A sign-in window can be drawn inside a page instead of opened by your browser. Researchers have just found a scam that copies Google’s login so closely that the only thing it cannot fake is how it behaves.
What Actually Happened
This week, the security company Island published a report on a phishing operation — phishing means building a fake page that copies a real one so you type your password into it — dressed up as advertising tools for AI companies.
The fake products borrow the names of tools millions of people already use: ChatGPT, Gemini, Claude, Perplexity, and Muse, the new assistant from Meta. Each one promises to help a business buy advertising online. Each one ends at the same single button: Connect.
Click it and a Google sign-in window opens, padlock and all, with an address bar that reads accounts.google.com. It is not a window. It is a picture of one, drawn inside the page.
Behind the picture sits a person. The researchers watched operators hold a victim on a waiting screen, reject a password on purpose, then choose which second step to show next: a code by text message, a code from an app, a tap on the phone. Every attempt is kept. The report logs hundreds of people handing over details, with the pages still running while it was written.
The same machine has also run fake refund pages and fake job interviews carrying the names of Tesla, Louis Vuitton, Nike and Red Bull. Only the front page changes.
Why It Matters to You
You do not have to own a business for this to reach you. The trick works with any brand, and the same pages have already pretended to be job interviews and refund desks. What the thief wants is one working login: your Google account, your work account, or the account that pays for your phone.
An advertising account is really a spending account: it holds a saved card and an approved budget, and one manager login can reach several client accounts at once. That is why stolen ad accounts get resold, and why older ones with a clean history fetch two to four times the price of new ones. If you sign in with a work account, the thief can reach your employer’s email and files too.
The twist is the second step. Most of us believe the code sent to our phone is what keeps us safe. Here, the person who takes your password asks for that code in the same conversation, and uses it within seconds.
What You Can Do
Nothing here needs a program you have never opened before. It needs one habit changed.
- Read the outer address bar. The drawn window can claim to be accounts.google.com. The browser around it cannot: that bar still shows the page you are actually on.
- Try to move the window. Pull it by its title bar, or drag a corner to resize it. A real window follows your mouse anywhere on the screen. A drawn one stays inside the page.
- Open the site the long way. Type the address yourself or use a bookmark, instead of arriving from a link, an advert or an invitation email.
- Treat “connect your account” as a login request. If a tool you have never heard of asks for your Google, Meta or work account, look it up on the company’s own website first.
- Never read a code out. No bank, police force or support desk that contacts you first needs the number that just arrived on your phone.
- Use a passkey where you can. It signs you in with your fingerprint or your face, so there is no password to copy, and Google, Apple and Microsoft accounts now offer it.
- If a code already went out, act today. Change that password, sign out of all devices, and check any advertising account for administrators you did not add.
What People Get Wrong
Myth: I would spot a fake page
Copying the looks is the easy part. The kit even carries notes about giving the fake toolbar a frosted finish so it does not look painted on, and newer versions copy Safari’s rounded address bar, Chrome’s tabs and dark mode. Judge the window by what it does, not by how it looks.
Myth: my code keeps me safe
A code defends you against a criminal who bought a list of stolen passwords months ago. It does not defend you against someone typing your password in right now and asking you for the next number. Keep using codes; just do not treat them as a wall.
Myth: I do not use AI tools, so this is not my problem
The AI names are only the newest wrapper. The same pages have posed as refund desks and job interviews since at least March, and a fresh product is the best moment for a fake one: the Muse pages appeared eight days after Meta announced Muse.
What to Watch For Next
Expect more of these, and better ones. Fake pages now follow the news within days, and the brand on top is chosen for whoever is most likely to click it.
Three things deserve a second look: an invitation from an AI company you never wrote to; a web address that is almost right (the researchers logged gemini-ads.ai, plus deliberate misspellings such as gemimi and gennini); and any page that asks to connect an account you already have.
The One Thing to Remember
If a sign-in window appears inside a page, treat it as a picture, not as a window. Pull at it, or read the address bar of the window around it, and you will know.
And the oldest rule still holds: nobody who contacts you first has a reason to know your password, or a code that was sent to you.
Image: “Username and password illustration” by Santeri Viinamäki, Wikimedia Commons, licence CC BY-SA 4.0.