10,000 Fake Stores Are Impersonating 166 Brands — and AI Chatbots Are Recommending Them to Shoppers

10,000 Fake Stores Are Impersonating 166 Brands — and AI Chatbots Are Recommending Them to Shoppers

Jasper Koers 12 min read Brand Intelligence

On November 12, 2026, NordVPN's Threat Intelligence team published an investigation that should alarm every brand that sells online. A single operator — likely Chinese-speaking, based on shared infrastructure patterns — has built a network of more than 10,000 fraudulent retail domains impersonating 166 fashion and luxury brands across 26 countries. As of November 3, at least 1,549 of those stores were fully active and accepting payments.

That alone would be a significant brand protection story. But the investigation revealed something far more dangerous: AI shopping assistants are recommending these fake stores to consumers as legitimate destinations.

How the Fraud Network Works

The operation is sophisticated in a way that specifically exploits how modern AI systems evaluate trust.

Each fake store is localized. It appears in the right language, shows the right currency, and lists a local address. The operator registered domains that look plausible — vero-moda.es for Vero Moda in Spain, drmartensie.com for Dr. Martens in Ireland. The sites display Visa, Mastercard, PayPal, and Klarna logos, though NordVPN found that some sites used separate systems to harvest card data rather than processing payments through those providers.

Adidas was the most impersonated brand with 52 domains. Ralph Lauren followed with 30, Dr. Martens with 29. The operator re-registered 100 expired domains to inherit their existing search history and backlink profiles — a tactic that gives new sites immediate search credibility without building it from scratch.

But here is the detail that matters most for the AI era: 597 of the fake domains had DNS records confirming access to Google Search Console. The stores maintained extensive sitemaps and supplied Google Shopping-compatible product data feeds. They were not just mimicking the visual appearance of brands — they were providing structured data that made them indexable and evaluable by the same systems AI assistants query when generating shopping recommendations.

AI Assistants Became Distribution Channels

NordVPN tested shopping queries in local languages across multiple AI platforms. The results were stark.

Perplexity recommended two counterfeit stores as trustworthy shopping destinations. In one case, vero-moda.es appeared alongside legitimate retailers Zalando and La Redoute — presented as an equally valid option. In another, drmartensie.com was described as Ireland's official Dr. Martens outlet. It is not.

ChatGPT and Gemini also cited sites linked to the network, extending the fake stores' reach beyond conventional search into conversational AI recommendations.

A separate study by Podqi, which tested 151 brands across ChatGPT, Gemini, and Perplexity, quantified the problem more broadly. Perplexity surfaced risky stores in 15% of shopping answers. ChatGPT did so in 8%. Gemini in 4%. Across all three platforms, 9% of shopping answers pointed to a risky site.

Scrunch's research, also from November 2026, widened the lens further. Testing across six AI platforms — ChatGPT, Perplexity, Gemini, Google AI Overviews, Google AI Mode, and Microsoft Copilot — Scrunch found that 17% of Fortune 500 companies had an impostor website cited in AI answers. In three out of four answers that cited a known impostor, the AI presented the site as the brand's official source.

Why AI Systems Cannot Tell Real from Fake

The traditional web had imperfect but functional trust signals. Google's search algorithm evaluated backlink authority, domain age, user behavior signals, and hundreds of other factors over two decades. A new fake site had to overcome significant barriers to appear alongside established retailers.

AI shopping assistants have inherited some of these signals but lack the verification layer that makes them reliable. When an AI assistant evaluates a potential shopping recommendation, it looks at the content of the page, any structured data the site provides, and whatever trust signals it can access. The problem is that every signal a fake store needs to game is on the public web:

Domain appearance. A domain that looks like the brand name with a country-code TLD passes the surface-level plausibility check. There is no mechanism for AI assistants to verify that vero-moda.es is not operated by H&M Group.

Structured data. The fake stores provide clean product data feeds, complete sitemaps, and schema markup. From a data-structure perspective, they look like well-maintained e-commerce sites — because the operator invested in precisely that infrastructure.

Payment trust signals. Displaying Visa and Mastercard logos does not require any relationship with those payment networks. The logos are freely available images. An AI assistant reading the page sees payment provider logos and interprets them as a trust signal, even though they prove nothing.

Content quality. Product descriptions, images, and category structures on the fake stores are copied or adapted from the real brand's site. The content quality is identical to the original because it is the original content, repackaged on a different domain.

What is missing is identity verification — a machine-readable way for AI systems to confirm that a website claiming to represent Adidas is actually operated by Adidas. Without this verification layer, AI assistants evaluate the signals that exist and conclude that a well-built fake store is a legitimate shopping destination.

The Brand Data Gap Behind the Trust Crisis

This problem looks like an AI safety issue, and it is. But it is also a brand data problem.

