Android AI App QA Audit — Charmlet UX & Functional Testing

I tested Charmlet, an Android AI dating assistant, from a first-time user's perspective. The review combined functional QA, usability checks, and evaluation of AI-generated suggestions. The deliverable was a structured report with five prioritized findings: two high and three medium.

Role: Manual QA Tester & UX Reviewer Device: Samsung Galaxy S25 Ultra Operating system: Android 16 Evidence delivered: Six screen recordings and a four-page QA report

I followed the core journey of adding a match, providing context, and generating a reply. I reviewed side-menu navigation, saved-profile behaviour, Topics selection, the Options modal, quick-instruction drafts, keyboard interaction, and the relevance and naturalness of generated messages.

The report paired each issue with observed behaviour, expected behaviour, priority, supporting recording references, and a practical recommendation. This made the feedback actionable for the development team rather than a list of general impressions.

High — Blank or prolonged loading states during navigation. Opening destinations such as About, My Account, and Tips & Tricks could replace the interface with a blank screen. Recordings 21434 and 21436 documented intermittent and prolonged cases. I recommended preserving the previous view until the destination was ready, with immediate loading feedback and a retry path.

High — A saved profile field changed when reopened. The report documented an unexpected change in a saved gender selection. I recommended checking persistence and load-time value mapping, followed by retesting after reopening the profile and signing back in.

Medium — A quick-instruction draft disappeared after opening Options. Entering text, opening Options, and returning to the composer cleared the draft. The keyboard also interfered with access to Generate. I recommended preserving draft state and keeping the primary action accessible above the keyboard.

Two further medium-priority findings covered Topics closing before the user had completed selection and background conversation scrolling while the Options modal remained open. The AI review also identified repetitive Topics, low-value text fragments, and openers that sounded unnatural.

These pages show the reporting format, observed behaviour, expected behaviour, recommendations, and evidence references. Video filenames refer to recordings delivered during the engagement; the raw recordings are not republished here.

In the published Upwork review, the client confirmed that several issues were reproduced and fixed from my notes and gave the engagement a 5.0 rating:

“Several of the issues he flagged were real and we were able to confirm and fix them from his notes.”

This is client-reported confirmation of several fixes, not a claim that every issue was resolved or independently retested.

Need an independent Android or AI-app review? Send your app, priority journeys, target devices, and release deadline through Contra.