Google now says bulk AI pages count as low effort
Google's own documentation now spells out what its human reviewers treat as real work on a page.
Google has added a new section to its guidance on creating helpful, reliable, people-first content. Barry Schwartz reported the update, which covers the main content of a page and how Google's search quality raters review it. Raters are the human reviewers Google pays to score pages, and their guidelines shape how Google trains and tests its ranking systems.
The part that matters is a written definition of effort. Glenn Gabe quoted the new section, which describes effort as the extent to which human work went into creating the content or the systems powering it. Writing original analysis, translating a poem manually, or building a custom interactive map for a game counts as high effort. Automatically generating pages from feeds, or using generative AI to produce large amounts of text without manual oversight or curation, counts as little to no effort. The section also says that crediting other sources does not replace the need for original effort.
Why this is worth two minutes
Until now, "low effort" was a vibe. Site owners argued about where the line sat. Google has now put the line in its own documentation, with examples on both sides. That gives its raters, and anyone trying to guess how Google thinks, something concrete to point at.
This lands in the middle of a spam update that has already appeared to hit mass-produced pages. Gabe said he is waiting for the quality rater guidelines themselves to be updated with more of this language, and read the recent run of documentation changes as Google trying to tell site owners something.
What this does not say
The guidance does not say that using AI makes a page low effort. It names two specific things. One is spinning pages out of a data feed with no human involvement. The other is generating large amounts of text with nobody checking or curating it. A researched article that a person planned, edited and stands behind is not what that sentence describes.
What to do
If you publish a handful of pages a month and a human reads every one before it goes live, do nothing. If you have a batch of pages built from a template or a spreadsheet, open your sitemap or Search Console and list them. For each one, ask whether a reader gets an answer they could not get from the page next to it. If the answer is no, merge them into one good page or delete them. Keeping a thousand near-identical pages because they used to rank is the bet Google just described in writing.
So if you are taking advantage of loopholes and scaling like crazy via AI/programmatic, BEWARE.