Name Watch: what AI answers say about a named doctor, lawyer or shop, captured as evidence

The problem is now well known. Aurascape, ZeroFox and Netcraft have each published research on it, and there are reported cases of AI tools handing a scammer's number to someone who asked for their bank's fraud line. At least one established brand-protection vendor, BrandShield, and one API vendor, cloro, already say they monitor AI answers for scam content. The incumbents that sell takedowns to banks (ZeroFox, Netcraft, Doppel, Bolster) can add an answer-probing module cheaply. So the general idea is taken, or soon will be. A new entrant probably can't win as a horizontal brand-protection tool. Its best chance is to go narrow and deep in banking. I found no one who checks contact details against an allowlist: extracting every phone number, URL and app-store link from an answer and comparing it to the bank's official list. I also found no one who probes by locale and phrasing at the scale of hundreds of queries per bank. Nobody is packaging this for the many credit unions and community banks that can't afford enterprise brand-protection contracts. Evidence formatted for regulators and audits, and handoff into a fraud team's existing case tools, are also open. Selling the probe data to brand-protection vendors is a real opening too, because most of them don't yet have an AI-answer data feed. The risks: incumbents can copy this quickly, AI engines may close these attack paths themselves, and abuse channels at AI engines are still weak, so the takedown packet may not get anything removed.

• BrandShield – AI Platforms Protection / Fraudulent AI Search Results

  The closest direct match I found. An established brand-protection vendor says it monitors ChatGPT, Gemini, Perplexity and Grok for phishing links, impersonation and scam content in AI answers, and says it gets that content removed. Its pages don't mention phone numbers, and they don't explain how it gets an AI answer taken down.

• cloro – Brand Protection API (SERP + AI-answer abuse detection)

  An API that says it covers seven AI answer engines (ChatGPT, Perplexity, Gemini, Copilot, Grok, AI Overview, AI Mode). It links AI citations of lookalike domains to an enforcement workflow, which makes it a likely data layer for brand-protection vendors. That could make it a competitor or a supplier for the takedown packet.

• ZeroFox – Digital Risk Protection (LLM SEO poisoning research)

  Has published research on fake support numbers being planted in ChatGPT, Gemini and Copilot answers. It recommends its own digital risk protection product for detecting this, and it already sells takedowns to banks.

• Netcraft (LLM login-URL research)

  Found that 34% of the login URLs an LLM gave for well-known brands were not owned by those brands, and one was a live phishing site. I found research, not a product. Netcraft is a major takedown vendor for banks and could easily turn this into a product.

• Doppel

  An AI-native impersonation detection and takedown platform with large funding. It covers social media, websites, apps and ads. I found no AI-answer monitoring feature yet, but it is the kind of vendor that could add one quickly.

• Bolster AI

  A brand-protection and takedown vendor covering domains, social media, ads, app stores and the dark web, and it uses LLMs to automate takedowns. I found no AI-answer monitoring yet.

• Aurascape (AuraLabs research)

  An AI security company that documented the first campaign of this kind it knows of, where Perplexity and Google AI Overviews gave out fake call-center numbers for Emirates and British Airways. I found research only, not a monitoring product for brands.

• Generic AI brand monitors (Apify AI Search Brand Monitor, PageCrawl)

  Scrapers and change monitors for ChatGPT and Perplexity answers. They are cheap to build on, but they don't check phone numbers or URLs against an allowlist and they don't do takedowns.

Checked 2026-10-10