I traveled to Scotland with a friend years ago. We still love to talk about the Scotch tasting we enjoyed at the Balvenie distillery.

Sitting around a table, sipping some amazing 30+ year old whisky, savoring the aromas and flavors, a person at our table described his dram more pretentiously than I thought possible.

He described tasting dewy lavender kissed by honeybees on the west facing slope of a meadow, or something as goofy as that. I tasted honey.

When you drink Scotch, as every tour guide told us, you get what you get. Everyone has a different palate, so everyone has a slightly different experience.

While it is possible we were both right, I know I thoroughly enjoyed drinking some of the best Scotch I ever tasted.

The other guy? The one performing?

His need to gush and draw attention to himself, instead of savoring the whisky, likely left him feeling as smug and self-satisfied as the cadres of the AI Demon Hunters now running rampant.

The Road to Hell

Yes, AI Demon Hunters. If you haven’t encountered them yet, you soon will.

They take different forms. Sometimes, they’re the folks who say they can tell when anything’s been written by AI. They just know by the word choice, the syntax, the punctuation, whatever.

Other times, they assume AI has been used to create something. With no evidence or grounding in fact, they lambaste and ridicule as on this Reddit post I read recently.

An International Master chess player had spent months building a program to help people learn how to think through positions. Instead of considered feedback, one commenter threatened to report him because they assumed he’d used AI.

Into this charged environment, which already has tools like the less-than-perfect Pangram now baked into Substack and others, Anthropic is about to begin watermarking text generated by its models released on or after August 2, 2026.

The AI Demon Hunters salivate, and are missing the point.

Touching Isn’t Creating

Just because an AI tool touches something does not mean the person who used the tool should be vilified. What matters is the intent of the action and the quality of the material.

As Anthropic pointed out in its announcement, “Detecting a Claude mark tells you that the content may have been processed by Claude. It does not, on its own, confirm the full provenance of the content.”

There are real reasons to care about provenance, but detection tools can’t touch them: contractual disclosure, academic integrity, legal filings, attribution. At best, a tool tells you a machine was involved, but it can’t say what that involvement looked like.

Which is the crux of the issue: when we’re so concerned about whether AI has or has not been used to do something like draft text, we miss the entire point.

So we need to go deeper.

Does the text convey meaning well? Do the chosen words, placed in a particular order, effectively convey a thought or a point of view?

If they do, the purpose of the text has been achieved.

But if it was wholly machine generated, it is soulless. That’s not something I would want my name on or to refer to in my own writing even though I can’t ignore its functionality.

Functional doesn’t equal quality.

As AI advancement accelerates and the technology becomes more entrenched in society, we should care more about identifying and valuing quality work than about whether AI played a part in it.

Why?

Do The Hard Work

Because generative artificial intelligence makes stuff up.

As a result, any time we use AI, we must check every output we receive. It’s not just something we should do. We have to incorporate it into how we engage with AI.

When I generate a client proposal, AI plays a formative role in how I think about the work and in creating the document. And then I pore over every word, every assertion, and every promise like my life and my reputation depend on it.

I need to know that, if somebody runs my document through an AI detector and gets a hit, I can defend everything with my name on it. Maybe I don’t get the business, but I know there’s enough of me in whatever I send out that I’m willing to put my name on it.

It’s the standard I hold myself to, and one I recommend you adopt especially when you hear a person say, “Hey, I used Chat to do this” or maybe, “I ran this in Copilot and I think we should do this….”

Even if the person hadn’t disclosed their AI use, I wouldn’t bother running anything through an AI detector, though. Instead, I’d use my BS detector to figure out if what they proposed made any sense whatsoever.

No AI required.

Defend Yourself Against Slop

We are all assailed every day by messages, by ads, and by requests for us to do things. If you’re in business, people often send you proposals and sales pitches.

Rely on your knowledge and your common sense to assess these requests. That’s really all you need.

Ask a consultant to defend a claim in the proposal. Suggest they explain how they arrived at a conclusion or maybe a number in the document.

Somebody who did the work, who completely thought through what they sent you will be able to answer simply and succinctly. The person who can’t shouldn’t get the work regardless of whether they used AI or not.

No AI detector needed. No Demon Hunters brought to bear. Just good old rational discretion.

We can all drink to that.