Google has updated its guidance on AI-generated content, and the message to publishers is blunt: a human needs to check the output before it goes live.

The new guidelines, surfaced by Search Engine Roundtable, tell site owners and content teams to manually factcheck and review anything produced by generative AI tools before publication. In other words, pushing raw model output onto a live page is no longer something publishers can treat as acceptable practice — at least not if they care about how Google evaluates their sites.

The update lands at a moment when publishers are wrestling with exactly this problem. Generative tools can produce a draft article in seconds. Verifying that draft — checking names, dates, figures, and claims — still takes a human being minutes or hours. Teams that skipped the second step were already running into trouble, and Google's refreshed guidance now makes the expectation explicit.

What does "review" mean here in practice? Based on the updated guidelines, it means an editor or writer reads the AI-generated text, verifies factual claims against reliable sources, corrects errors, and takes responsibility for the finished page. The human review step is the point. Content that ships without it carries the risk — both of factual errors reaching readers and of quality problems that reflect poorly on the site in Google's systems.

For publishers, the operational impact is straightforward. If your workflow is "prompt, copy, publish," Google is telling you that workflow is out of step with its guidelines. If your workflow already includes an editorial pass, the update mostly confirms what responsible teams have been doing all along.

The guidance also puts a number on nothing, and that is worth noting. There is no quota, no percentage of content that must be human-written, no disclosure requirement buried in this update. It is a process requirement: review and factcheck AI-generated material manually. How publishers implement that — full editorial review, spot checks, subject-matter expert sign-off — remains their call.

Why does this matter for the trade? Because AI-generated content has flooded the web over the past two years, and quality varies wildly. Some of it is serviceable. Some of it contains fabricated facts, outdated figures, or confident claims that fall apart under a simple check. Google's guidelines update is a line in the sand on the side of verification.

Publishers using AI at scale should audit their pipelines now. Questions worth asking: Who reviews AI drafts before they publish? What sources do they check claims against? Is there a named editor responsible for each piece? If the answers are "nobody," "none," and "no," the updated guidelines give you a reason to fix that before it costs you.

The bottom line for operators: Google's refreshed AI content guidelines require manual factchecking and human review of AI-generated content. The tools can draft. People still have to verify. That division of labor is now formal policy guidance, not a suggestion buried in a help doc.

We will track how Google enforces or expands this guidance in future documentation updates.