Does sayit work well for an Arabic accent?

sayit works for an Arabic accent through the same mechanism that makes it useful across first languages generally: it scores the phonemes you actually produced against the target sounds, rather than a transcript that autocorrects past the error. That matters specifically for Arabic-influenced English because standard Arabic has no /p/ phoneme, so "p" and "b" are a genuinely common confusion ("pig" and "big" landing the same way) — the kind of substitution a word-guessing model tends to smooth over but a phoneme-vs-target-IPA comparison is built to name directly.

30-second version: Arabic isn't one of the four accent groups sayit names as explicitly tested (Indian, American, British, European), so this is a mechanism-based answer, not a published benchmark. What's documented is that sayit measures your actual phonemes against target IPA — a better fit than word-recognition for catching the p/b merge and the narrower set of vowel qualities that English distinguishes and Arabic doesn't mark the same way.

Why the mechanism is the honest answer here

English has more distinct vowel sounds than Modern Standard Arabic marks in its own vowel system, so contrasts like "ship" versus "sheep," or "bit" versus "bet," can collapse for Arabic-influenced English in a way that a lenient, context-guessing speech recognizer usually won't catch — it hears a plausible word and moves on. sayit's scoring instead compares the vowel you produced to the target phoneme directly, which is what surfaces a genuine vowel-quality gap instead of letting a forgiving transcript absorb it.

Where this is genuinely strong

• Per-word, per-phoneme feedback with target IPA next to your actual sound makes a p/b substitution or a vowel-quality gap visible rather than silently passed.
• Minimal-pair drills (ship versus sheep is a documented example) are exactly the exercise type for tightening a vowel contrast, and they're free on every tier.
• IELTS and TOEFL exam-format practice (Pro/Max) is relevant if you're studying toward either exam.

Where to keep expectations honest

• Arabic isn't among the four accent groups sayit publicly names as tested — treat this as an argument from how the engine works, not a published Arabic-specific benchmark.
• Consonant clusters and emphatic consonants vary a lot by Arabic dialect, and sayit doesn't publish dialect-specific guidance — its feedback is general phoneme scoring, not tuned per Arabic variety.
• It scores pronunciation and fluency only, with no grammar or vocabulary checking.

A concrete way to test this yourself

Rather than taking any of this on faith, the fastest way to check the fit is to record a small set of minimal pairs that specifically target the p/b contrast — "pig" and "big," "pack" and "back," "cap" and "cab" — and look at whether the feedback correctly separates them. Because the scoring works on your actual phonemes rather than a guessed transcript, it should treat these as genuinely different sounds rather than letting context paper over the distinction the way a forgiving speech-recognition model might.

Who this is for

If the p/b distinction or a specific vowel contrast is the thing you're trying to fix, per-phoneme feedback with target IPA and a minimal-pair drill built around exactly that contrast is a direct fit, and both are free features in sayit.

Try it

Record a word pair like "pig" and "big," or "ship" and "sheep," in the app — no card needed for the first take — and see the verdict.