Scaled content abuse and location pages: what crosses Google’s line after the 2026 updates
In March 2026 Google ran a spam update and a core update within days of each other, and named scaled content abuse as a priority. Here is where the line actually falls for a location-page estate, and how to recover if you are on the wrong side of it.
What changed in March 2026
Two separate events ran days apart. The March 2026 spam update completed first, in under twenty hours, the quickest spam update Google has confirmed; the March 2026 core update started afterwards and finished in early April. They are different mechanisms judged on different signals, so it matters which one moved your pages. A short, abrupt drop that lines up with the spam update points at a policy violation Google’s systems flagged directly. A slower drift over the core update’s longer rollout points at a broad reassessment of how helpful your pages are, page by page, against everything else competing for the query.
Neither update changed the underlying rule. The scaled content abuse policy has read the same way since March 2024: creating many pages whose primary purpose is manipulating search rankings rather than helping people. What the 2026 updates changed is enforcement weight. The thing that was always against the rules is now one of the first things being acted on.
What actually crosses the line
The policy is about purpose and value, not tooling or count. Google’s examples of scaled content abuse are pages generated to fill results slots rather than answer a search, scraped or lightly reworded copy that adds nothing, and pages stitched together from fragments. The method is explicitly irrelevant: a page is judged the same whether a person, a template or a model produced it. So the question is never “did I use automation” but “does each page carry something true and specific to its place”.
For a location-page estate, the failing pattern is substitutability. If two of your town pages read as the same page with the place name swapped, they meet both the scaled-content definition and the older doorway definition at once. The swap-the-town test is the fastest way to see it: change the town name on one page and, if nothing else has to change, the page was never about that town. The line is crossed not by publishing many pages, but by publishing many pages that are interchangeable.
What does not cross the line
The policy does not ban location pages, does not set a page count that is too high, and does not penalise AI assistance on a page that is genuinely useful. Ten thin pages are inside its target; two hundred pages that each carry real local detail are not. Programmatic generation is fine when there is real data underneath it – the difference between a defensible programmatic local estate and a doorway estate is whether the template is filled with substance or with synonyms. Publication speed is not named in the policy either, though a large batch of thin pages appearing at once is exactly the footprint the policy describes, so tranches beat one big drop.
How to recover a flagged estate
Recovery is not a single switch. The work is to move every page out of the substitutable bucket, by giving it something to say or by taking it down. These five steps are the order that wastes the least effort.
- Triage every page into keep, rewrite or retire. A keep page already answers a real local search with detail no sibling has. A rewrite page is about a place you genuinely serve but currently says nothing specific. A retire page is one you cannot honestly distinguish, even with effort – the area you barely cover, or the page that exists only because the spreadsheet had a row for it.
- Consolidate the duplicates first. Before writing anything, fold near-identical pages together. Two pages that share most of their sentences should become one stronger page with a redirect, not two pages you try to differentiate after the fact. This is the step that removes the footprint fastest, because it cuts the count of interchangeable URLs directly.
- Rewrite the keepers around local substance. Replace swapped-name filler with things that are only true of that place: jobs you have done there, local conditions you account for, access or regulation specific to the area, the parts of the service that change with location. Use your own photos from local jobs, not stock – team shots, completed work in the area, process photos – with honest captions. What counts as real local evidence is the same standard as experience and trust for a local business: first-hand detail a competitor cannot copy.
- Hold back the pages with nothing to say yet. Retiring a page is not a defeat. A page held back until you have real work or knowledge in that area is a page that will pass when it returns. Publishing it empty now is the exact behaviour the policy targets. Bring it back as a defensible tranche when the local substance exists.
- Measure before and after, and keep measuring. Recovery is judged on the same signals the update used, so read them directly: impressions, clickless impressions, query mismatch and page-two stalls in Search Console tell you which surviving pages earned their place and which are still drifting. Re-run the similarity check after consolidation so you can prove the estate is no longer interchangeable.
Is this the end of location pages?
No, and the framing matters because it changes what you build next. The pages that lost traffic in March 2026 were interchangeable, not local; the pages that held were specific. That is the same distinction the policy has always drawn, now enforced harder. The honest reading is not that the format is dead but that the cheap version of it is – the longer argument is in whether programmatic SEO is dead. A small business with genuine reach into a handful of areas can still earn every one of those pages; it just has to earn them.
Did the March 2026 spam update target AI-generated pages specifically?
It targeted scaled content abuse, which is method-agnostic. AI-generated pages were hit heavily because mass-generating thin pages is easy with a model, but the trigger is value, not the tool. A useful page assisted by AI is not a violation; a thousand worthless hand-typed pages are.
How do I tell whether the spam update or the core update hit me?
Line your traffic drop up against the dates. A sharp fall around the spam update’s short rollout points at a policy violation flagged directly. A gradual slide across the core update’s longer rollout points at a broad helpfulness reassessment. The two call for different work, so it is worth knowing which moved you before you start fixing.
Will deleting pages hurt more than keeping thin ones?
Retiring a page that adds nothing removes part of the footprint the policy describes, which helps the estate as a whole. Keeping a page only to preserve a URL count keeps the liability. Redirect the duplicates you fold together, take down the ones you cannot honestly distinguish, and bring them back when there is real local substance to publish.
Where Townsmith fits
Townsmith builds the policy’s questions into the workflow as maths rather than hindsight. Every page is scored against its closest sibling before it goes live, and the score is documented so the credit is withheld until the page carries local substance. Twin Finder highlights the sentences a page shares with its closest sibling, the duplicated copy that marks a page as a candidate to fold together in the consolidate step above. The same tooling that keeps a new estate defensible is what reads a flagged one back into shape.