Can two strings compare equal for sorting but differ exactly?

Measured inputs and outputs

These are locally constructed test strings, not evidence that Claude inserts these characters. We executed the saved homepage script snapshot with a minimal DOM harness and compared exact strings. The invisible-checkbox state and dash mode appear in each row. The observations below test this page's specific question. We make no detector-score or statistical-watermark removal claim.

Two different equivalence decisions

The observer compares each string against AB twice: strict equality checks the sequence, while Intl.Collator compares according to its configured locale rules. In this runtime, the selected-marker and retained-direction-mark fixtures both compare as collation zero before cleaning while strict equality is false. Cleanup makes the selected-marker fixture strictly equal to AB. The direction-mark fixture remains strictly unequal because U+200E is retained. The visible reference is equal by both methods throughout. Each row includes the resolved collator settings rather than silently assuming them.

Zero is a sorting result, not an identical sequence

Intl.Collator can treat some characters as ignorable under a locale implementation. That is a distinct question from whether two strings contain the same code points. We did not test a database collation, operating-system file ordering, another locale or another ICU version. The runtime and resolved options in results.json delimit the observation. The homepage does not call a collator; it deletes selected code points. A visually identical sort grouping can therefore coexist with a failed exact match in a different stage of a workflow.

Record the comparator your destination uses

When a sorted list seems correct but a lookup fails, compare exact strings and record the sorting comparator separately. Do not deduplicate records just because a locale comparator returns zero. Confirm the identifier policy and inspect original code points before applying a targeted edit. If the destination defines its own collation, test that actual implementation. This experiment explains one discrepancy and preserves its reproduction data; it does not prescribe a universal equality policy or certify that a sorted collection contains unique identifiers.

Reproduce this test

Save reproduce.cjs and tested-app.js in the same folder. Run the command below with Node.js. The harness prints its runtime, script SHA-256 and every measured row. Compare those rows with the original record. Using a newer script or runtime creates a new experiment; retain the version information with your rerun.

node reproduce.cjs