The bottleneck moved from drafting to designing the work. A primer for marketing teams.

What actually changed

Somewhere in the last year, "should we use AI in content?" stopped being a question. In June 2025, HubSpot's blog team made the switch to an AI-first writing model — one of the most sophisticated content operations in B2B, rebuilding its editorial engine around machine drafting.

The interesting part isn't that it happened. It's what the team learned on the other side. The public conversation assumes that because AI can write passably, the hard part is over: draft with a model, apply a light polish, ship. The manager who actually ran the transition — three years leading that blog team — says the assumption is wrong from the inside. And she isn't selling a tool. She's describing what broke and what held.

For marketers, that gap between assumption and reality is the whole story of AI marketing right now.

The ceiling is generic, not generated

The objection to AI content has always been quality. But quality was never the binding constraint — sameness is. As the HubSpot account puts it, a blog post