The official site of Hoshi-Bond, an AI companion chat app on Android where characters remember the user and do role-play.
HoShi (also called Hoshi-Bond / Bondmate internally) is an AI companion chat app from the Hong Kong company Huayun Innovation & Idea Technology Limited. The domain was registered in September 2025, and the app pitches role-play chat with characters that "remember you", aimed at US Android users. The website is just a landing page and web version that catches brand searches, while new users mainly come from Google Play; it monetizes with in-app ads through ad networks such as Bigo, plus a Premium subscription that removes ads, lengthens replies and strengthens memory. An invite-link subdomain went live six months ago, showing it is working on invite-based acquisition. No direct evidence was found for the specific acquisition channel (paid ads or organic app store traffic); from "a China-background team going overseas + ad monetization + almost no social traffic", it is inferred to rely mainly on app store promotion and store search.One-line growth story, compiled by us from public sources; speculation is marked separately below.
The main engine is app store acquisition (inference): direct visits are 73% and all search is organic, so the official site mainly catches people who have already installed the app, and new users are acquired on Google Play.
The lubricant is long-term chat that remembers you: existing users come back to keep chatting with their characters, and the search keywords are mostly people looking for Discord and error fixes; an invite link also launched in 2026, and visits rose 119% month over month last month.
By estimate, about two thirds of visits are existing users coming back to keep chatting; the visit count is an estimate, and about 19 clicks to the checkout page per 100 visits.
Model: Lenny's racecar growth framework. Engine = app store acquisition (inference: new users are acquired on Google Play, and the official site only catches brand searches and the web version); lubricant = long-term chat that remembers you, so existing users keep coming back, with free users seeing in-app ads and heavy users paying for Premium; the invite link is an attempt at an acquisition loop.
Real data: The largest channel direct at 72.83%, the top-three channel ranking (direct, organic search, referrals), organic search 100% of search, 24 search keywords in total, top search keywords (hoshi ai friends discord, hoshi app, hoshi ai chat, hoshi app not working, hoshi-bond), referring site category (Accounting and Auditing 94.36%, Other 5.64%), YouTube as the only social source, countries (US 97.3%, Spain 2.7%), bounce rate 34.35%, 7.58 pages per visit, average time 4 minutes 16 seconds, month-over-month +119.2% and global rank #604690 come from the Similarweb public page for last month; the public page gives no total visits, nor the shares of the second and third channels. 9,950 (about 9,950) visits to the Stripe checkout page come from quietmoney checkout-page data.
Supporting evidence: Made by the Hong Kong company Huayun Innovation & Idea Technology Limited, with the domain registered in 2025; the invite subdomain invite.hoshi-aifriends.com launched in 2026; app-ads.txt is mainly a direct Bigo Ads account plus a dozen or more ad resellers, which shows the app has programmatic ads; the Bondmate AI Discord server has only 7 members, so community is not a growth tactic; the web version supports 9 languages, but 97% of traffic is in the US.
Inference: The 52,500 total visits is estimated from Similarweb global rank #604690 (fitted on 437 sites with real visits, with half of sites within 17% error), so both channel circles carry ≈, and the largest channel's share is real. How each path is split is an inference: direct 38,200 = existing users 29,000 + saw it on Google Play and in ads 5,200 + friend invites 4,000; other channels 14,300 = searched the brand name after seeing it on Google Play and in ads 7,300 + existing users searching Discord and error fixes 6,000 + friend invites 1,000. New users coming from app store search and paid ads is the profile's guess, with no direct evidence on this site; the 5% who go to the official site after seeing it and the 40% who open an invite are guesses.
Checkout: buy.stripe.com
“'Want to talk with other Hoshi-Bond users or share your own experience?' (community Discord, not a growth mechanic)”Quote from the site
New-user source: Other (Direct / referral / email / paid). Direct 72.8%, search 20.4%, brand keywords 93.2%; search terms are mostly existing users looking for the Discord and error fixes (hoshi ai friends discord, hoshi app not working); social is only YouTube. New users mainly come from Google Play, with the website catching brand searches.
Top channel measured by Similarweb, others estimated by rank
| Attributable new users, last month | Share |
|---|---|
| Other | 100% |
Direct + brand search 93%. Direct and brand search count as viral only when the site shows sharing mechanics; the largest share wins, under 5% means Other.
| Country | Share of visits |
|---|---|
| United States | 97.3% |
| Spain | 2.7% |
English
YouTube
24 keywords · brand ~93% · organic 100% / paid 0%
| Similarweb top 5 keywords |
|---|
| hoshi ai friends discord |
| hoshi app |
| hoshi ai chat |
| hoshi app not working |
| hoshi-bond |
| Category | Share of referrals |
|---|---|
| Accounting and Auditing | 94.36% |
About 9,950 last month (Similarweb estimate), #327 of 606; the month before was 768. Opening checkout is not paying, so this is not revenue.
New-user source: Other (Direct / referral / email / paid). Mostly referrals, email or paid, or mostly direct with no clear source Direct 72.8%, search 20.4%, brand keywords 93.2%; search terms are mostly existing users looking for the Discord and error fixes (hoshi ai friends discord, hoshi app not working); social is only YouTube. New users mainly come from Google Play, with the website catching brand searches.
Has a Premium tier (user reviews mention longer replies, better memory, no ads); amount Not found
Dossier confidence: medium. Parts 1 and 2 come from reading the site and Similarweb data; part 3 from public reporting, founder posts, Wayback snapshots, WHOIS and similar, with speculation marked.
The number: checkout visits is Similarweb's estimate of visits from the site to a checkout payment page; opening checkout is not paying, so this is not revenue. Small sites land on fixed buckets.
Data through last month. Page updated 2026-10-31.