Island maps 186 AI lookalike domains and three routes
Island says it traced 186 lookalike domains across 27 AI brands and identified three routes from AI lookalikes to malicious activity. Its TypoPrompting research was published on November 12, 2026.
Key takeaways · 4
- 01
ChatGPT was the most frequently imitated brand in Island’s inventory, with 25 lookalike domains; Claude had 18 and DeepSeek had 16.
- 02
One Claude typo, cloude[.]ai, led to a ClickFix lure that runs a hidden PowerShell command fetched from a decimal-encoded IP address.
- 03
The traffic system can select a destination based on a visitor’s location, operating system and browser.
- 04
Treat the 186-domain figure as a count of lookalike addresses, not infections or measured financial losses.
Lookalike domains are the entry point
Island traced 186 lookalike domains across 27 AI brands.[2] ChatGPT led its inventory with 25 domains, followed by Claude with 18 and DeepSeek with 16.[2] Island says these addresses can turn attempts to reach a familiar AI app into prompts to verify, install or run something else.[2] The traffic distribution system selects a destination based on the visitor’s location, operating system and browser.[2] That means a familiar-looking address does not, by itself, establish what a visitor will be shown.
Three routes lead to different risks
Island traced three routes from AI lookalikes: a ClickFix lure, browser-hijacking malware and a search-hijacking extension.[2] In one example, the Claude typo cloude[.]ai ended at a ClickFix lure.[2] Island says the gate collects a FingerprintJS profile, requires an MD5 proof-of-work challenge, looks up the visitor’s public IP and posts the profile to a server.[2] The recovered command downloads a PowerShell script and runs it in a hidden PowerShell window with execution-policy bypass enabled.[2] Island says the download address 3112851459 represents IP address 185.138.88[.]3 as a decimal number.[2]
Extension findings include a mismatch
Island says eight typos involving Claude, DeepSeek, Copilot, ElevenLabs and Voice AI led through deezyj[.]com to fake “Security Check” pages.[2] The MSIX file from the funnel matches the hash of the Net Guard Microsoft Store listing.[2] Island says the analyzed Net Guard browser-control features are enabled by default and classified each analyzed item as a malicious browser hijacker.[2] The extension listing promises Bing search, but its manifest sets the default search provider to find.flashysearches[.]com.[2] Island reported that 13 extensions were removed while 17 still listed showed 972,000 declared users combined.[2]
Counts do not establish infections
Island’s 186-domain figure counts lookalike addresses, not confirmed infections, and Island has not established financial losses from the campaign.[3] The wider gate corpus contained 550 matching scan records across 211 entry domains, including lookalikes of Scotiabank, Charles Schwab, Verizon and Microsoft Azure.[2] Separately, the installer funnel spans 191 related landing-page domains, about 45 product names and 1,437 source domains across AI tools, banks, insurers and telecoms.[2] Island also linked a group of 81 domains, including about 23 AI lookalikes, through shared publisher keys and affiliate IDs to the redirector a11[.]click.[2]
For security and IT teams, the findings make lookalike domains and unexpected install or verification prompts a practical part of AI-tool risk reviews. The figures describe infrastructure and observed routes, not proof that a particular organization’s users were infected or suffered losses.
Why it matters
AI training for your team
kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.
How this developed
14 November 2026
Island maps 186 AI lookalike domains and three routes
