What did Scrunch find about impostor websites in AI answers?
AI assistants are citing fake websites as if they were the real company. Scrunch, the AI search company Sitecore owns, checked ChatGPT, Perplexity, Gemini, Google AI Overviews, Google AI Mode and Microsoft Copilot and found that 17% of Fortune 500 companies had an impostor website cited in AI answers. It published the research on November 12, 2026.
In three out of four answers that cited a known impostor, the AI presented the fake site as the brand's official source. Scrunch estimates AI assistants give more than one million answers a year that cite websites impersonating real brands. The numbers come from Scrunch's November 12 release, issued at Sitecore Symposium in Orlando, and The Agile Brand Guide reported the same figures the next day.
These are Fortune 500 companies, with legal teams, brand protection budgets and domain monitoring. A fake site still got through for roughly one in six of them.
Why would an AI assistant send customers to a fake site?
AI assistants pick sources by what looks relevant and consistent, and they don't verify who owns a domain. A lookalike site that copies a brand's name, logo and product language can read as official to a model, so the answer cites it with the same confidence it gives the real site.
The customer often never checks. Separate research Scrunch published with Sitecore, called The New Shape of Search, found that more than 40% of AI sessions involved no open-web search beforehand and no click to the open web afterward, against 21.5% of traditional search sessions. If the answer hands over a phone number or a link, many people act on it as given.
Scrunch also told the Symposium audience that when an AI platform recommends a brand to someone new, that person is 182% more likely to search for the brand on Google and 117% more likely to visit its site within a week, according to CMSWire's coverage.
What did Scrunch launch alongside the research?
Scrunch launched two tools at the same event. Citation Optimization shows which third-party pages AI answers cite for the prompts a brand tracks and helps the team pitch those sources. Imposter Detection flags suspected fake sites and gives the steps to start a takedown.
Keep the source in mind. Scrunch sells the product that fixes the problem its research describes, and The Agile Brand Guide flagged the figures as vendor-sourced. I still think the pattern is real, because it matches how these assistants pick sources. Treat the 17% as Scrunch's number from Scrunch's method.
| Finding | Number |
|---|---|
| Fortune 500 companies with an impostor site cited in AI answers | 17% |
| Answers citing a known impostor that called it the official source | 3 in 4 |
| Estimated AI answers per year citing impersonating sites | over 1 million |
| AI sessions with no web search before and no click after | over 40% |
| Traditional search sessions that stay self-contained | 21.5% |
What does this mean for a small business owner?
If AI can mix up a Fortune 500 brand, it can mix up a local plumber, med spa or law firm with a similar name. The usual damage is smaller and quieter: an old phone number, a competitor's listing or a directory page that forwards calls to someone else. The customer thinks they called you.
I see the same leak with paid traffic. A business spends money on ads, someone hears the name, asks an assistant for the number later, and the call lands somewhere else. That person never shows up as a missed call, so it never gets counted. We've been tracking what missed calls cost for years, and this is a version of it you can't see in your phone logs.
How do you check what AI says about your business?
Ask the assistants directly. Open ChatGPT, Gemini, Perplexity and Google AI Mode and ask for your business's phone number, website, hours and reviews, the way a customer would. Write down every answer, and click every link it cites.
- Ask each assistant "What's the phone number for [your business] in [your city]?" and "What's the website for [your business]?"
- Click every cited source. Note any domain, directory or listing you don't own.
- Fix your own sources first: your website, your Google Business Profile and the main directories. Use the same name, phone number and address everywhere.
- Put your phone number, hours and service area on your site as plain text, not inside an image, so a model can read it.
- Report a site that copies your brand to Google Safe Browsing and to the domain's registrar.
- Repeat the check every month. Answers change when the models change.
Reviews on your real Google profile also help. They give the assistants more current, consistent facts about you to pull from. Our free Google review link generator makes a direct link customers can tap to leave one.
Make sure the right call gets answered
When a customer does find your real number, KenjiAI makes sure someone picks up. Our Voice AI agents answer every call, qualify the person and book them on your calendar, and the review requests go out by text after the job so your real profile keeps growing. We've been running ads since 2013 for 500+ businesses, and losing a ready buyer to a wrong number is one of the leaks I hate most.
What should you do this week?
Run the six-question check above on all four assistants today, and save the answers in a doc with the date. If any answer shows a number, site or listing you don't own, correct your own listings first, then report the copy. Do it again next month and compare.
If you want help with the answering side, book a call and we'll show you how the agents handle calls to your number.