One Manual Action Now Has Two Geographies

Google split one of its spam penalties along a geographic line this week. On 28 August, Search Central published an update to the site reputation policy: from 30 August, a manual action applied under that policy has a different effect depending on where the searcher is. Outside the European Economic Area, the action directly affects search results for the affected portion of the site, while the rest of the site is untouched, as before. Inside the EEA, "the impact of the manual action won't apply" — instead, the affected section "may be separated in our systems so that, over time, it ranks independently from the rest of the site."

The change follows discussion with the European Commission, and Google says it remains "concerned that an overbroad application of the DMA could prevent us from addressing real threats" to result integrity. Notification still lands in Search Console, reconsideration requests still exist, and — new in this announcement — eligible sites can escalate to an external mediation scheme after reconsideration.

The tempting reading is that the EEA became a safe harbour for parasite SEO. It did not. The sharper reading is operational: one violation is now two different objects with different symptoms, different timelines, and different escalation paths, and most monitoring stacks are not built to tell them apart.

The repeated angle to avoid

The ten most recent posts here covered Copilot approvals as revocable attestations, cache storage versus HTTP representations, variant matrices versus sellable catalogs, htmx state ownership, Shopify's three order clocks, verified bots as policy joins, AI catalogs before conformance harnesses, security dashboard denominators, chat adapter semantics, and agent compute planes. Older overlapping posts covered situation-specific AI search pages, vary/caching content contracts, bot-contaminated analytics, and Turnstile-pass lead quality.

The weak version of this story would revive the old X needs Y formula — publishers need to prepare for the policy update — or reduce to a news recap: Google changed enforcement in the EEA, here is what changed. Reading the announcement alone gets you that.

The sharper thesis is that a manual action is now a region-scoped object with divergent observability. Outside the EEA it is a step change you can date from Search Console. Inside the EEA it is a slow separation that may not be datable at all, because "ranks independently over time" is an algorithmic process, not a labelled event. Different symptom shapes require different detection, different expectations for remediation timelines, and a forked dispute path. That is the part no single source spells out.

Fresh evidence and background serve different jobs

Source	Freshness	What it contributes
Update to the Site Reputation Policy	28 August 2026, effective 30 August	The two-outcome enforcement split, Search Console notification, reconsideration, and the mediation escalation
Defending Search users from "Parasite SEO" spam	13 November 2025	The regulatory pressure behind the change: the Commission investigation, Google's DMA argument, and the German court dismissal claim
Site reputation policy, spam policies	Current documentation	The criteria, the EEA note, and the violating and non-violating examples that define the line
March 2024 core update and spam policies	5 March 2024	Policy origin: site reputation abuse introduced alongside the March 2024 core update
Manual actions report	Current documentation	How actions surface in Search Console and what a match means
Reconsideration requests	Current documentation	The shared first step of both dispute paths
Google Search Mediation Scheme (CEDR)	Current scheme page	Free mediation for EEA business users over Google Search conditions of access

The fresh announcement supplies the event. The November 2025 company post supplies the why. The policy documents supply the line an operator actually has to stay on the right side of.

One violation, two enforcement surfaces

The mechanics are worth separating precisely, because the failure mode is assuming they are the same thing with different severity.

Outside the EEA, the manual action is direct and portion-scoped. It affects search results for the affected portion of the site. Google states the rest of the site is not affected, which was already the design intent of this policy: the weapon has always been the host's accumulated ranking signals, so the sanction targets the section that borrowed them.

Inside the EEA, the manual action's impact does not apply — but separation does. The affected section "may be separated in our systems so that, over time, it ranks independently from the rest of the site." Read that mechanism carefully. Independent ranking is not the absence of treatment; it is the withdrawal of inherited authority. A coupon directory that ranked because it lived on a news domain now has to rank as what it is. If the content's own quality would not earn that position on a standalone domain, independence usually means lower visibility — that is the point of the policy. But Google's wording is deliberately softer than a penalty: separation is described as something that may happen, over time, rather than a demotion applied on a date.

Three consequences follow:

1. The symptom shapes differ. Outside the EEA, expect a cliff you can correlate with the Search Console notification. Inside the EEA, expect drift — a section whose head terms erode over weeks or months while the rest of the site holds. Drift is easy to attribute to seasonality, algorithm updates, or measurement noise.
2. The timelines differ. A reconsideration request that succeeds lifts the manual action, but the EEA separation is a system behaviour that unwinds "over time," not on request. Teams that promise stakeholders a recovery date will be wrong twice.
3. The diagnosability differs. Search Console will notify you about the action regardless of geography. What it will not give you is a per-region view of the separation effect. The observability burden moves to your own stack.

The dispute path forks after the shared first step

Both geographies share one entry point: fix the issue, then file a reconsideration request. The fork appears after that.

