Human writers vs AI drafts: the 50/50 experiment

Human writers vs AI drafts: the 50/50 experiment

Month 1: A pet-care site. Writers cost me $0.08/word — about $120 per article. Painful at scale.

Month 3: I split production. Half the new posts went to my human writer, half to AI drafts I'd edit myself for 30 minutes each. Same topics, same templates, published in alternating order so timing wasn't a factor.

Month 7, the data: human posts averaged position 6.2, AI-edited posts position 9.8. Close-ish. BUT the AI posts had a 3x higher "helpful content" decay — they ranked, then slid, because they lacked the specific anecdotes and product quirks my writer naturally included.

The turn: I stopped framing it as human OR AI. I had AI build the skeleton and research, then paid my writer half-price to inject real experience — the "my cat actually did this" details. $60/article, near-human performance.

The takeaway: raw AI content competes on quantity and loses on durability. The winning combo isn't choosing one — it's AI for the scaffold, humans for the lived detail that survives helpful-content updates. Pay for the part machines can't fake.