How AccentMatch.US is made
Every page on this site is built from a research bundle that a separate pipeline rebuilds nightly: one record per sound, per rhythm feature, per resting-mouth setting, per exercise and per daily tip, for each native language. Bundle date: 2 Oct 2026.
How this material is made
A research pipeline (the research harness) researches each language pair nightly with a language model. For every record it keeps a source list, validates the data against a schema, and exports the bundle this site is built from. The schema checks shape, not truth.
Each record carries an evidence badge. Sourced: at least one reference attached and used. Consensus: agrees with standard references; no single citation attached. Generated: ai-inferred; no supporting reference attached. A badge describes the support behind a record; it does not mean a human checked it.
None of it has been reviewed by a phonetician or a speech-language pathologist. Illustrations are generated by an AI image model and reviewed by an AI vision model; some may still be anatomically wrong.
The AI research runs continuously, so this material changes over time, hopefully for the better: an article you read today may be revised next month, and each article shows when it was last updated. This is material that evolves with its readers: reader ratings and feedback are coming soon and will help us identify improvements for future update rounds.
Evidence levels
- sourced At least one reference attached and used. (300 of 894 records)
- consensus Agrees with standard references; no single citation attached. (533 of 894 records)
- generated AI-inferred; no supporting reference attached. (61 of 894 records)
Scope
The target accent is General American English. Native languages are added in a fixed order; a pair that is not yet covered says so on its page. Learner text is written for adult intermediate learners. There is no audio yet.
Your feedback and your data
Reader ratings, comments and language wishes are coming soon. No registration or feedback data is collected in this release; hosting logs are described in the privacy notice. Privacy notice.
Source code
The research pipeline is the research harness project; this site's source is accent-match-us.
Maintainer pages: review.