For eligible sites, Google's announcement adds an escalation: "following the reconsideration request," disputes can be brought to mediation. The linked scheme is the Google Search Mediation Scheme administered by CEDR, which offers free mediation to business users in the EEA over Google Search's conditions of access — a dispute channel of the platform-to-business type European regulation requires, not a court and not a Google product.

The sequencing rule matters more than the plumbing:

• Reconsideration first, always. Mediation is positioned after reconsideration in Google's own wording. Arriving at a mediator without a documented fix and a rejected or unresolved reconsideration weakens the dispute and burns calendar time.
• Know what each track can produce. Reconsideration can remove a manual action match. Mediation can produce a negotiated outcome about access conditions with a platform that is not obliged to reinstate a ranking. Neither is an entitlement to the traffic the section used to borrow.
• Preserve the evidence both tracks will ask for. What the third-party relationship was, who authored the content, whether it was promoted to readers, and when it was removed or restructured. This is also the evidence that decides whether the section was inside the carve-outs in the first place.

Detection now depends on where you sample

This is the operational core of the change. A rank tracker that samples from one region, or an analytics stack that reports global blended traffic, cannot distinguish the two states.

Observed state	What you are likely seeing	First move	Escalation	What to monitor afterwards
Search Console shows a site reputation match; non-EEA visibility dropped sharply	Direct manual action effect outside the EEA	Fix or remove the section, then reconsideration request	Mediation if rejected and eligible	Per-region recovery of the affected section only
Search Console shows a match; EEA visibility eroding gradually, no cliff	Separation effect inside the EEA	Same fix; expect slow, un-dated unwinding	Same sequence; timeline expectations set lower	Section-level EEA trend versus a control set of first-party pages
No match; one section drifting in the EEA only	Possible algorithmic separation or ordinary volatility	Do not file reconsideration for a match you do not have; audit the section against the policy examples	None available for unlabelled effects	Whether drift correlates with third-party sections specifically
No match; a genuine drop everywhere	Core updates, seasonality, technical faults	Standard traffic-drop debugging before assuming policy	Not applicable	Indexing, crawl, and query mix before rankings

The third row is the uncomfortable one. Google's language — "may be separated," "over time" — leaves room for separation-like effects that never become a notified manual action. You cannot appeal what is not labelled. The only defence is knowing which sections of your site depend on borrowed authority, before a regulator-adjacent policy change makes that dependency visible.

Where the policy line actually sits

The policy text is more precise than most summaries of it. It applies where third-party content is published on a host site mainly because of that host's already-established ranking signals, which the host earned primarily from its first-party content. Third-party content alone is fine. The violation is the intent: hosting content so it ranks better than it could on its own.

Google's own examples draw the line in a way an operator can audit against:

• Inconsistent with the policy: an educational site hosting sponsored payday-loan reviews distributed to many sites; a medical site carrying an uninvolved third-party "best casinos" page that exists to bank on the medical domain's trust.
• Not inconsistent: wire and press-release services, syndicated news between publications, genuine user-generated content sites, editorial columns and opinion pieces, advertorial or native content whose purpose is reaching the publication's readers through the publication's own promotion, and affiliate links used appropriately throughout an otherwise independent page.

The carve-out with teeth is the advertorial one. The distinction is not who wrote the content or whether money changed hands — it is whether the content is being promoted to readers or hosted to harvest rankings. That is a behavioural test, and behaviour leaves records. A publisher that can show the newsletter send, the on-site placement, and the social promotion has evidence of a reader-facing purpose. A partner section that exists only in a sitemap and an internal-link block from the footer has evidence of the other thing.

Pre-flight for any host of third-party content

For publishers, marketplaces, local-service directories, and commerce sites that carry partner or syndicated sections, the audit is cheap before a manual action and expensive after one:

1. Inventory sections by authoring entity. Freelancers, white-label providers, user content, and syndication partners are all third-party under the policy's definition. Most hosts find they have more third-party surface than they assumed.
2. Apply the "own domain" test per section. If this content moved to its own domain tomorrow, would it plausibly rank for the terms it wins today? Where the honest answer is no, that section is the exposure.
3. Keep promotion records, not just contracts. Contracts establish the commercial relationship; promotion records establish reader-facing purpose, which is what the carve-out actually rewards.
4. Segment monitoring by authoring entity and searcher region. Blend first-party and third-party pages into one trend and a policy effect becomes invisible until finance asks about it.
5. Name the escalation owner before the notification arrives. Reconsideration and mediation both have timelines; ownership disputes inside the company usually cost more weeks than the process itself.

The practical conclusion

The EEA adjustment did not soften the site reputation policy — it split its enforcement into a direct penalty and a slow separation, with a forked dispute path behind them. The November 2025 fight with the Commission explains why; the 28 August announcement defines what. For operators, the actionable shift is that detection and remediation are now regional and section-scoped, and the only durable protection is third-party sections that would survive being ranked independently — because inside the EEA, that is no longer hypothetical: it is the enforcement.