How do you distinguish AI content from real, human-made content?
AI is incredibly good, I’d even say almost perfect.
And for many people, that uniformity of perfect templates is starting to feel annoying.
For example, a few days ago, someone publicly showed that they built Anti-Grammarly – a tool that intentionally adds mistakes to text instead of removing them (to make it feel more human). But the tool itself is AI, so it’s a bit contradictory.
1) When we’re so flooded with AI-generated content, do you have any methods to recognise it?
For example, I keep noticing the same patterns:
– long dashes,
– phrases like “It’s not X, it’s Y,” and similar structures.
2) But what about beyond text, like images or video?
1.3K views
Replies
Bond
Why does it matter?
Why focus on how it was produced instead of the outcome?
Do you care if you asked someone what's 24x5 if they did the math in their head vs on a calculator?
Honest answer: AI-vs-human detection is mostly broken now. What still works is detecting whether a piece of writing was edited by someone with an opinion versus sitting on the surface of every angle. The METR research is the closest public reference for this kind of operator evaluation: https://metr.org/
There's not a reliable way to detect it with the text. Even the general structure doesn't tell you much, at least for articles where humans were pushed into the same handful of SEO-friendly templates before generative AI became widespread.
Honestly, you can't reliably - and chasing that is a losing battle. The better question is: does it add value? AI content that's accurate, useful, and sounds human beats robotic human content every time. The game has already changed.
I think we're entering a phase where AI isn't being detected by quality anymore, but by patterns. It's less about whether the writing is "too good" and more about whether everything starts to sound like it came from the same invisible editor.
For text, I notice repeated structures and a certain "over-optimized" tone. But for images and video, it's changing even faster. Early AI content had obvious tells (weird hands, strange motion), but newer models are getting good enough that context matters more than artifacts. If every image and every video can be generated perfectly, maybe the new signal of authenticity will be specificity — personal details, imperfections, and a point of view that only a real person would bother to include.
the tells for me are weirdly specific. real people contradict themselves mid-paragraph or drop a thought halfway and move on, AI is too evenly good, no throwaway lines. the adding-mistakes-back-in thing is funny, total arms race lol. do you even care if somethings AI as long as its useful, or is it more that you want it to feel like a person wrote it?
The biggest giveaway in text is the complete lack of strong, polarized opinions. AI is built to be helpful and inoffensive, which makes it incredibly bland. For video/images, it’s all in the micro-details: weird physics, nonsensical background text, and lighting that doesn't quite match the environment. Give it 6 months, though, and we might not be able to tell at all!
Honestly, I don’t think there’s a reliable way anymore, especially with images and video.
With text, I notice the same patterns you mentioned, but once people start intentionally editing AI output to make it look “human”, it gets even harder to tell.
I’m actually more interested in the other side of this — how AI itself decides what content or brands to trust when so much of the internet is AI-generated. That feels like a much bigger problem than just spotting AI content.