Fake AI Sites Target Advertising Accounts With Credential Theft

A phishing campaign targeting advertising account managers uses fake ChatGPT, Gemini, Claude, and Perplexity sites to steal login credentials and multi-factor authentication (MFA) codes, according to researchers at browser security company Island.

The sites claim their AI products can help advertisers reach buyers, obtain ad briefs, and plan or audit campaigns and spending. To use them, visitors are prompted to connect an account. That button opens a fake Google sign-in window embedded within the page, with an address bar displaying accounts.google.com. The window is an iframe designed to capture credentials, rather than a genuine browser pop-up.

Island says the phishing kit adapts its interface to Windows, macOS, iOS, and Android. During the sign-in process, a human operator can request a password up to three times, ask for an SMS or authenticator code, display Okta push requests or Google approval prompts, and show a QR code. The operator can also reject submitted codes or keep victims on a waiting screen.

The researchers linked the campaign to a broader operation involving fake recruitment and refund pages. They traced the activity through exposed source code in misconfigured public GitHub repositories, with evidence reaching back to March. The operation’s pages share a Next.js and Socket.IO stack; many use Vercel frontends and Railway or Render backends. Island found that the platform supports Google, Meta, TikTok, and Okta sign-in workflows. Researchers also observed hundreds of victim submissions to the campaign’s Telegram control channel, a figure that does not establish how many accounts were successfully compromised.