Duplicate pages without confirmed consolidation

Why it matters

Finding repeated main content helps you decide whether each page has a useful purpose and which version you want visitors to use.

What to do

1. Read the actual pair in the finding and compare the two pages. Check whether visitors need both.
2. If the repetition is unintended, give each page useful distinct main content, or consolidate the duplicate into the appropriate representative with a canonical or permanent redirect.
3. Keep intentionally similar pages when they serve distinct needs. Do not add unrelated hreflang or pagination tags to silence the review.

When to leave it as it is: Keep distinct variants when they serve a real purpose. Inspect near-only pairs before changing content; a lexical match alone does not establish that the pages mean the same thing.

What we read

The exact comparison uses the complete normalized text extracted from captured HTML, including the title, menus and footer. It does not need an explicit main region, a language declaration or 50 tokens. Indexability, render state and original head relationships must be known; noindex pages and script-only shells are outside this comparison.

The near comparison retains its separate gates: complete server HTML, one explicit main region, confirmed indexability, a usable declared language and at least 50 tokens. Counted five-token sequences are compared by multiset Jaccard at the recorded threshold, 90% by default, within the same declared primary language.

Each finding belongs to one anchor URL and includes only its direct peers. Exact and near proofs are labelled per pair; similarity is not transitive. Canonical handling requires a read, measured, matching representative. An ordinary self-canonical does not handle a duplicate pair.

Exact pairs are excluded by proven reciprocal hreflang declarations matching both pages’ declared languages, reciprocal head prev/next links, or matching canonical representatives. An unproven alternate or canonical remains unknown; a one-sided pagination link excludes nothing.

Example & evidence

Illustrative example — one pair, not a live audit.

Two pages have the same complete extracted HTML text. Neither has proven matching canonical, reciprocal hreflang or reciprocal pagination handling. Each anchor receives a high-confidence issue with an exact whole-document identity pair. A near-only pair at the recorded threshold instead has medium confidence.

Confirm & check again

Check for yourself: Open the two addresses and compare their main content. Check their original head declarations and whether any canonical representative is accessible, indexable and actually matches the content.

Check again: Full audit — Run a complete audit of the same anchor page. A fix requires no counted exact or near pair and sufficient checks in both comparisons. Changed text or proven canonical, reciprocal hreflang or pagination handling can remove a pair. An unread page, missing identity, incomplete comparison or unavailable near preparation cannot prove a fix.

Limits & sources

Text identity concerns extracted server HTML, not CSS visibility, rendered appearance, meaning or rankings. Scripts are not executed. Partial render dependence is disclosed even when an exact pair gives high confidence. Unmeasurable near-only peers remain in coverage but are omitted from the positive pair list and disclosed separately. A peer already proven by an exact pair stays listed; its unproven near comparison still leaves coverage incomplete.

An exact positive can still be found on a short page or a page without a unique main region or usable language. Without near preparation, an empty exact comparison is unknown coverage, never a pass.

The exact comparison allows at most 10,000 candidate pages and materializes at most 25,000 same-hash pairs, including excluded and unknown pairs. Incomplete page coverage or a budget limit preserves positive witnesses but prevents every exact clean and fix.

Confidence: Medium Any counted exact pair gives the anchor finding high confidence. Near-only findings have medium confidence; near partners in a mixed finding remain lexical review evidence.

Sources

• Ahrefs, Duplicate pages without canonical in Site Audit — Background on duplicate pages and canonical consolidation; the audit defines its own pair proofs and confidence. Checked 7 October 2026.
• IANA, Language subtag registry — The pinned registry validates the original declared HTML language tag; it does not detect the language of the text. Checked 3 October 2026.
• IANA, Language tag extensions registry — The pinned registry validates the original declared HTML language tag; it does not detect the language of the text. Checked 3 October 2026.

Editorial review: 2 November 2026