What Google's Scaled Content Abuse Policy Actually Prohibits

Google's published spam policies do not prohibit writing with AI. What they prohibit is defined precisely: "Scaled content abuse is when many pages are generated for the primary purpose of manipulating search rankings and not helping users," and the same policy names the tool only inside that frame, calling out "using generative AI tools or other similar tools to generate many pages without adding value for users." Three conditions, all doing work: many pages, ranking as the primary purpose, no added value for readers. Enforcement runs through automated systems and human reviewers, and when a human is involved the outcome is that "some or all of that site will not be shown in Google search results," with reconsideration taking "several days or weeks." That timeline is the real cost, and it is why this is worth ten minutes of reading before your next publishing sprint.

Four beliefs, checked against the actual text

"Google penalises AI-written content." Not in anything Google publishes. The spam policies page conditions the violation on scale and purpose, not on which tool produced the draft. A single well-researched article drafted with AI assistance and edited by someone who knows the subject does not meet the definition. Two hundred near-identical location pages do, whoever or whatever typed them.

"Disclosing that we use AI makes it compliant." Disclosure sits in a different document and answers a different question. Google's helpful content guidance asks, under "how" content is produced: "Is the use of automation, including AI-generation, self-evident to visitors?" That is a reader-trust question. It is not an exemption from the scaled content policy, and a disclosure line at the bottom of 200 thin pages changes nothing about the three conditions above.

"This only catches content farms." The neighbouring policies are aimed squarely at small commercial sites. Thin affiliation is defined as "publishing content with product affiliate links where the product descriptions and reviews are copied directly from the original merchant without any original content or added value." Doorway abuse covers pages "created to rank for specific, similar search queries" that lead to intermediate pages less useful than the destination, which describes a lot of city-by-city service pages. Neither policy requires industrial volume.

"Letting a partner publish on our domain is free money." That arrangement has its own name. Site reputation abuse is "a tactic where third-party content is published on a host site mainly because of that host's already-established ranking signals." If someone is paying for a subfolder on your domain because your domain ranks and theirs does not, that transaction is the policy's definition, not an edge case near it.

The page-level audit, from Google's own questions

Google publishes the self-assessment it wants site owners to use. Turn each one into a pass or fail per page rather than a score, because a score lets you average away the failures. Run it on ten pages, weighted toward whatever you published fastest.

1. "Does the content provide original information, reporting, research, or analysis?" Fail if everything on the page could be assembled from the first three results for the same query.
2. "Is this the sort of page you'd want to bookmark, share with a friend?" Fail if you would not send it to a customer who asked the question directly.
3. "Is this content written or reviewed by an expert?" Fail if no named person with relevant experience touched it. Google's "who" guidance is about clear authorship, including bylines and author background.
4. "Is the use of automation, including AI-generation, self-evident to visitors?" Fail if a reader would feel misled on learning how the page was made.
5. "Do you have an existing or intended audience for your business or site that would find the content useful?" Fail if the honest answer is "people arriving from search," full stop.

Above these sits E-E-A-T: experience, expertise, authoritativeness and trustworthiness, of which Google says "trust is most important." Content does not need to demonstrate all four equally, which is a more forgiving standard than the acronym's reputation suggests. A shop owner writing about what actually breaks on the products they sell has experience, and experience is one of the four.

The scale rule

Here is the test we use, drawn from Google's own framing that "why" is "perhaps the most important question," with content existing to help people rather than to "attract visits from search engines."

If a batch of pages could not survive being read one at a time by your intended audience, the batch is the problem, and the drafting tool is not.

That rule cuts cleanly through most decisions. Twelve pages, one per product you actually sell, each containing something you know and the manufacturer does not publish: fine, whether you dictated them or drafted them with a model. Twelve hundred pages, one per city you have never worked in, differing only in place names: not fine, and it would not have been fine in 2015 either. The publishing method never changed the answer. Cheap generation just made the wrong side of the line much easier to reach in an afternoon.

What enforcement looks like from your side

Algorithmic demotion is invisible. Manual actions are not, and they show in the Manual Actions report in Search Console. The listed types include thin content with little or no added value, scaled content abuse, site reputation abuse, user-generated spam, cloaking and sneaky redirects, hidden text and keyword stuffing, and unnatural links in both directions. Affected pages are "ranked lower or omitted from search results without any visual indication to the user," so the first symptom is usually a traffic chart, not a warning.

Recovery is a sequence: fix every violation across all affected pages, submit a Request Review describing what you changed and what the outcome was, then wait for email. Google says "most reconsideration reviews can take several days or weeks," and link-related requests take longer. Price that as downtime. Several weeks without organic search is a long time for a business whose lead flow depends on it, which is the strongest practical argument we know for the position we took in distribution is the moat: build at least one channel whose rules you are not subject to, starting with the email list you own.

A rule you can hand to whoever publishes next

Before anything goes live, one person has to be able to name the specific thing on the page that a reader could not have got from the first page of results, and say it in one sentence. If that sentence exists, the page passes the policy's value test regardless of how it was drafted. If it does not exist, no amount of editing, disclosure or word count will supply it, and you should not publish the page at all, let alone another 199 like it. The reliable source of that sentence is not a prompt. It is the thing you already know from doing the work, or from the complaints your customers are posting in public.