Template Diversity — Why HTML Structure Counts as a Spam Signal

What it detects

30% is the default minUniqueRatio pseolint warns below — the floor at which Google's SpamBrain (rebuilt August 25, 2022 alongside the Helpful Content System launch to score site-level helpfulness alongside per-page signals) starts treating a domain as a single template rather than N designed pages. Each parsed page carries a `structureSignature` — a hash of its HTML structure that ignores text content but preserves the sequence and nesting of element types. The rule counts how many distinct signatures exist across all pages and divides by the page count to produce a unique-ratio in [0,1]. If that ratio falls below `minUniqueRatio` (0.30 default), a single warning-severity finding (weight 12) is emitted at the site level — versus error=25, critical=40, info=5 elsewhere in the engine. This is a holistic signal, not a per-page one: there is no list of 'failing' URLs because the problem is the site's design system, not any individual page. Powered by @pseolint/core v0.4.3, MIT-licensed at github.com/ouranos-labs/pseolint.

Why it matters

Templated HTML is not in itself a spam signal — every modern CMS produces it. The signal is when templated HTML combines with templated content. SpamBrain (publicly named April 12, 2021 and rebuilt across the August 25, 2022 Helpful Content System launch) reads the combination as 'one piece of low-effort programmatic output,' even if the underlying data is rich, because there is no surface variation for the classifier to latch onto. Field reports following the March 5, 2024 scaled-content-abuse update show 60% to 80% organic-traffic loss within a 6-week window for sites whose unique-ratio sat below 10%, and a 90-day recovery window once the structure was diversified. Industry crawlers like Ahrefs, Sitebulb, and Screaming Frog all surface comparable template-fingerprint counters, but the 30% floor is specific to pseolint's measurement (powered by @pseolint/core v0.4.3). Sites with diverse structure (some pages have a comparison table, some don't; some have a video embed, some don't; some have a sticky TOC, some don't) communicate to the classifier that real per-page editorial decisions were made. Sites with one signature for every URL communicate the opposite. The fix is to introduce conditional structure, not to randomise it artificially. The current implementation lives in @pseolint/core v0.4.3 with site-type-aware weighting — programmatic-directories tolerate slightly higher template homogeneity than small-marketing sites.

A page that fails

A page that passes

How to fix it

1. 1Identify which sections in your template should be optional. Anything that doesn't apply to every entity is a candidate: pricing tables, video embeds, timelines, FAQs, comparison widgets.
2. 2Wrap optional sections in conditionals that key off the underlying data, not random booleans. 'If the entity has a video URL, render the video block' produces meaningful diversity; 'if Math.random() > 0.5' produces nothing.
3. 3Vary the order of secondary sections by entity type. A restaurant page might lead with menu, a hotel page with rooms — same template, different priority.
4. 4Add per-entity media variations. Some pages have hero images, some have hero videos, some have galleries. Each renders different HTML.
5. 5Don't fix this rule by adding random structural noise. The rule is a holistic warning; if the underlying content is differentiated, the warning is acceptable on a homogeneous content type.

SpamBrain context

Structural homogeneity has been a feature in spam classifiers since at least the 2009 'doorway page' updates, but it took on new prominence after the August 25, 2022 Helpful Content System rollout introduced site-level helpfulness scoring. Sites that look the same on every URL communicate 'mass production' to a classifier whose entire job is to find mass production. Google's December 14, 2022 link-spam update mentioned 'sites that exist primarily to feed link signals' — those sites are almost always structurally homogeneous because they were built off a single template with no per-page editorial input. The March 5, 2024 scaled-content-abuse policy formalised this: 'producing many pages with little unique value' is structurally measurable, and the May 7, 2024 site-reputation-abuse update extended the same logic to hosted third-party content. While alternative fingerprinting approaches like SimHash (used by spam/near-duplicate at the 85% threshold), Jaccard set similarity, and BERT structural embeddings exist, the structureSignature hash pseolint uses keeps the rule deterministic and runnable in under a 1-second wall-clock per 100 pages — well within the 60-second free-tier audit budget. Implementation: @pseolint/core v0.4.3 (current), MIT-licensed, github.com/ouranos-labs/pseolint. Comparable template-diversity counters surface in Ahrefs, Sitebulb, and Screaming Frog, though each defines the floor slightly differently.

Frequently asked questions

Isn't every WordPress site structurally identical?

Not really. Default WordPress themes produce slightly different HTML for posts, pages, archives, single-product, and category templates — usually 5-8 distinct signatures across a typical install. The rule will fire on heavily-templated WordPress builds (especially those using a single page template for every URL) but not on default editorial sites.

Why is this only a warning, not an error?

Because structural homogeneity is acceptable for some content types — a glossary, a directory of API endpoints, a product catalogue. The rule surfaces the signal so you can make an informed call; it doesn't assume the call. If your content type genuinely demands one template, document the decision and ignore the warning.

Does the structureSignature ignore CSS classes?

Yes. Class names and attribute values are stripped before hashing — only element types and their nesting pattern are considered. This means restyling a page (CSS changes only) doesn't change its signature, which is the right behaviour for a structural signal.

How does this differ from boilerplate ratio?

Boilerplate ratio measures shared text content. Template diversity measures shared HTML structure. A page can have low boilerplate (every page has unique paragraphs) but identical structure (every page renders those paragraphs in the same shell). Both rules need to be green for a site to look genuinely diverse.

What's a healthy unique-ratio target?

0.30 is the default minimum and the floor at which most pSEO sites stop reading as templated. 0.50+ feels like an editorial site to a classifier. Below 0.10 is almost always a single template — fine for some content types, dangerous for others.

Related rules

• spam/boilerplate-ratioBoilerplate Ratio60% is the default boilerplateMaxRatio: pseolint identifies sentence-level blocks appearing on 80%+ of pages, then flags any URL whose word count is dominated by those repeated blocks (warning severity, weight 12).Read →
• spam/doorway-patternDoorway PagesGoogle has banned doorway pages since the March 16, 2015 Search Central post — pseolint's spam/doorway-pattern rule mirrors SpamBrain's convergence logic by requiring 3 independent signals to stack (SimHash near-duplicate above 0.85, entity-swap, and structural confirmation) before firing at error severity (weight 25), the highest-confidence spam pattern reported by @pseolint/core v0.4.3.Read →
• spam/near-duplicateNear-Duplicate Pages85% SimHash similarity is the pseolint default threshold — every page pair at or above that mirrors the near-duplicate canonicalisation ceiling Google's web indexing team has used since adopting Charikar's 2002 SimHash paper in 2007, and which the March 5, 2024 scaled-content-abuse update reaffirmed as policy via SpamBrain's 60-second triage queue.Read →