July 30, 2026
This week, two of the platforms I live on and support clients with both declared war on AI content in very different ways.
Substack rolled out a built-in AI detector that scores every new post for how "human" it is. LinkedIn just launched a report button that lets anyone flag your post as "AI Slop."
Two different companies, two different tools, and one very messy debate going on: who do we trust to tell the difference between AI and human writing?
Let's get into it.
What Substack did
On July 21, Substack CEO Chris Best announced the platform was adding an AI detection tool, built on a service called Pangram, to flag how much of a post looks AI-generated. Every new post over 100 words now gets scored, and readers can see that score if they go looking for it. Writers can turn detection off for their own work, and Substack has added a voluntary "how I made this" disclosure for anyone who wants to get ahead of the conversation.
The backlash was immediate.
Writers started calling it a "witch hunt." Ghostwriter and writing coach Alice LemΓ©e put it bluntly: these detectors are "notoriously, wildly inaccurate," and a single false accusation can follow a writer's reputation for a long time.
Julie Laufer said she's turning off the detection on her account for ethical reasons (even though Pangram gave her a 100% human score).
Pangram's own CEO says the false positive rate is around 1 in 10,000, but critics point out the errors aren't evenly spread. Non-native English speakers and neurodivergent writers get flagged far more often, simply because their natural rhythm and phrasing doesn't match whatever "human" pattern the model was trained to expect.
What LinkedIn did
LinkedIn has been building toward this for months. Back in May, the platform announced it would start quietly downranking posts that read like generic AI output: the recycled "thought leadership," the engagement bait, the "it's not X, it's Y" formula we've all seen a thousand times. LinkedIn said its internal detection was hitting 94% accuracy, though it never released numbers on how many real, human posts got swept up as false positives.
Today, that effort became visible and personal: LinkedIn rolled out a report option that lets any user flag a specific post as "AI Slop."
Here's the tension: LinkedIn is simultaneously encouraging people to use its own "write with AI" button in the post composer, while asking the community to police each other's writing for looking too AI. That is, at best, a confusing message. At worst, it hands out a new weapon disguised as a content-quality tool.
The problem: these tools don't actually work that well
If you want a case study in why AI detectors can't be trusted at face value, look no further than the Bible. When researchers ran a section of the Book of Genesis through the AI detector ZeroGPT, it came back 88.2% AI-generated.
The same tool has flagged the U.S. Constitution, passages from Harry Potter, and the lyrics to Bohemian Rhapsody as machine-written, too. In controlled tests, ZeroGPT's overall accuracy sits around 73.8%, meaning roughly 1 in 5 pieces of human writing gets wrongly flagged.
GPTZero's CEO Edward Tian has said publicly that many detectors are built to lean toward false positives on purpose, so that AI content doesn't slip through undetected. The tradeoff is that formal, clean, well-structured human writing (the Bible, legal documents, a tightly edited essay) reads as "predictable" to these tools, and predictable gets mistaken for machine-made.
So if a 2,000-year-old religious text and the founding document of the United States can't clear the bar, what chance does your Tuesday LinkedIn post have?
The other problem: it's not just the tools, it's the people using them
Even if the detection technology gets better, LinkedIn's report button adds a second layer of risk that has nothing to do with algorithms: human subjectivity. A report button run by real people isn't just a quality filter, it's a popularity contest.
Will people use it in good faith, flagging posts that genuinely feel hollow and generic? Or will it become one more way to report accounts they simply don't like, disagree with, or feel threatened by?
We already know how "report" buttons get used on other platforms. There is no reason to assume this one will be different.
I think I've seen this film before
This brought me back to the Facebook Election Era (yes, I'm getting old). Remember Facebook's "fake news" flag? In December 2016, right after the election, Facebook let users report stories they suspected were fake. Reported posts got sent to third-party fact-checkers (FactCheck.org, Snopes, and others), and anything they disputed got hit with a red "Disputed" flag right on the post.
It didn't work the way anyone hoped. Research found the flag sometimes backfired, making people more likely to share a disputed story, not less (a version of the backfire effect), where telling someone something they believe is false only entrenches them further. And the reporting itself was never purely about accuracy. It became just another lever people pulled on posts and pages they already disagreed with.
By December 2017, Facebook quietly killed the red flag altogether, replacing it with softer "related articles" links instead. Less than a year, and the company had already abandoned its own fix.
It's a recurring pattern with social media: a platform under pressure ships a reporting tool fast, the tool gets used for reasons that have nothing to do with its stated purpose, and the platform ends up walking it back or reworking it entirely. AI Slop reports may follow the same arc. That doesn't mean ignore it, it means don't build your sense of your own work around whether a crowd of strangers, or an algorithm trained to be trigger-happy, approves of it.
So where does that leave us?
Honestly, I'm not worried. Not for my own writing, and not for any of my clients.
These moments present an opportunity to go back to your why. Why are you creating content in the first place? Take a hard look in the mirror and ask whether you're proud of what you're putting out, and whether you'd stand behind it if someone questioned it line by line.
We always, always have a human writing, editing, and adding what I call signs of life into every piece of content before it goes out. Signs of life are the small, specific, unmistakably human details that no detector or algorithm can fake: the story only you would tell, the offhand observation, the sentence that sounds like you talking and not you reading a script.
If your content already has that, an AI detector score or a stray report from a stranger shouldn't shake you. If it does, that's worth addressing regardless of what any tool says.
Before you publish your next post, ask yourself:
- Could anyone else have written this? If you swapped your name for a competitor's and nothing felt off, it's not finished. Find the detail (the story, the number, the observation) that only comes from you actually living your work. If it's not there, add one.
- Is there anything here that builds real connection? Not oversharing, but the kind of honesty that makes a reader feel like they're talking to an actual person. A moment of uncertainty. A lesson that took longer than it should have. A thing you changed your mind about.
- Read it out loud. Does it sound like you, or like a script? If you'd never say it that way over dinner, your reader will feel the gap. Tighten it until it sounds like something you'd actually say out loud.
Detectors will keep getting better and worse in cycles. Report buttons will get used well and abused. None of that is fully in our control. What is in our control is whether the thing we publish could only have come from us.
Keep showing up as the very human, very imperfect version of yourself. That's the version no algorithm can flag and no stranger can convincingly report away.
You're doing great,
Brynne
p.s. Have you had a post flagged, scored, or questioned by one of these tools? I'd love to hear how it went.
When you're ready, here are 3 ways you can work with me:
- Send me one of your geeky topics and I'll send you an free idea for a post
- Ask me to review and edit your LinkedIn profile
- Have our team of humans manage your LinkedIn & Substack on retainer
Just reply here and I'll send you more info.
π New to Socially Minded? Read previous newsletters here and subscribe to get more of this in your inbox.
β
β