Consider what would happen if every brand published a verifiable, machine-readable identity fingerprint: the official domains, the authorized color palette, the correct logos, the verified social profiles, the authenticated contact information. An AI assistant could then compare what a website claims to be against what the brand's verified data says it should be.

Today, that verification layer does not exist in a standardized, machine-readable form that AI systems can consume. Brands publish their identity across dozens of surfaces — their own website, social profiles, retail partner listings, directory entries — but none of those surfaces is authoritative in a way that AI can verify programmatically.

The result is the identity gap that the NordVPN investigation exposes. A fake store provides structured data claiming to be the official Vero Moda outlet in Spain. The AI assistant has no machine-readable source of truth to check that claim against. The fake store's data is clean, well-structured, and self-consistent. It passes every quality check except the one that matters: is this actually Vero Moda?

The Scale Is About to Get Worse

The NordVPN investigation describes one network. A competent one, but one network. The economics of AI-mediated brand impersonation are compelling enough that this will not remain a single-operator problem.

Consider the attack surface. Traditional SEO fraud required gaming a specific algorithm with specific signals. The cost of maintaining a fake store in Google's top results was high because Google's spam team actively hunted these operations. The effort-to-visibility ratio kept the problem manageable for most brands.

AI shopping assistants change the calculus. They aggregate data from multiple sources, so a fake store does not need to rank first in Google — it just needs to appear in enough places that an AI model encounters it during retrieval. The 597 domains with Search Console access show that the operators understand this: they are not trying to outrank the real brand in organic search. They are creating enough structured data presence that AI systems include them in the recommendation pool.

Black Friday amplifies this. Shopping query volume spikes, consumers are primed for deals, and AI assistants are being marketed as tools to find the best prices. The combination of increased demand, deal-seeking intent, and AI-mediated discovery creates the ideal conditions for fake stores to reach consumers at scale.

What Brands Should Do Now

The NordVPN investigation, Podqi study, and Scrunch research all point to the same structural problem: AI systems lack a reliable way to verify brand identity claims. Fixing this requires action from brands, not just from AI platforms.

Audit your domain footprint. Search for domains that resemble your brand name with different TLDs, country codes, or slight spelling variations. NordVPN found that every major impersonated brand had multiple lookalike domains. Knowing which domains exist is the first step to challenging them.

Verify your structured data is authoritative. If your competitors in AI recommendations include fake stores, it means your structured data is not distinctive enough for AI systems to prefer it. Ensure your product data includes attributes that a fake store cannot easily replicate — verified business information, authenticated social profiles, and brand metadata that cross-references across multiple authoritative sources.

Establish cross-platform brand consistency. The NordVPN operators exploited inconsistency. A fake Dr. Martens store in Ireland works because there is no single, machine-readable source that says "these are the only authorized Dr. Martens domains in Ireland." When your brand data tells a consistent story across your own site, retail partners, directories, and social profiles, anomalies become detectable.

Monitor AI recommendation surfaces. Run regular shopping queries for your products across ChatGPT, Perplexity, Gemini, and Google AI Mode. When a fake store appears alongside your brand, document it. Most AI platforms have feedback mechanisms, and consistent reporting helps the platforms identify fraudulent sources.

Build your machine-readable identity layer. Your brand should have a structured, verifiable data fingerprint that AI systems can check against. This includes logos in standard formats with predictable URLs, a consistent color palette implemented in both CSS and structured data, verified contact information, and authenticated social profiles. The more structured and consistent your brand identity data is, the harder it becomes for a fake store to replicate it convincingly.

The Trust Verification Layer Is the Missing Infrastructure

The internet has had authentication for transactions (SSL/TLS), authentication for identity (OAuth), and authentication for email (SPF/DKIM/DMARC). What it does not have is authentication for brand identity in a format that AI systems can verify at retrieval time.

The NordVPN investigation shows that this gap has become commercially exploitable. 10,000 domains, 166 brands, 26 countries, and AI assistants serving fake stores as trusted recommendations. This is not a theoretical risk. Consumers are entering payment information on fraudulent sites because an AI assistant told them it was safe.

For brands, the immediate response is defensive: audit, monitor, report. But the structural fix is offensive: make your real brand data so well-structured, so consistently deployed, and so verifiable that AI systems can distinguish you from every impostor by checking the data, not by guessing from content quality.

Brand intelligence APIs exist to build exactly this layer. By extracting, structuring, and verifying brand identity data — logos, colors, fonts, contact information, social profiles — across every surface a brand appears on, they create the machine-readable fingerprint that AI systems need to tell real from fake.

The 10,000 fake stores in NordVPN's investigation did not need to build brands. They copied them. The defense is making your brand's identity verifiable in a way that copying alone cannot replicate.

Verify your brand's identity layer. Extract your brand fingerprint with Fetching Company and make sure every AI system can tell the real brand from the fake — before Black Friday, not after.

Share this article

Ready to try the API?

Extract brand data from any website with a single API call. Start free.