80k installs land on Taiwan's Top Grossing chart: an adult-oriented AI companion disguised app, 7 weeks of paid acquisition, switch category, stop spending and it's gone

An app called Thousandfold Reverie AI, still on version 0.2 with no updates since Aug 5, is sitting at #10 on Taiwan's Google Play Utilities Top Grossing chart. Half its reviews complain about crashes and diamond purchases not crediting. It didn't take off on its product; it rode a roughly 7-week paid acquisition push, an ad hinting that adult topics are allowed, and a cleverly chosen category. Below we break down where the money came from, where the traffic came from, why it sits under "Utilities", and whether you should follow it.

A review that the ad tricked into the store

Late night, Sep 17, 2026, a one-star review written in Korean appeared in the Taiwan store:

This ad promises adult-oriented sexual chat. But when I installed it and picked the "student" persona, any turn toward explicit or suggestive content just shut the conversation down. I feel like I was tricked, and I'm not paying for a subscription at all. I don't believe that unlocking "19+" content will actually deliver what the ad promises.

This review lays out the whole business: ads are running, the pitch is adult chat, and inside the app there's a paywalled "18+" unlock. But the store page describes it differently: "a creative tool that integrates AI image generation and video animation," where each character "has a unique personality and storyline, and switching characters lets you experience different romantic vibes." The website calls it a generic "AI assistant" and spends most of its space on safety, reporting, and community rules. One pitch for ads, another for the store.

What it is and who made it

An AI character chat + image and video generation app: you chat with characters, and the chats can generate images and short videos. Launched Jul 2026 21, charted only in Taiwan, with reviews almost entirely in Traditional Chinese.

The developer is Hong Kong company WANWIN LIMITED (萬呈有限公司). The company's listed apps were originally tools like Cyber WiFi and Local Weather Wherever (the latter now delisted per April, AppBrain). Its website wanwin.cc still carries "銀賀的每一款应用" in the company intro, which points to another Hong Kong shell company, 銀賀有限公司, built from the same site template, likely mass-produced (inferred). The package name com.thousandfold.reverie.gp.blansh carries a random suffix, and the store also lists a nearly identical "Reverie AI" under developer AntCollect, last updated Sep 2, still live. The team behind it appears to be an overseas-expansion outfit that mass-creates shell companies and mass-publishes apps, most likely based in mainland China (inferred, not verified).

Where the traffic came from: one round of paid acquisition, with a cliff in mid-October

Connecting the cumulative installs (8 through October are AppstoreSpy estimates; from 10-29 onward are public Google Play install counts):

Mid-September through early October saw two to three thousand installs a day, then they abruptly stopped between Oct 9 and 18. Organic growth doesn't look like this. It's the shape of a paid push that was switched on and then off. I couldn't find media coverage, X discussion, or TikTok videos about it, and I didn't directly see where the creatives ran. But Taiwanese users on Threads complained that Meta's ad slots were "always flooded by other AI chat apps," with "otome games and AI chat" ads everywhere. In Taiwan that mostly means Meta, so it's probably running there too (inference).

Where the money comes from: diamonds deducted per use, with membership upsells on top

In-app purchases range from $0.99 to $99.99. Reviews reveal the pricing structure: generating one video costs 450 diamonds; there's a weekly Pro membership, but "after buying a week of Pro, generating a video with sound still pushes you to an upgraded plan"; and daily check-ins give points. Users were blunt: "Generation is strong, I'd give it 90 stars, but it's not cheap!" and "The customization is great… but the overall price is a bit high."

AppGazers estimates its Google Play IAP revenue in October at $25.7k to $123k, with 5.7k to 28k downloads that month (the estimate range is very wide). Even at the low end, 80k installs producing over $20K a month in revenue means each user pays far more than a typical utility app. That's how strong willingness to pay is for adult companion chat.

The cost shows up in reviews too: "Paid but never got the diamonds," "Used 450 diamonds to generate a video and can't find where the video went," "Stuck on a black screen, quality not worth paying for," "Unexplained charges." There are 3.6,242 ratings in total: 1 one-star reviews, 5 two-star reviews 75 and 136 zero three-star reviews, 2 four-star ones. Two five-star reviews that said only "讚" ("like") were posted in the same minute on the night of Oct 18. Whether the good reviews were solicited can't be seen from outside, but the pattern looks suspicious.

