Our POV on AI and Content

Anthropic just started watermarking content that runs through Claude. Not just what Claude writes from scratch, it’s also the content we wrote ourselves and ran through it for a copyedit. They both get the same mark.

It’s a small detail, but it matters to us. It raises a bigger question we’d rather answer directly than let a badge answer for us: what’s actually ours, once AI has touched it?

Here’s how we think about it:

AI doesn’t get the raw material. The observation, the client detail, the thing someone said in a room that changed how we thought about a problem. All of that comes from being there. AI can’t generate that. It can only work with what we hand it. So we hand it drafts, bullet points, and half-formed thinking. Never a blank page and a topic.

Editing isn’t authorship. We use AI the way we’d use a sharp editor: to catch a flabby sentence, tighten a structure, or push back on a paragraph that’s explaining instead of showing. That’s a different act than generating the paragraph in the first place, and we don’t think it should be treated the same way — even if a watermark can’t currently tell the difference.

If we can’t tell you what a machine added, we don’t publish it. The test we use internally: could this sentence have been written about a different company? If AI smoothed something into that kind of generic, we rewrite it before it goes out.

We’re not precious about the tools. Actually, we love our tools. They make our work better and faster. What we are precious about is the thinking. AI drafts faster. It doesn’t watch a client’s face change when a plan finally clicks, and it doesn’t know which detail from a messy discovery call is the one worth keeping. That part stays human, on purpose, regardless of what any watermark says about the file.

So when you read something from us, the honest disclosure isn’t “AI-assisted” or “human-written” as a binary. It’s this: the thinking is ours. The tools just help us say it faster — and sharper.