Site Reputation Abuse: What Google Suspended in Europe — and What It Did Not

01What the policy covers, and what changed on August 30

The site reputation policy — introduced in 2024 and widely known as "parasite SEO" — does not target third-party content as such. It targets a specific arrangement: third-party content published on a host site mainly because of ranking signals that host earned with its own, first-party work. A news publisher rents out a subdirectory, a white-label provider fills it with coupons or casino reviews, and the section ranks on borrowed standing. Google names the exception explicitly: wire services, syndicated news, user-generated forums, opinion columns and advertorials aimed at readers rather than at rankings are all outside the policy.

On August 28, 2026, the Google Search Quality team announced a change to how the policy is enforced in the European Economic Area, effective August 30. Google frames it as the result of regulatory pressure and is unusually blunt about its reservations, writing that it remains concerned "that an overbroad application of the DMA could prevent us from addressing real threats to the integrity of our search results."

The change is narrower than the headlines suggest. Outside the EEA, a manual action still affects the portion of the site concerned, and as before the rest of the site is untouched. Inside the EEA, the impact of the manual action does not apply — but something takes its place, and the notification in Search Console arrives either way. The policy itself was not withdrawn, the detection was not switched off, and the human review that triggers both outcomes applies globally.

02The sentence that gives away the mechanism

Buried in the enforcement section of the policy page is a sentence that is far more interesting for E-E-A-T than the European news: "Google generally applies a presumption that individual pages (including new pages) match the overall quality of other pages on the domain." That is Google, in its own publisher documentation, confirming site-wide quality inheritance as an operating principle.

Until now, that principle could only be reconstructed from leaked field descriptions. The Content Warehouse leak documents siteAuthority as "converted from quality_nsr.SiteAuthority. NOTE: this is applied in Qstar" — a site-level aggregate, distilled from NSR, that feeds the Q* score. NSR aggregates content quality at chunk level and passes averages on to new areas of a site. The policy sentence describes exactly that behaviour in plain English, without naming a single field.

Site reputation abuse is, mechanically, the deliberate exploitation of this presumption. The rented section inherits a quality assumption it did not earn. And a second system measures the same arrangement from another angle: topic embeddings place every page in the site's topic space, where siteFocusScore records how tightly the site sticks to one topic and siteRadius records how far a page deviates from the site vector. A casino directory under a medical domain is not a policy judgment there — it is a distance.

03Why separation is not amnesty

What replaces the manual action in the EEA is the part worth reading twice. Google writes that the affected section "may be separated in our systems so that, over time, it ranks independently from the rest of the site" — and the policy FAQ then says outright what that does. The categorization "tells our systems that the presumption that we generally apply globally — which is that individual pages (including new pages) match the overall quality of other pages on the domain — no longer applies."

That is not relief. It is the withdrawal of exactly the presumption the arrangement was built to exploit. A section rented for the standing it could borrow has, by construction, little standing of its own: no accumulated NSR average, no share in the host's siteAuthority, no history — which is the position the hostAge signal describes, a passive dampening rather than a penalty. Google is careful about the timing, and so should anyone reading this be: separation "doesn't mean that the separate portion of the site immediately loses the ranking signals of the main site," it is explicitly "not automatic," and the effect arrives as the ranking systems "learn to rank these parts of a site independently." Slow is not the same as harmless.

The rest of the reporting's framing does not survive the source either. None of this runs quietly in the background: detection leads to human review — "If we detect that a portion of your site may be out of line with this policy, a site will be subject to human review" — and that review applies globally. Separation is the EEA form of the enforcement decision, not its absence. Google does clean the slate properly, though: previous EEA manual actions under the policy are lifted, and it states that neither a past manual action nor the categorization itself is used as a ranking signal. Search Console notifies either way, and reconsideration and, for eligible sites, mediation remain open.

04What the algorithmic layer was never asked to suspend

Google also published the criteria the human review applies, and they read like E-E-A-T translated into operational questions. Is the content presented consistently with the host — design, formatting, typography, UX? Are there quality problems on the page that do not appear on the main domain? Is authorship explicitly acknowledged, and does anything contest it? Does the same or near-identical content appear on multiple other sites? Google adds that no single factor is necessary or sufficient on its own.

Every one of those four has a machine counterpart in the system matrix. Presentation and topical fit are what siteFocusScore and siteRadius measure. Departure from the domain's quality standard is what chard and contentEffort estimate per page. Near-identical content across sites is what OriginalContentScore addresses. Only authorship has no field of its own — a gap that is itself informative, since the raters are instructed to research the reputation of creator and operator separately whenever they are not the same entity.

This is the part that survives the European change untouched, because it was never a manual action to begin with. Panda evaluates the share of weak pages across the whole site rather than its best work — quality as a distribution, not a peak — and a large rented section is part of that distribution. The neighbouring scaled content abuse policy is unaffected by the EEA adjustment entirely. The manual action explicitly spares the rest of the host site; the site-wide mechanics carry no such promise. That asymmetry is the reason to remain careful even where enforcement has been relaxed.

05References