Fake ChatGPT ad tools turn account linking into a login trap

Island’s October 6, 2026 research describes counterfeit AI advertising tools whose “Connect” buttons display fake Google or Okta login windows inside the attacker’s page. Entering credentials or approving authentication requests in that flow can put advertising accounts at risk.

A plausible pitch for people who manage ad budgets

One fake ChatGPT page promises a weekly Google Ads briefing. A fake Gemini offering advertises support for manager accounts and linked clients, while other lures borrow the Claude, Perplexity and Manus names. Muse Ads was the newest version described by Island; these are counterfeit offerings, not announcements from the named vendors.

Island says the targets include advertising agency staff, media buyers and manager-account administrators. Advertising accounts carry budgets and stored payment methods, and a manager identity may reach several clients’ accounts—so the consequences can extend beyond one advertiser.

The address bar is part of the bait

The technique is called Browser-in-the-Browser: the page draws a second browser window inside the real one. The imitation can display a Google or Okta sign-in address and a lock icon, even though the real browser remains on the phishing domain. The kit also copies Windows, macOS, iOS and Android browser styling, including dark mode.

A human operator watches what the visitor submits and chooses the next screen. They can retain up to three password attempts, reject an entry to prompt a retry, and request an SMS or authenticator code, a Google approval or an Okta push.

Check the entry point—and the accounts it can reach

For an AI advertising beta or account-connection invitation, go independently to the vendor’s official site to check that the offering exists. Inspect the real browser’s outermost address bar, not the one drawn inside the page: the imitation cannot change the real page’s origin.

Island also recommends origin-bound passkeys and phishing-resistant hardware-backed authentication to protect against this type of credential collection. SMS and authenticator-based MFA remain useful, but a request for a code does not prove that a page is genuine.

If you submitted credentials or approved a request through a suspicious page, review every client advertising account that identity could access. Look for new managers or partners, changed recovery details, and campaigns or spending nobody authorized.