
    5. Decline<\/strong>. The people who made it worth showing up for get fed up and leave.<\/li>\n<\/ol>\n

      Her read is that LinkedIn just crossed into stage four. The tell is its new Creator Marketplace<\/a>, a feature that literally puts your reach openly up for sale.<\/p>\n

      (If your own posts have been reaching fewer people lately, you’re not imagining things … this has been engineered.)<\/p>\n

      The shape of Enshittification is a five-stage decline, and most of the social media platforms we use are somewhere at stage 4 or 5 right now.<\/p>\n

      Futurist Sinead Bovell<\/a> goes further, and argues we’re watching the beginning of the end of the social media era itself<\/strong>.<\/p>\n

      The reality is that people don’t really post for friends\/social circles like we used to even just a few short years ago.<\/p>\n

      Bovell argues that the entire reason we post is to be seen by other humans.<\/p>\n

      That’s the whole deal.<\/p>\n

      We post to signal that we’re employable, or interesting, or worth following, or because we want to sell something …<\/p>\n

      And we do that, because real people are on the other end, watching us.<\/p>\n

      Take those real people away, and the entire thing stops making sense …<\/p>\n

      But that’s exactly what’s happening<\/strong>.<\/p><\/blockquote>\n

      Personally, I think LinkedIn hit stage four a lot sooner than this, almost certainly because it originated as a business-oriented platform. The owner of a company I worked for in the 2000s required<\/em> that all managers have active LinkedIn accounts, so I was “active” there for a couple of years, but I felt it quickly lost any actual benefits and became a forum of boastfulness and sycophancy. There were serious people on the platform, providing useful and insightful posts, but the vast majority of content was self-promotion and empty flattery.<\/p>\n","protected":false},"excerpt":{"rendered":"

      On Substack, Ryan Levesque explains the major differences between human writing and AI-trained-on-human-writing: It turns out, slop has a shape. And it’s the reason why AI generated writing sounds the way it does. In a new study, a team of researchers at the University of Maryland and Google DeepMind ran an experiment. They took 10,272 […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_feature_clip_id":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_post_was_ever_published":false},"categories":[32,28,15],"tags":[1506,1563,513,593,134],"class_list":["post-103289","post","type-post","status-publish","format-standard","hentry","category-books","category-media","category-technology","tag-artificialintelligence","tag-enshittification","tag-research","tag-socialmedia","tag-writing"],"jetpack_featured_media_url":"","jetpack_shortlink":"https:\/\/wp.me\/p2hpV6-qRX","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/quotulatiousness.ca\/blog\/wp-json\/wp\/v2\/posts\/103289","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/quotulatiousness.ca\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/quotulatiousness.ca\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/quotulatiousness.ca\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/quotulatiousness.ca\/blog\/wp-json\/wp\/v2\/comments?post=103289"}],"version-history":[{"count":3,"href":"https:\/\/quotulatiousness.ca\/blog\/wp-json\/wp\/v2\/posts\/103289\/revisions"}],"predecessor-version":[{"id":103294,"href":"https:\/\/quotulatiousness.ca\/blog\/wp-json\/wp\/v2\/posts\/103289\/revisions\/103294"}],"wp:attachment":[{"href":"https:\/\/quotulatiousness.ca\/blog\/wp-json\/wp\/v2\/media?parent=103289"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/quotulatiousness.ca\/blog\/wp-json\/wp\/v2\/categories?post=103289"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/quotulatiousness.ca\/blog\/wp-json\/wp\/v2\/tags?post=103289"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}