Why it ranks under "Tools"

This is the most instructive move here. In Taiwan's Google Play Tools Top Grossing chart, #4 Hope Pro has only about 18k installs, and #10 is this app (Google Play chart, 11-06). The Entertainment category's #10 is Crushie at 790k installs. In Entertainment, it can't even crack the top 30. In Tools, a little paid spend gets it charted. Other apps on the same chart use the same trick: 感性AI男友 has only 22k installs yet ranks #2 in Comics Top Grossing, and Pogee, newly launched in September, ranks #8 in Lifestyle. Tools probably also draws less scrutiny of adult content than Entertainment or Social categories (inference).

On Oct 27, right after the paid push stopped, the website's privacy policy was updated to state compliance with Google Play's AI-generated content policy. Whether that's because it was reviewed or because it's preparing materials for the next round, I can't tell.

Who it's competing with

Its only distinguishing trait is bundling companion chat with image and video generation in one package, while the ads hint at adult content. Each half on its own has better competitors. In this niche, it's just a 80k-vs-2M small package.

If it were you

The verdict: there's something to learn here, but not something to copy. The demand and willingness to pay are real, but this approach is already crowded in Taiwan, and its real "technology" (shell companies, random package names, ads and store pages with two different pitches) buys a package that can be delisted at any time and drops to zero the moment spending stops. Grade C.

Who it's for: teams already running going-global paid acquisition, with multiple developer accounts, who can calculate the cost per install and first-week payment; or people who want to learn the trick of charting niche categories and apply it to a legitimate product.

Who it's not for: indie developers who haven't run paid ads (there's no organic traffic here, so stopping spend means zero); people building a long-term product who need to build reputation; anyone who can't accept the risk of Google Play's adult content policy (one delisting and both account and revenue are gone).

How to do it (if you must try): don't fight head-on in Taiwan's AI companion market. The bet worth considering is switching language and audience: Traditional Chinese women-focused content and BL are already held by WHIF and 感性AI男友, so start by checking the Play Top Grossing charts in each country's Tools, Lifestyle, and Comics categories. Look for a language where the top 10 still sit around 10,000 to 20,000 installs and no AI companion package has taken hold. At minimum the product must not crash, and diamond purchases must actually deliver. It fails on this, and half of its 1-star reviews are about exactly that. Keep the store page, ads, and in-app content aligned, and don't rely on suggestive hints to drive clicks.

How to acquire users: paid acquisition only, mainly Meta. Test creatives on a small budget first, and see whether first-week payments cover the install cost. This kind of package's lifecycle is a single campaign, so payback should be calculated over 6–8 weeks (inference).

Is there a deadline: no clear one. It isn't an open slot waiting for you. It's a category that many packages have already been milking.

What would show I misread it: ① Thousandfold Reverie, or the similarly named Reverie AI, regains momentum before December and installs climb back to the thousands per day, which would show this is a repeatable model, not a one-off package; ② it stays in the top 10 of the Tools Top Grossing chart more than a month after spending stops, which would mean the paying users who remain are worth more than estimated and the unit economics are better than assumed; ③ conversely, if it or Reverie AI gets delisted before year-end, that would show the "switch categories, two pitches" route is being shut down by Google.

Data and sources

• Installs, ratings, IAP range, update dates, and Taiwan category charts: public Google Play pages and charts, 10-29 through 11-06; 8–October installs are AppstoreSpy estimates.
• Revenue and monthly downloads: AppGazers model estimates (October), wide range, unverified.
• Review quotes: Google Play Taiwan reviews, 2026-09-09 through 10-27.
• Developer and shell companies: Google Play developer page, wanwin.cc, yinheehk.cc, thousandfoldreverie.chat (including the privacy policy from 10-27), AppBrain.
• Taiwan ad environment: posts by Threads users @zuchrou.art and @hsiyu_yeh.
• No discussion of it found on X, TikTok, or in media.