A score on the phone, on the commute, before the upload. The detector runs in three honest shapes: the responsive web app at app.textsight.ai works in iOS Safari, iOS Chrome, Android Chrome, and Samsung Internet right now; the Android native app is live on Google Play (v1.0 on Internal testing, v1.0.2 polish build ready as a signed AAB) with paste, share-target, scan, three modes, and history sync against the same backend the web uses; the iOS native app is not shipped yet and is on the 2026 roadmap with no waitlist and no email gate. This page covers what the detector does on a phone today and what is honestly not ready.
One Authenticity Score, one quota, one account no matter which surface a user opens. The three mobile entry points are different shells around the same detection engine.
The web app at app.textsight.ai is the same React build that runs on desktop. The layout collapses cleanly under 768px, the editor takes the full width, the score gauge stacks above sentence highlights, and the result panel sits under the input so reading on a 6-inch screen is comfortable. iOS Safari, iOS Chrome, Android Chrome, Samsung Internet, and Firefox for Android all render the same UI. Paste from the standard iOS or Gboard keyboard works without permission prompts. This is the universal path and covers every phone in service today.
The Android app is live on Google Play under the Internal testing track. Version 1.0 (versionCode 1) shipped to invited testers, and the v1.0.2 polish build is ready as a signed AAB with onboarding redesign, the tool family launcher, the result hero family, and the latest five QA fixes baked in. Detector is the first tab: paste editor, document picker, the five-band score read-out, sentence highlights in a scrollable list, and the system share sheet wired as a receiver. Production roll-out is gated on screenshots, the feature graphic, and the remaining Play Console forms; the build itself is not the blocker.
There is no TextSight Detector app on the App Store. The iOS native build is on the 2026 roadmap and has not started development. The honest framing matters because most search results on "ai detector iphone app" silently bury this. The web app on iOS Safari covers the full scan, sentence-highlight, history sync, and copy workflow at parity with Android, and Apple supports Add-to-Home-Screen for a chrome-less full-screen window that behaves like a native app. No waitlist, no email collected, no gate. iOS development starts after the Android production listing settles.
The Android app is the most-shipped surface after the web. Both versions call the same backend; v1.0.2 is mostly UI polish and QA fixes rather than new detector flows.
Detector tab with a paste editor and a document picker, up to 5,000 characters per call on free and 50,000 per day on Pro. The 0-100 Authenticity Score read-out with five colour-coded bands (Original, Mostly Human, Mixed, Likely AI, AI Generated). Sentence highlights in a scrollable list with per-line scores. Three modes: Fast, Balanced, Deep. Recent scans saved per account. The Android share sheet wired as a receiver so any app on the phone (Chrome, Gmail, WhatsApp, Notion) can send text into the detector in one tap. Stub auth during the developer-verification bridge phase; Google Sign-In and email sign-in arrive before the public listing.
Onboarding redesign that ports the four-step welcome flow from the web. The tool family launcher that groups detector, AI rewriter, paraphrase, summarize, plagiarism risk, and grammar check into one screen so the detector is one tap from any other tool. A result hero family that reuses the same score header pattern across every output screen so the score gauge is consistent everywhere. Five QA fixes from the round-four review covering keyboard handling, score-band contrast, share-target preview, deep-link state, and remembered mode preference. No new endpoints, no new permissions.
Voice-to-text input through the standard Android speech intent for dictating a paragraph straight into the detector. Offline scan cache for the last ten results so the score is readable without a connection. Share-to-image for the score card. Full Material You dynamic colour where the app picks up the user's wallpaper-derived palette. None of these block the production listing; they wait until v1.0.2 ships.
No Play Billing during the developer-verification bridge, because the bridge does not satisfy Play Billing's verified-account requirement. Subscriptions are taken on the web at app.textsight.ai through Stripe and the entitlement carries to the Android build at sign-in. No on-device detector model; the scan compute stays server-side on every platform so the score never drifts between phone and laptop. No ad surfaces, no analytics SDK beyond Firebase Crashlytics for crash triage.
One account, one quota, one plan across mobile web, the Android app, and desktop. No platform surcharge, no app-store-only tier, no Pro-on-mobile-only feature.
Billed $89.88/year — Save $30
Billed $179.88/year — Save $60
Billed $359.88/year — Save $120
Yearly billing saves 25%. View full pricing →
The detector ships with three modes on every surface (web, Android, iOS Safari). They share one backend, so the score is consistent; only the model count and the latency change.
