"Every model it lists is fake": how Zenda hit 130k installs

Why Zenda: AI Chat Bot Assistant (ai.starvia.zenda) got this far

Bottom line: Zenda wasn't built by a product; it was bought. It's one of ten apps batch-launched by a Hanoi app factory (STARVIA COMPANY LIMITED, the same people behind Growly Limited). It uses the most basic "AI chat wrapper" template and runs a heavy volume of video install ads on Meta (Facebook / Instagram / Audience Network) through two Pages with only a few dozen likes each. The volume goes into cheap countries like Nigeria, India, and Turkey, then high-CPM countries like Germany, UK, and Netherlands carry ad revenue. The steps in the install curve match the spend rhythm. The product itself has no differentiation; the edge is paid acquisition capacity and a multi-app portfolio.

1. Basic data (googleplay library, 2026-10-29 to 11-06)

• Launched 2026-09-22 (store launch date); category Productivity; version has stayed at 1.0.5; contains ads, no in-app purchases (app_snap.offers_iap = False, ad_supported = True). The store listing says "Some features may require a subscription", but our crawl shows no IAP, so subscriptions may not be live or may be hidden (unverified).
• Cumulative installs: 10-28 was 38,651,11-06, now 131,785; +93,134 in the last 7 days.
• Daily step jumps: 11-01 +10,443, 11-02 +46,475, 11-03 +16,990,11-04 through 11-06 for +19,226 over three days (app_day).
• Curve since launch (AppstoreSpy estimates, app_day_ext): 10-09 about 7 thousand, 10-12 about 14k, 10-18 about 24k, then another step at 10-25 through 10-27 (about +12k), reaching about 39k by 10-28. The curve jumps in steps rather than growing smoothly.
• Ratings: 727 (10-30) → 2,246 (11-06), adding +120 to +290 per day; ratings per install about 1.7%, a normal range, no sign of rating manipulation.
• Charts: only on the Productivity Top Free category chart, never overall or Top Grossing. Charted countries went from 10-29 (1) to 11-05 and 11-06 (20); best rank is Nigeria #42 (11-05). India's best is #98 (11-02), Turkey climbed from 179 to 107, Germany 197→144, UK 193→166 (chart_rank).

2. Growth tactics: Meta paid acquisition (confirmed)

metaads library (Facebook ad library scraper, tracked 11-04):

• The 65 ads we saw are all video, all with CTA INSTALL_MOBILE_APP, all landing on play.google.com/ai.starvia.zenda. Placements are Facebook + Instagram + Audience Network + Threads, i.e. the default all-placements setup of Advantage+ app ads.
• Likes are only in the double digits while hundreds of ads are running. That's the signature of a "pure paid-acquisition small app," as the metaads docs put it.
• The timing lines up: the ads we saw started running 10-15, and the AppstoreSpy curve steps up from 10-25 to 10-27. 10-30 brought a new batch of creatives. On 11-02, about 25 new video ads started in one day (metaads ad table), and Play installs rose +46,475 that same day, the biggest day in the whole dataset.
• Countries: ads are aggregated with country=ALL, so the exact target countries aren't available. The charting countries include low-price markets (ng, in, tr, th, my, sa, ae) and high-price ones (de, gb, nl, fr, it, es, pl). It looks like a global spray.
• This app's advertisers don't appear in the Google Ads or TikTok ad libraries (googleads / tiktokads: no results; they aren't on those trackers, so no result doesn't mean no spend). Google UAC is the other most common channel for apps like this, unverified.
• Organic channels: X (x-radar found 0 posts for "Zenda AI" in the last 30 days), media, and Reddit show no discussion. Searchable results are only auto-aggregator pages: APK sites, AppBrain, and CNET download pages.

3. The company behind it: a Vietnamese app factory

• STARVIA COMPANY LIMITED: its site starvia.ai calls itself an "AI Software Studio". LinkedIn-type pages list headquarters in Hà Nội and a headcount of 11-50 (from Jina search snippets). AppBrain shows it started publishing apps only 2026 years ago, with 10 on Google Play and roughly 800k downloads combined (per the AppBrain page).
• Our database lists 10 apps under Starvia totaling about 2.65M installs (developer table). All were published between 2026-08 and October: Amora AI role-play (826k), Battery Emoji Buddies (782k, #6 in Vietnam's personalization chart), Yolo Fitness lazy workouts (670k), Zenda (132k), PixGen photo editing (113k), Pictora AI image generation (83k), Sleepzy sleep, Glowy beauty filters, Talkie English learning, and Home Workout. The store listings also include Lingua, SnipFlix short dramas, and baby photo albums.
• Several apps use the developer emails [email protected] and [email protected]. That matches Growly Limited in Hanoi (growly.life, "AI-powered apps… Powering millions of devices", founded 2025), so it's the same team or a related company. This is the common Vietnamese playbook: multiple entities, many apps, covering every popular category, then paid acquisition plus ad monetization.
• Starvia's Facebook Page, "Starvia: AI Product", has only 206 followers but shows "84,341 talking about this". That engagement is driven by ads, not organic followers (from Jina search snippet).

