Updated October 13, 2026. This guide explains a documented scam and practical ways to avoid it; it does not claim that every ChatGPT advertisement is malicious.
You search for ChatGPT, click a sponsored result, and land on a page that looks familiar. Then a message says the service is busy and asks you to visit a backup site. The next screen looks like a routine human-verification check. That is exactly the sort of detour security researchers have linked to a malware-delivery trick called ClickFix.
The most important warning sign is simple: a genuine website verification check should not ask you to open Windows Run, Terminal or PowerShell and paste a command. If a page does, close it. Do not follow the instructions just because the page displays a recognizable brand or security logo.
What researchers found in the ChatGPT-themed campaign
In research published on October 1, 2026, Island security researchers described sponsored Google results that led to attacker-authored Custom GPTs or shared-chat content on the legitimate ChatGPT website. In the confirmed cases, unrelated prompts received the same fabricated service notice, steering visitors toward a supposed backup domain. That destination then led to a counterfeit ChatGPT and Cloudflare-style verification page.
Island connected the broader delivery activity to roughly 850 paid-ad landings, 26 lookalike ChatGPT destinations and 71 Google Ads campaign IDs observed from late May through August 24, 2026. Those numbers are not counts of infected computers. Researchers confirmed the exact lure on only a subset of the destinations. Their analysis of one malware-delivery path found behavior consistent with a Windows remote-access tool.
The important distinction: researchers did not report that someone had hacked ChatGPT or Google to make this work. The campaign relied on misleading sponsored results, attacker-controlled content, redirects and social engineering.
How a ClickFix trap works
- A credible starting point. A search ad, link or compromised website directs someone to a page that appears legitimate.
- A manufactured problem. A notice claims the service is overloaded, a download has failed or a security check must be completed.
- A fake verification step. The page imitates a CAPTCHA or Cloudflare check, but instead of asking the visitor to identify pictures, it instructs them to run a command on their computer.
- Execution by the visitor. If the visitor follows those instructions, the command can retrieve and run malicious software. The exact payload and consequences vary by campaign.
The U.S. Federal Trade Commission’s June 2026 warning describes this same fake-CAPTCHA tactic. Microsoft’s August 28 security research also documents a ClickFix variant called TerminalFix, where fake verification screens pushed users toward PowerShell or Windows Terminal.
Seven warning signs worth remembering
- A backup-site demand: a page unexpectedly claims a familiar AI service has moved or requires a different address.
- A suspicious sponsored result: the ad label or recognizable brand is treated as proof of authenticity. Neither guarantees that a destination is safe.
- A command disguised as verification: the page tells you to use Windows Run, PowerShell or Terminal.
- A clipboard trick: a website says it has copied verification text for you to paste into a system window.
- Urgency: the page says you must act immediately or lose access.
- Lookalike addresses: the URL differs from the official service even if the logo and layout look convincing.
- An unexpected installation: a supposedly simple chat or CAPTCHA task suddenly requires software or an extension.
One warning sign is enough to stop. You do not need to identify the malware or prove that the page is fraudulent before closing it. For broader advice, see our guide to phishing warning signs.
What to do if you encounter one
If you saw the page but did not run anything
Close the suspicious tab. Open a new tab and type the official service address yourself rather than following the redirect. For ChatGPT, use chatgpt.com and check OpenAI’s official verification guidance if a message or download seems questionable. If you entered a password on a suspicious page, change it from a trusted device and review account security.
If you pasted and executed a command
- Disconnect the affected computer from the internet. Turn off Wi-Fi or unplug its network cable.
- Use a trusted, separate device to change passwords for important accounts, starting with email, banking and the service involved. Enable multifactor authentication.
- Run a reputable security scan and follow your security software’s removal instructions. Keep Windows and protection tools updated.
- Get professional help if needed. A remote-access infection can be more serious than a single unwanted file. A workplace computer should be reported promptly to the IT or security team, which can investigate the device and related accounts.
- Review account activity. Look for unfamiliar sign-ins, recovery changes and suspicious email forwarding rules. Our email account recovery and protection guide covers useful next steps.
These steps follow the FTC’s published consumer guidance. A clean-looking scan is reassuring, but it does not by itself prove that credentials or sessions were never exposed.
How to reduce your risk before the next search
Bookmark the genuine sites you use frequently. Keep your browser, operating system and security software updated. Pay attention to the actual web address, especially after a redirect. Never treat a sponsored placement, a professional-looking interface or a familiar CAPTCHA logo as a substitute for checking where you are.
For families, students and small businesses, the most effective rule is easy to teach: websites can ask you to click a CAPTCHA; they should not ask you to run computer commands to pass one. If a colleague encounters a suspicious verification screen, encourage them to report it rather than experiment with the instructions.
Quick questions
Does this mean ChatGPT itself is infected?
No. The October 2026 investigation described abuse of legitimate platforms and attacker-authored content, not a confirmed vulnerability in ChatGPT.
Are all sponsored ChatGPT results dangerous?
No. The research documented a particular malicious delivery cluster. Still, ads should not be treated as proof that a destination is trustworthy.
Can this happen outside Windows?
Yes, related social-engineering schemes have targeted other operating systems. The particular Windows command flow described here is especially relevant to Windows users. Microsoft has also documented ClickFix-themed targeting of macOS users.
Sources and image credit
Reporting and practical guidance are based on the original Island research (October 1, 2026), FTC consumer alert (June 2026), Microsoft Security Research (August 28, 2026) and OpenAI’s official account guidance. The featured photograph is an illustrative stock image, not a screenshot of the campaign. Photo: Kowalski7cc / Wikimedia Commons (CC0).
