How Google Detects Site Networks

You build four sites from the same codebase. Different logos, different color schemes, different content. You deploy them on separate hosting accounts. Traffic starts, pages get indexed, and then — quietly, without any warning in Search Console — everything stalls. Pages sit in "Discovered — currently not indexed" limbo. Crawl frequency drops. Rankings plateau or regress.

No manual action. No penalty notification. Just algorithmic silence.

What happened is that Google looked at your four domains and saw one website wearing four masks. Not because of the colors or the logos. Because of the DOM structure underneath.

This is template fingerprinting, and understanding how it works is the difference between a network that scales and a network that gets clustered.

What Google Actually Measures

Google holds multiple patents related to boilerplate detection. The core mechanism works like this: when Googlebot renders a page, it separates template markup from unique content. Headers, footers, navigation elements, sidebars, cookie banners — that is boilerplate. The article text, product descriptions, and unique data — that is content.

The critical part for network operators: Google does not just measure boilerplate ratio on a single site. It measures boilerplate similarity across domains.

When Googlebot crawls healthyjoints.com and finds the exact same DOM tree structure as flexibilitytips.com — same <nav> nesting pattern, same <footer> with the same number of <ul> elements containing the same number of <li> items, same <aside> widget arrangement, same <main> wrapper hierarchy — it does not matter that one is blue and the other is green. The structural fingerprint is identical.

The Consequences of Clustering

When Google detects high structural similarity across multiple domains, the effects cascade:

• Crawl budget throttling: Each domain gets crawled less frequently. Not deindexed — just deprioritized.
• Index consolidation: Google may index only one version of structurally similar pages, suppressing duplicates.
• Quality scoring depression: The entire cluster gets evaluated as a single entity. The weakest site drags down the strongest.
• Discovery suppression: New pages on clustered domains take dramatically longer to get indexed.

None of this appears as a manual action. There is no message to appeal. The throttling is algorithmic, silent, and persistent.

The Signals Google Uses

1. DOM Structure Analysis

This is the primary signal. Google's renderer builds a DOM tree for every page it crawls. It can compute tree-edit distance — a mathematical measure of how many insertions, deletions, and modifications are needed to transform one DOM tree into another. Sites with low tree-edit distance across domains get flagged as structurally similar.

What triggers it:

• Identical nesting depth in navigation elements
• Same number of footer columns with the same number of links
• Matching sidebar widget order and structure
• Identical <main> content wrapper hierarchy

What does not matter:

• CSS class names (Google renders the page; class names are irrelevant to visual structure)
• Color values
• Font choices
• Image sources

2. Boilerplate Ratio

Google computes the ratio of unique content to shared template markup on every page. A healthy article page should have unique content comprising at least 60% of the visible text. If your navigation, footer, sidebar widgets, and repeated elements account for more than 40% of the visible content, you are in the danger zone.

Across a network, the boilerplate ratio becomes a fingerprint. If every site in your network has a 35% boilerplate ratio with structurally identical boilerplate, that is a strong clustering signal.

3. SpamBrain Pattern Detection

SpamBrain is Google's AI-based spam detection system, introduced in 2022 and significantly upgraded since. It does not look at individual signals in isolation. It looks at patterns across signals.

SpamBrain can detect:

• Multiple domains registering within a short time window
• Domains using the same Google Analytics or Tag Manager accounts
• Sites sharing identical robots.txt patterns
• Domains hosted on the same IP ranges with the same deployment timestamps
• Content publication patterns that are suspiciously synchronized

SpamBrain is probabilistic. No single signal triggers action. But the accumulation of signals — same template structure, same hosting fingerprint, same registration timeline, same content cadence — builds a probability score that eventually crosses a threshold.

4. HTML Source Patterns

Beyond DOM structure, Google can analyze the HTML source for telltale patterns:

• Identical HTML comments (build tool signatures, template engine markers)
• Same meta tag ordering
• Identical schema markup structures (same optional fields included in the same order)
• Matching CSS and JavaScript file naming conventions
• Same favicon dimensions and format

The 60/70 Rule

A useful heuristic for network differentiation: at least 60% of your visible text should be unique content (not template boilerplate), and at least 70% of your DOM structure should differ between any two sites in the network.

The 60% content ratio keeps individual pages looking like genuine content pages rather than template-heavy thin pages.

The 70% structural difference ensures that DOM tree comparison algorithms do not flag your sites as variations of the same template.

The 15-Point Differentiation Checklist

Before launching any site in your network, audit it against these fifteen checks. At least twelve must pass.

Scoring: 12-15 = deploy. 9-11 = tweak before launch. Below 9 = redesign the template variant.

How to Differentiate from One Codebase

The challenge is obvious: if you are running a monoclone architecture (one codebase, many sites), how do you produce genuinely different DOM structures?

The answer is multiple base layouts within the same codebase. Not CSS variations — structural alternatives.

You need three to four fundamentally different layout templates:

• Layout A: Top navigation bar, full-width hero, single-column content, three-column footer grid
• Layout B: Left sidebar navigation, two-column content area, single-row footer bar
• Layout C: Minimal top bar (logo + search only), card-grid homepage, article pages with floating table of contents
• Layout D: Magazine-style multi-column homepage, hamburger mobile-first navigation, full-width article pages with pull quotes

Each layout produces a genuinely different DOM tree. Assign each site in your network to a layout via sites.json, and the build system selects the correct template set automatically.

CSS class names should use a per-site namespace prefix to prevent cross-site pattern matching in the source code. Instead of .nav-item, use .bpw-nav-item for BestPressureWashers and .lcd-nav-item for LawnCareDaily. This is a small detail, but source-level analysis is one of the signals Google evaluates.

What Not to Worry About

Some network operators over-optimize for stealth and end up wasting time on signals Google does not meaningfully use:

• Different hosting providers per site: Useful for other reasons, but IP diversity alone is not a strong clustering signal. Shared hosting puts thousands of unrelated sites on the same IP.
• Different registrars: Google does not have access to WHOIS data at the crawling level in a meaningful way. Use whoever offers the best price.
• Unique content management systems: The CMS is invisible to the crawler. It sees rendered HTML. Whether that HTML was generated by Eleventy, Hugo, or WordPress is not detectable from the output alone (unless the CMS leaves signatures in the HTML, which static site generators do not).

Focus your differentiation effort on what Google actually renders and compares: DOM structure, boilerplate ratio, and visual layout.

Going Deeper

This post covers the detection mechanisms and the audit checklist. The full implementation guide — including the four base layout templates, the CSS namespacing system, the automated boilerplate ratio calculator, and the pre-launch audit script that scores your site against all fifteen checks — is in The $100 Network by J.A. Watte. Chapter 3 is devoted entirely to defeating template fingerprinting, and Appendix B provides the complete checklist in a format you can run through before every launch.

If you are running multiple sites from a shared codebase and have not audited for structural similarity, do it today. The clustering may already be happening.

The SEO foundations for these techniques are covered in The $20 Agency, Chapters 3-5. This article builds on those basics with advanced multi-site strategies from The $100 Network.