Google has been explicit, repeatedly, about one point that gets lost in most discussion of AI and search: the company does not penalize content for being produced with AI assistance. What it penalizes is content produced primarily to manipulate search rankings, regardless of how it was written. That distinction matters more than most publishers using AI tools currently act like it does.
The practical signals Google's own guidance points to are consistent with what has always separated durable publishers from content farms: does the page demonstrate direct experience or expertise with the subject, is there a clear, accountable author or organization behind it, does the content add information or perspective beyond what is already available elsewhere, and would a reader feel the page satisfied their query well enough to trust the source again.
Why so much AI content still fails to rank
The failure mode is not the AI. It is publishing volume without editorial judgment — template-driven articles that restate a press release or a competitor's article with different phrasing, no original sourcing, no named author with any visible track record, and no updates when facts change. That pattern was penalized by search algorithms before generative AI existed; AI has simply made it cheap enough to produce at a scale that made the problem visible again.
What durable AI-assisted publishing looks like
The publishers who are holding up well combine three things: transparent sourcing (linking out to the primary reporting or research a piece draws on), a consistent, named editorial voice that takes accountability for accuracy, and a willingness to correct or update articles rather than leaving stale claims live. None of that is exclusive to human-only writing. It is entirely achievable with an AI-assisted workflow, provided the workflow treats AI as a drafting and research tool operating under editorial standards, rather than as a replacement for having any standards at all.