Single-model classifier optimised for latency. On a phone over 4G a 5,000-character scan returns in roughly two seconds. Use Fast when triaging a feed of social posts or doing a quick sanity check on a paragraph before posting. The five-band verdict is still accurate; the per-sentence breakdown is skipped to keep the response small.
The default mode and the one the phone hero CTA opens. A multi-model ensemble that mixes DeBERTa, ELECTRA, and the closed-loop scorer in roughly four seconds on a 5,000-character paragraph. Sentence-level highlights are included. This is the mode you want for an essay before submission, a blog draft before publishing, or a client email before sending.
The most thorough mode. Runs the per-sentence classifier in full plus the plagiarism risk indicator on Pro and above. Latency on a phone is around eight to twelve seconds for a 5,000-character document. Use Deep for the final read before a high-stakes upload (Turnitin sanity check, agency hand-off, scholarship submission).
Mode picker is a three-state segmented control above the editor on both Android and the mobile web. The chosen mode persists per account so the next scan opens in the same mode without re-tapping. No paid feature lock on Fast or Balanced; Deep's per-sentence classifier is unlocked on Pro and above.
A phone is not a small laptop. These four detector workflows are actually faster on the phone than on a desk, and they are the ones we see most in session analytics.
An essay written the night before needs one detector pass before the 9am submission. On the bus or train, open the Android app, paste the draft from the team chat, run Balanced, read which paragraphs flag red, swap to the writing app, soften two sentences, re-scan to confirm the score moved into the green band, share into Google Classroom or the LMS. The whole loop fits inside fifteen minutes and the laptop never leaves the bag.
A forwarded WhatsApp essay, a viral LinkedIn post, an investor update copied from a Substack newsletter. Long-press the suspicious text, copy, open app.textsight.ai in mobile Safari or Chrome, paste, scan. Two to four seconds later the five-band verdict tells you whether to trust the source. The detector is the first place to send anything that arrives in a chat app and feels too clean.
A LinkedIn post or X reply drafted with help from ChatGPT needs a quick check before posting. Paste into the mobile web detector, run Fast for triage, see the top two flagged sentences, rewrite those before tapping Post. The whole loop takes under a minute and prevents the "this is obviously AI" comment that kills the reach of the thread.
A cold pitch lands in the inbox at the airport. A vendor drops a 200-word capability blurb in the Zoom chat. Paste into mobile TextSight, see the score in two seconds, ask the right follow-up question. The score is not the decision but it shapes the next message, especially when a real human reply would change the outcome.
Most users do not need a native app for the detector. The web app on phone browsers covers scan, sentence highlights, history sync, paste, and copy at parity with native, and skips the install step entirely.
iOS Safari is the canonical iPhone path. The web detector handles the paste editor cleanly, paste from the standard iOS keyboard works, copy uses the navigator.clipboard API with no permission prompt, and the score gauge plus sentence highlights adapt to the safe-area inset on notched and dynamic-island devices. iOS Chrome is a WebKit shell under the hood, so behaviour is identical. Add to Home Screen gives a dedicated icon and a chrome-less full-screen window that behaves close to a native app.
Android browsers all work, and the web app installs as a Progressive Web App from the three-dot menu in Chrome and Edge with a home-screen icon. The Android native app is still the right pick if the share-target workflow matters, but the mobile web is the right pick if the user wants zero install friction and is fine with copy-paste between apps.
The detector compute happens on the server in every case. A scan from iOS Safari hits the same endpoint, runs against the same multi-model ensemble, and returns the same Authenticity Score as a scan from the Android app or from a desktop laptop. The only thing native adds is OS integration (share sheet, deep links, system tray), not accuracy or throughput.
No system share-sheet integration; the user has to copy and paste between apps. No background processing if the screen sleeps mid-scan; the call resumes on wake. No offline mode at all (this is true on every surface). File upload works on phone but the picker UI is rougher than on desktop. None of these block normal use; they are listed because most marketing pages quietly skip them.
The sister page for rewriting on a phone — same Android-live, iOS-roadmap story.
Open the AI rewriter mobile guide →Right-click any text on the web to score it in a hover card without leaving the page.
See the extension →3 scans/day with no signup needed for the first scan. Works in any mobile browser.
Open the free detector →The main landing page covering every source model and the full Authenticity Score system.
Open the overview →Mobile web on iOS Safari and Android Chrome covers everyone today. The Android app is on Play Store for testers and the v1.0.2 polish build is signed. iOS is on the 2026 roadmap; no waitlist, no email collected.