Aug 25, 2026 · From Substack

A Tip for Developing Your SpAIdey Sense

How to spot AI-generated writing without watermarks — or em dashes.

With all the talk about Anthropic’s new watermark, here’s a little experiment that may help you spot AI generated text without watermarks —* or even our beloved em dashes.

Step 1. Go to YouTube.

Step 2. Find one of those completely AI-produced videos. (You know, the ones that have turned one of the most prestigious brands on the internet into its biggest slophouse.)

Step 3.​​ Listen to the AI-generated voice awkwardly read the AI-generated script.

If pattern follows, you’ll hear what some call the uncanny valley of language. And once you do, you can’t unhear it. It even transfers back to text-only content.

No idea how it works, but here’s my theory: When we read text, whether human-written or AI-generated, we project our internal monologue onto it. We instinctively / automatically smooth it out, insert nuance and add emotional inflections, etc.

On the other hand, listening to a synthetic voice read a synthetic script strips away our subconscious editorial charity, thereby exposing the uncanny valley.

Give it a shot. After just a couple videos, I felt more in tune with the sound of AI writing and way more confident in my ability to spot LLM-generated text, no watermarks (or em dashes) needed.

Anyone else experienced this?

*100% ORGANIC, AI-FREE. No vectors were compared in the creation of this em dash.