Cross-Page Duplication: When Different Pages Say the Same Thing

Cross-Page Duplication detects content repetition between different pages on your site. While the Duplicate Content Blocks criterion checks for repetition within a single page, this criterion checks whether different URLs share substantially similar paragraphs.

The scorer collects paragraphs from your homepage, key pages, and a sample of blog posts, then compares them pairwise using shingle-based Jaccard similarity. Any paragraph pair from different pages that exceeds the similarity threshold is flagged. The key innovation is boilerplate detection: paragraphs that appear on more than 40% of pages (by shingle fingerprint) are classified as template elements (CTAs, author bios, footer content, disclaimer blocks) and excluded from the comparison. This means your standard "Subscribe to our newsletter" footer block does not trigger a penalty.

The remaining non-boilerplate duplicates represent genuine content reuse - paragraphs where two different pages make the same claims in the same language. This is a problem because it creates content cannibalization at the most granular level. AI engines encountering the same paragraph on two different URLs have to decide which one to cite, and the confusion often results in neither being selected.

The cross-page duplication scorer operates at the site level (not the page level) and follows these steps:

1. Collect pages: Gather the homepage, key pages (about, services, products), and a sample of blog posts (typically the 5-10 most recently published).

2. Extract paragraphs: From each page, extract meaningful text paragraphs (excluding navigation, headers, footers detected as boilerplate).

3. Detect boilerplate: Calculate a fingerprint for each paragraph using its first 5 shingles. Paragraphs whose fingerprint appears on more than 40% of pages are classified as boilerplate and excluded.

4. Pairwise comparison: Compare all non-boilerplate paragraphs from different pages using shingle Jaccard similarity. Pairs exceeding the threshold are flagged as duplicates.

5. Score calculation: Based on the number of unique page pairs with duplicated content and the severity of duplication, calculate a 0-10 score. Zero or one duplicate pairs with low severity scores 8-10. Three to five pairs scores 4-6. Six or more pairs with high similarity scores 0-3.

If fewer than two pages are available for comparison (e.g., a single-page site), the scorer returns 5/10 with a "not enough pages" finding.

Step 1: Review the audit evidence

The audit report shows each duplicate pair with: the two URLs involved, the duplicate paragraphs, and the similarity percentage. Start with the highest-similarity pairs.

Step 2: Differentiate overlapping pages

If two blog posts cover similar topics, differentiate them by angle, depth, or audience: - "Live Chat for E-Commerce" vs "Live Chat for SaaS" - same tool, different audience - "Live Chat Pricing Guide" vs "Live Chat Feature Comparison" - same tools, different focus - "Beginner's Guide to Live Chat" vs "Advanced Live Chat Strategies" - same topic, different depth

Step 3: Rewrite shared introductions

The most common cross-page duplicate is the introductory paragraph. Many sites use the same "context-setting" opener across related posts. Write a unique opening for each post that ties directly to its specific angle.

Step 4: Link instead of repeating

If two pages need to reference the same concept, explain it fully on one page and link to it from the other:

<!-- Instead of repeating the explanation -->
<p>For a detailed breakdown of live chat pricing tiers,
see our <a href="/blog/live-chat-pricing">pricing guide</a>.</p>

Step 5: Audit template-injected content

Check whether your CMS injects the same content blocks (beyond CTAs) into multiple pages. If a "Why Choose Us" block appears in the body of 5 different pages, that is cross-page duplication that the boilerplate detector may not catch if it appears on fewer than 40% of total pages.

ChatGPT encounters cross-page duplication when it retrieves multiple pages from the same domain. When two pages contain the same paragraph, ChatGPT has to decide which page is the canonical source for that information. If neither page provides unique context around the shared paragraph, ChatGPT may cite neither and instead find a more differentiated source from another domain.

Claude builds entity models across multiple pages from the same site. When Claude detects substantial content overlap between pages, it reduces the overall trust score for the domain because repeated content suggests thin editorial process. Claude also uses cross-page uniqueness as a signal for topical authority - a site where every page makes distinct claims demonstrates broader expertise than a site where the same claims appear on multiple pages.

Perplexity assembles answers from passages across sources. Cross-page duplication means Perplexity encounters the same passage twice from the same domain, which wastes its source diversity budget. Perplexity prefers domains where each page contributes unique information to its answer.

Google AI Overviews deduplicates at the URL level before selecting sources. Pages that substantially duplicate content from other pages on the same site are less likely to be selected because Google has already captured that information from the canonical page.