Translate 137+ languages including Hausa, Yoruba, Igbo, Urdu, Bengali, Swahil…
TexAISome warnings 2 things we found but could not confirm. Nothing confirmed. Checked: 400 reviews and the Android app’s code.
From our automated analysis, 25 Oct 2026. Every label links to its evidence in the full report below.
This analysis describes version 5.3.2.005.xm of the app (the version number recorded inside the package we inspected), analyzed 2026-10-25 .
The store currently lists version 6.3.0.001.gp, updated 2026-10-19.
Store screenshots supplied by the developer.
Translate 137+ languages including Hausa, Yoruba, Igbo, Urdu, Bengali, Swahili, Tagalog, Thai, Vietnamese, Arabic, Persian, Hindi and more — including African and South Asian languages other apps don't support.
Trusted by 500M+ users worldwide. Ranked #1 in WMT25 Low-Resource Indic Language Translation by ACL — the world's most prestigious machine translation competition.
The only AI translator built for global languages, not just English and Spanish.
📱 FLOATING SCREEN TRANSLATOR
Instantly translate WhatsApp, Instagram, Messenger and any on-screen text without switching apps. Drag the floating icon anywhere for real-time translation — perfect for chatting, browsing, or traveling abroad.
📞 CALL TRANSLATOR (Real-Time)
Translate phone calls instantly with AI.
· Talk to international clients with ease
· Handle overseas business calls
· Built for remote work and global teams…
From the developer’s store description.
For developers and the curious
The evidence, sources, measurements and versions behind every label above.
Specific evidence of a useful property, within the stated scope.
No advertising or mediation SDK among the 5 SDKs detected in the package
No advertising or mediation SDK of any kind was found in the package.
Package artefact
Appworthy
sha256:fdfd37ac42357b27c323fc8951e81c8d0223856c0ffe8dac79e46f4b0427060bThe developer's store data-safety disclosure states that data deletion can be requested; we have not seen a deletion control in the app's own text
The developer's data-safety disclosure on the store states that users can request deletion of their data, or the app's own text offers a way to delete the account. This restates what the developer declares; it is not a test of whether deletion works. Apple publishes no equivalent disclosure, so on an Apple-only listing only the app's own text can support it.
Open observations that have not cleared the evidence bar.
Progress itself is for sale: "Get more coins"
This observation has not cleared the required evidence standard.
Advancement or competition is gated behind spending money rather than playing.
Package artefact
sha256:fdfd37ac42357b27c323fc8951e81c8d0223856c0ffe8dac79e46f4b0427060bThe app's own text says: "Out of coins. Buy more to keep using captions"
This observation has not cleared the required evidence standard.
The core of the app — not extras — sits behind payment, or previously-free function moved behind a paywall.
Package artefact
sha256:fdfd37ac42357b27c323fc8951e81c8d0223856c0ffe8dac79e46f4b0427060bThe app's own text says: "Continue as guest"
This observation has not cleared the required evidence standard.
The app's own screens offer a way past sign-up — guest access, skipping, or browsing before registering — rather than gating the product behind an account. The store's privacy disclosure is checked alongside it: that disclosure is self-reported, so it corroborates this credit rather than establishing it.
Package artefact
sha256:fdfd37ac42357b27c323fc8951e81c8d0223856c0ffe8dac79e46f4b0427060bThe app's own text says: "Enjoy stable offline translation and save data, even with no network"
This observation has not cleared the required evidence standard.
The app's own copy describes working without a connection — offline mode, downloading for offline use, or content saved for later. Shown because readers filter on it; it is a feature, not a fairness property, and never counts for or against an app.
Package artefact
sha256:fdfd37ac42357b27c323fc8951e81c8d0223856c0ffe8dac79e46f4b0427060bNothing was published as a thing to watch for in this assessment. That is not a finding that nothing is wrong with the app.
No review-based observation is published in this assessment. That does not establish that reviewers had no problems.
“Not assessed” means we could not measure it. It is never a zero and never a pass.
Android package 5.3.2.005.xm, scanned 25 Oct 2026. These describe that package, not the current store version.
Supporting measurements: not a complete list of possible harms or positives. Measurement vocabulary 1.4 · how these are measured
This analysis describes version 5.3.2.005.xm of the app (the version number recorded inside the package we inspected), analyzed 2026-10-25 .
The store currently lists version 6.3.0.001.gp, updated 2026-10-19.
Each store’s release notes, as the developer wrote them. They are the developer’s words, not ours.
No release notes were provided for this version.
hi-translateAssessed 2026-10-25 from 400 user reviews and the app's package.
What the stores publish. Store ratings are the stores’ numbers, not Appworthy’s, and are never averaged across stores.
This assessment is produced automatically. Four independent signals are read — user reviews, SDKs in the app package, the app's own on-screen text and the app manifest — and each candidate finding is filed under the sources that support it.
No charge stands on one kind of evidence — with one published exception. A finding is either a measured fact that anyone can reproduce (an ad SDK is in the package; every subscription bills weekly; the privacy label omits a tracker we can see), or it is an interpretation — and an interpretation needs independent sources to agree before it reaches the catch list. The exception is a harm only users can see: a charge may publish from reviews alone, as “Reported by N users”, once enough reviewers raise it to clear that charge’s own published bar. That agreement rule is what replaced the human reviewer, not a model’s confidence.
There is no overall score, on purpose. A 0–99 number was published here until 2026-09-15. It was withdrawn because it discriminated too weakly to be the instrument — 44.7% of every pair of apps tied — and because a single number is a verdict where a count is an observation. What publishes instead is a row of measurements, each of which says what was counted and, where we could not look, says that instead.
Findings under One source only are real reports we do not stand behind as findings. They are shown because withholding them entirely would hide a genuine complaint, and labelled because publishing them as charges would be an accusation off a single source.
Write to [email protected], or use the form. Name the finding, the app version and the evidence that needs correcting. Factual corrections are free and never depend on buying anything.