Duplicate content across location pages: how similar is too similar

Location pages are never accused of being copies of other sites. They are accused of being copies of each other. That makes near-duplication measurable, and worth measuring before Google does.

Near-duplicate is the real category

Nobody publishes forty identical pages. They publish forty pages that are 90% identical: same structure, same sentences, a different town and the odd swapped adjective. Search engines fold near-duplicates together, pick one canonical-ish survivor, and quietly stop showing the rest. The estate does not get penalised so much as collapsed.

How to measure it, in plain words

Break each page’s visible text into overlapping five-word sequences, called shingles, and compare the sets between two pages: shared shingles divided by total shingles, a ratio known as Jaccard similarity. Two crucial refinements. Strip the swapped values first, because “boiler installation in Rye” and “boiler installation in Lewes” are the same five words pretending otherwise. And compare each page against its most similar sibling, not the average, because the worst pair is the one that looks like a doorway.

Thresholds that bite

In Townsmith’s scoring, a page keeps full similarity credit below 55% overlap with its closest sibling and none above 85%, on a straight line between. Those numbers are opinionated rather than official, but they match what reads as distinct: below half shared text, a page feels like its own; nearing nine-tenths shared, no reader would call them different pages.

Fixing a too-similar estate

• Do not rewrite the template into forty bespoke pages; you will never maintain it. Shared structure is fine.
• Add or deepen the local paragraph on each page you keep. Unshared substance is what moves the ratio.
• Vary genuinely variable passages (openings, FAQs) rather than spinning synonyms.
• Consolidate or delete pages for areas you cannot defend. Fewer, stronger pages beat a long thin tail.

Where Townsmith fits

Townsmith computes exactly this measurement on every save and shows it before publishing: “89% similar to Boiler Installation in Lewes” is a finding you can act on, not a feeling. The quality score documentation sets out the maths, the doorway pages guide covers the policy side, and a well-designed template stops the problem arriving at all.