The question "does Google penalize AI content?" has been answered, repeatedly, by Google itself and by two years of ranking data: no — it penalizes unhelpful content. The distinction matters enormously, because the sites losing traffic aren't losing it for using AI. They're losing it for publishing raw output at scale and calling that a strategy.
What the policies actually say
Google's guidance has been consistent: automation is fine when used for useful content; spam is spam regardless of tooling. The scaled-content-abuse policy that rolled out targets the actual failure pattern — many pages produced primarily to rank, with little added value, however they were generated. A thousand raw chatbot posts per week fit that description. Ten deeply-edited, genuinely-expert-reviewed pieces don't, regardless of which tools touched the draft.
What the data shows about winners
Look at the sites thriving with AI-assisted workflows and the pattern is boring: human expertise decides the angles, AI accelerates drafts and research, a human edits for accuracy and voice, and subject-matter review happens for anything consequential. The sites cratering share the opposite pattern: keyword in, prompt out, publish, repeat. No experience signals, no original anything, sentences that sound like everyone else's sentences.
The detector confusion
People conflate "AI content" with "AI-detectable content". Detectors are unreliable in both directions — flagging human writing, passing edited machine drafts. But the practical signal engines evaluate isn't statistical: it's whether the content demonstrates experience, adds something original, and serves the query better than alternatives. Detectors guess at process. Rankings reward outcome.
The working rule
Use AI like a fast junior researcher with a good keyboard: it drafts, it summarizes, it suggests — you verify, you add the parts only you can add (your data, your examples, your hands-on observations), and you edit until it sounds like your site. If you can't point at what the human contribution was, the page isn't ready. That rule survives every algorithm update announced so far, and it's the one the rankings have been quietly enforcing all along.
