The AI watermark: Do we care?

Does the Claude watermark actually matter? (spoiler: probably not how you think)

Anthropic now watermarks everything Claude writes, and we all got our pants in a twist. As though we're about to be exposed for using AI to write content.

The watermark is near-invisible, woven straight into the text. But Anthropic have only done it to comply with EU law, and Google say they won't penalise watermarked content – only crap writing.

The watermark has nothing to do with the reader. Once you realise that, the watermark stops mattering.

In marketing, PR and journalism, the only thing that actually matters is whether the reader keeps reading. You can be as literary and high-brow as you like, but if the reader doesn't like what you write, they'll bin you off.

That was the biggest lesson I took from working in tabloid journalism. Put your "maybe one day I'll write like Hemingway" hat aside and focus on the reader.

The real problem with the watermark, and with AI content generally, is this: we don't know if the reader cares whether a human or a robot wrote what they're reading.

We don't care that CGI replaced hand-drawn animation in Hollywood. We say we care – "oh, the loss of the craft" – but the market proves we don't care enough.

We don't care that music is formulaic. We say we care – "oh, everything sounds the same" – but homogeneity is clearly working.

If we as a society don't actually do anything about the things we supposedly care about, don't be surprised when content – movies, music, marketing material – becomes formulaic. AI content is just the latest version of that same formula.

AI content written by LLMs is formulaic by design, and we've all adopted it anyway. Do we care that AI is undercutting original writing? Us marketers, journalists and PRs might. Clearly the audience doesn't – or not enough of the audience, anyway.

Until we know whether the reader actually cares about AI content, we can't prove its true value. Brands need to know if their audiences and customer bases care, or whether they even notice.

So here's a question: do you care if a brand's voice is clearly AI-generated?

Yes? No? No idea?

Because I'd bet the decision on whether to use 100% AI content, human-edited AI content, or 100% human content is based purely on the preference of the marketing team in charge of the brand – not on the audience at all.