4. Competitors (similar apps on Google Play, googleplay library 11-06)

A huge number of apps do the same thing. Zenda is small among them:

Bottom line: this category is already crowded. There are official big-tech apps and established reskins already earning through subscriptions, and new reskins keep appearing every month. Zenda added +93k in 7 days, which ranks roughly #20 among this category's gains over the last 7 days.

5. Differentiation

• Features (Q&A, writing, assistant templates, text-to-image, photo recognition) are identical to this kind of reskin. The "chat + image generation + camera recognition" that the CNET download page describes is standard.
• One store review with 20 people marking it helpful: "Fake app, full of fake AI models. they are listing gpt 6 astra and fable…they are all fake". In other words, it lists a bunch of nonexistent or unlicensed model names in its model picker as selling points. Another review says ads pop up as soon as you open it (Play listing, ng store, read via Jina 11-06).
• So the differentiation isn't the product. It's distribution: lots of creatives (hundreds of video ads running at once), multiple Pages, global geo spread, plus one company's experience running ten apps on shared paid acquisition and monetization. That really is why it grew, but that capability can't be copied from the product.

6. Where the money comes from

• No in-app purchase records; contains ads. Monetization is mainly interstitial and app-open ads through something like AdMob (reviews mention ads right on opening). The description leaves room for subscriptions, so a subscription wall may come later.
• The play is paid acquisition → ad monetization arbitrage: low-CPI countries (Nigeria, India) bring volume, while countries like Germany and the UK push up eCPM. There's no evidence it actually makes money. 11-04 Installs later slowed from 46k a day to about 6k a day, which may mean the budget was cut after testing ROI (unverified).

7. Visibility and copycats

• Visibility: almost nobody knows about it yet. X has 0 posts, no media coverage, no public revenue. It's only indexed by APK sites and data aggregators.
• Copycats: no one is copying Zenda specifically. The template itself is copied. Other products doing the same thing in the same window include SHI & COMPANY's "AI Chat - Ask AI Chatbot" (10-13 on the store) and Prop studio's Gene Ask.

8. What to borrow

• The playbook, step by step: use one company entity to mass-launch apps in hot categories (AI chat, AI photo editing, battery emoji, lazy workouts). Give each app one or two newly created Meta Pages, run hundreds of video creatives on Advantage+ app install ads, spread across countries, use low-price markets first to push chart rank and install counts, then add budget to whichever app's ROI works. The company's real asset is creative production capacity and the acquisition-and-monetization loop, not any single app.
• For anyone thinking about building an AI chat app, there's no opportunity signal here. The category is crowded, big tech offers it free, and reviews are already calling out "fake models".

9. Uncertain

• No retention, ARPU, or ROI data. Whether subscriptions have launched is unclear (the description says yes, but our scraped listing shows no in-app purchases).
• The country breakdown of spend, and whether Google UAC is used.
• Whether model names like "GPT 6 astra" and "fable" actually connect to other models or are just relabels, and who the backend model provider is.
• Whether the 11-04 slowdown is due to budget cuts or to the collection interval (11-04 and 11-05 show the same install counts, so the figures are multi-day totals).
• The storefront (ng) shows 4.8 stars and our details page (us) shows 4.23 stars. The histogram numbers repeat in a regular pattern (e.g. 103/103/0/103), which suggests Play is serving estimated values, so don't take the absolute rating too seriously.

Sources

• googleplay library (DuckDB on the agent-im machine): app_profile, app_day, app_day_ext (AppstoreSpy estimates), app_snap, chart_rank, developer, queried 2026-11-06.
• metaads library: advertiser and ad tables (page_id 1354177281110169 etc.), tracked 2026-11-04; ad library link www.facebook.com/ads/library/?active_status=active&ad_type=a…
• Play store page play.google.com/store/apps/details?id=ai.starvia.zenda (read via Jina, 2026-11-06)
• AppBrain developer page www.appbrain.com/dev/STARVIA+COMPANY+LIMITED/; starvia.ai; growly.life; CNET download page download.cnet.com/zenda-ai-chat-bot-assistant/3000-android-z… (from Jina search snippets)
• x-radar search for "Zenda AI" over the last 30 days: 0 posts.