gempai.app is an impersonation phishing page that presents itself as a playful Japanese tool for generating sparkly mahjong hands. It is listed by the MetaMask and SEAL blocklists, which points to crypto wallet users as the likely targets. PhishDestroy first flagged it on 2026-10-07, three days after the domain was registered. As of 2026-10-19 the site was still online and serving content.
How the scam works
The page title, "Gem牌 | キラキラ麻雀牌姿ジェネレーター", promises a "sparkly mahjong hand generator", while our classification is impersonation phishing. Schemes of this kind copy the look or identity of a trusted service so visitors lower their guard, then steer them to an action that hands over control: typing a password or seed phrase, or connecting a wallet and signing a request. Anyone who complies can lose account access or have funds moved out of their wallet.
Evidence
Nine of 92 VirusTotal engines flagged the domain on 2026-10-11. The vendors include alphaMountain.ai, BitDefender, CRDF, Forcepoint ThreatSeeker, Fortinet, G-Data, Gridinsoft and Kaspersky. Most engines have not flagged it, but these detections come from independent sources. MetaMask and SEAL both list the domain alongside PhishDestroy, so wallet software that relies on those lists may warn users before the page loads. Four AlienVault OTX threat pulses also reference the domain, showing it is being tracked by researchers outside our project.
Infrastructure
The domain was registered on 2026-10-04 and caught within three days, which fits the short life of disposable phishing domains. It was registered through CloudFlare, Inc. and sits behind Cloudflare's network (AS13335), with the address resolving to the US. Because Cloudflare acts as a proxy, the visible IP belongs to the network edge rather than the operator's real server.
What to do
Do not enter any credentials, recovery phrases or private keys on gempai.app. If you connected a wallet or signed anything there, revoke the site's token approvals with a trusted approval-checking tool. Move remaining assets to a fresh wallet if you shared a seed phrase. If you typed a password, change it everywhere you reused it and turn on two-factor authentication. Defenders should block the domain at DNS and proxy level and report it to Cloudflare's abuse channel. They should also check logs for users who visited after 2026-10-04.
注册信息创建 2026年10月4日 (16d · Very New!)—Expires 2027年10月4日
HTTP 状态码200
技术详情Provenance & timestamps
首次发现2026年10月7日
DOM Analysisanalyzed 2026年10月9日score 100/100
Submitted URLhttp://gempai.app/
Favicon Hash
ICANN OVERSIGHT
认证和 RAA 背景
Registrar accreditation and DNS abuse obligations
For this gTLD, the registrar above operates under an ICANN accreditation agreement. The links below provide the official fee schedule and current DNS abuse compliance guidance.
Accreditation is a contract, not a safety certification.
RAA §3.18 establishes abuse-contact and handling requirements. This report can document stored outbound notices and later technical observations; it does not by itself establish receipt, investigation, remediation, or contractual non-compliance.
Certificate fingerprintdd63edd8717b2c2c81dd63714af8c8e1fe1fb8533ab9cb0b935c4a39dab8ebe2Certificate names · 1 SANs
www.gempai.app
Saved certificate metadata. Certificate dates without a timezone are shown as stored. Transport encryption does not establish that the site is trustworthy.
Evidence layersAvailability: Content servingContent: Content served at rootDNS: ResolvedRegistration: Unknown
Latest HTTP observation内容在线Ordinary HTTP content served90%2026-10-19 22:39:24 UTC
RDAP registration未知
Observed timelinelast reachable: 2026-10-19 22:39:24 UTClast content: 2026-10-19 22:39:24 UTC
Availability, content, DNS and registration are independent evidence layers. NXDOMAIN, an unreachable origin or missing content alone does not prove registrar action. A registrar or provider is credited only when a direct technical marker identifies that actor. Report causality is shown separately.
技术 · 3 identified
Cloudflare Browser Insights
AnalyticsRUM
Cloudflare Browser Insights is a tool that measures the performance of websites from the perspective of users.
Cloudflare is a web-infrastructure and website-security company, providing content-delivery-network services, DDoS mitigation, Internet security, and distributed domain-name-server services.
Template-based draft · optional AI wording assistance requires separate consent
AI辅助草稿——事件详细信息由AI提供商处理 · 自行审核并提交
After a phishing incident
针对受害者的建议与指导
Choose the steps that match what you shared. Keep a copy of this report alongside your own records.
我应该立即采取什么措施？
紧急
Stop using the suspicious site. Open the real service through its official app or an address you already trust.
更改所有密码 — 电子邮件、Exchange 账户，以及任何使用同一密码的账户
启用双因素认证 使用身份验证器应用（而非短信）。禁用基于短信的恢复功能
冻结卡片 如果您在钓鱼网站上输入了银行信息
我应该为报告收集哪些信息？
联邦调查局（FBI）指南
Keep original emails, receipts, messages and screenshots in a secure place. A factual timeline helps the receiving authority understand what happened. IC3 evidence guidance ↗
Payment records — if you paid, save the receipt, amount, date, payment provider and recipient details.
Your bank or payment provider — contact its fraud team promptly through official contact details if money or card information was exposed.
当地警方 — 即使无法立即采取行动，也能留下正式记录
A report is not a guarantee of recovery or investigation, but prompt, accurate transaction data can help authorities and service providers trace the incident.
How do phishing scams typically work?
虚假网站 — 与正规网站完全一致的克隆网站，仅域名略有改动
猪的屠宰 — 通过Telegram/WhatsApp/交友软件花数周时间建立信任，随后钱被骗走
退款诈骗 — fraudsters pose as recovery agents and demand upfront fees. Never share a seed phrase or pay before independently verifying the provider
About this reportSources, observation dates and corrections
This report combines saved page captures, registration and network records, scanner results and PhishDestroy reporting history. Each source keeps its own observation date. Open a source report to inspect the original result.
The AI-assisted analysis was written on . It describes the evidence available at that time; later source checks can change.
A capture documents what was shown. A scanner result records that provider’s assessment at the time of its check. Neither a valid certificate nor a zero-detection result is a guarantee of safety.
(US)
