Ghost stores, deepfakes and fake reviews: AI's con trick on fashion shoppers

Between 1 January and 30 April 2025, researchers in Vienna counted 36,725 Facebook and Instagram adverts shown in Austria for so-called ghost stores: online fashion boutiques that presented themselves as family businesses in Vienna, Salzburg or Munich, announced a painful closure and invited shoppers to a final sale. According to the Österreichisches Institut für Angewandte Telekommunikation (ÖIAT), the campaigns reached about 85.9 million people across Europe, and 61% of the users targeted were women, especially those aged 45 to 65.

Generative AI did not invent online shopping fraud. What it has done is make the most persuasive parts of a fake fashion business almost free to produce: the founder with a moving backstory, the storefront photograph, the model shots, the product copy and the review text. The result is a shadow retail sector that borrows the emotional language of independent boutiques and the visual polish of established labels, and that can rebuild itself within days of being reported.

Regulators have responded. The United States now bans fake and AI-generated reviews outright, the UK competition authority can fine companies directly, and European consumer groups are pressing the European Commission to use the Digital Services Act against the platforms that carry the adverts. Yet the enforcement figures suggest the fraud is scaling faster than the response, and legitimate fashion companies are paying a hidden price: the trust signals they depend on, from a founder story to a five-star rating, are being devalued.

What are AI-generated ghost stores and how do they work?

Australia's competition regulator gave the clearest official description. In July 2025 the ACCC said ghost stores falsely present themselves as local Australian boutiques that are closing down and selling high-quality clothing and footwear, when they are allegedly based overseas, are not closing and drop-ship low-quality products. It estimated at least 360 consumer reports about 60 online retailers since the start of 2025 and named four websites in public warning notices. The stores typically advertise through paid social media posts, appear to use Shopify, and rebrand after closing. "This conduct preys on the empathy of consumers who have a genuine desire to support local businesses," said Catriona Lowe, the ACCC's deputy chair.

The pattern is international. In August 2025 The Guardian reported that Madrid-based cybersecurity researcher Marta Mallavibarrena had found four shops using identical closing-down language; a search of Meta's ad library returned 50,000 ads with the same phrase, of which 1,600 were still active. Pictures of founders can be AI-generated, the paper noted, while real images from reputable shops are sometimes reused. Fraud consultant Serpil Hall called the schemes "industrialised social engineering".

In July 2026 The Canadian Press identified at least 21 Facebook pages claiming to be Canadian boutiques, created between March 2025 and June 2026. Eighteen were managed from the Netherlands, and most of the linked shops used images that appeared AI-generated. One page claimed: "My great-grandmother started this boutique in Victoria in 1956." A shopper reported receiving only 30% of the purchase price back on a return. The ÖIAT study identified the same three levers in German-speaking markets:

• An emotional closure story, often a retiring owner or a family member saying goodbye after decades.
• Promises of quality and service that a drop-shipping operation cannot deliver.
• A regional identity, with city names such as Vienna, Salzburg, Berlin or Munich in the store name.

How big is the fake shop problem in Europe?

The Austrian figures are a rare measured snapshot. ÖIAT, which acts as a trusted flagger under the Digital Services Act, used Meta's Ad Library interface to trace 71 deceptive stores with their own landing pages and 84 Facebook pages. One store alone, schneidersalzburg.at, ran 2,785 ads reaching more than 5 million people. After one page was removed, nine new account variations appeared, and an estimated 5 million additional people were reached while reports were pending. By the end of May 2025 Meta had not removed any of the pages ÖIAT had reported.

Germany's federation of consumer organisations, vzbv, describes a broader fake shop economy. Its November 2025 paper counts more than 10,000 complaints to consumer advice centres in 2024, up 47% on 2023, and more than 5,000 in the first half of 2025. Twelve per cent of online shoppers surveyed said they had fallen victim to a fake shop in the previous two years. When vzbv checked 653 websites that had received fake shop warnings in 2025, half had advertised on Google or Meta; the 227 shops found on Meta platforms had placed more than 403,000 ads. The vzbv data are not broken down by product category or by AI use, so they show the size of the channel rather than the share of fashion within it.

How are deepfakes and AI fake founders used to sell fashion?

The founder story is now a manufactured asset. In May 2026 PYMNTS summarised an ABC News investigation that found dozens of operations on TikTok and YouTube using generative AI to create fake founders, factory footage and brand stories to sell low-quality imported goods at premium prices. In one Instagram video a granddaughter described her grandfather's decades of hand-stitched leather bags; the story was entirely fabricated. A purported New York clothing retailer posted an AI image of its own storefront with shattered glass and police tape to announce a big sale. The sites often disappear before customers can leave reviews or file complaints.

Real designers are also being cloned. Cassey Ho, founder and lead designer of the activewear label Popflex, told Forbes in 2024 that a seller had used AI to swap her face onto another person in a video of her modelling her own skort, which then appeared in a counterfeit Amazon listing. Amazon removed three listings after Forbes flagged them. "It feels incredibly violating," Ho said. Juozas Kaziukėnas, chief executive of Marketplace Pulse, warned at the time: "We are all about to have our friends see 'ourselves' promoting goods we had no part in."

Platforms have begun to respond to the celebrity version of the problem. In October 2024 Meta said it would resume facial recognition testing to fight so-called celeb-bait ads, enrolling about 50,000 public figures and blocking ads when a face in a suspected scam matched their profile pictures. The UK and the EU were initially excluded pending regulatory approval. Such systems protect the famous; they do little for a mid-sized label whose designer, models and product photos are copied.

Are AI-generated fake reviews illegal now?

In the United States, yes. The Federal Trade Commission's rule on consumer reviews and testimonials, announced in August 2024, took effect on 13 October 2024. It bans fake or false reviews and testimonials, explicitly including AI-generated ones and testimonials from people who do not exist; buying positive or negative reviews; undisclosed insider reviews; review suppression; and the purchase of fake followers or views to misrepresent influence. Then FTC chair Lina Khan said fake reviews "pollute the marketplace and divert business away from honest competitors". On 22 December 2025 the FTC sent its first warning letters under the rule to 10 companies, reminding them that civil penalties can reach USD 53,088 per violation.

In the UK, the consumer provisions of the Digital Markets, Competition and Consumers Act 2024 came into force on 6 April 2025. Posting and commissioning fake reviews is now on the list of banned practices, and the Competition and Markets Authority can decide infringements itself and fine companies up to 10% of global turnover without going to court. "Consumers deserve to know that the CMA has their back," said chief executive Sarah Cardell. Because the powers are not retrospective, the CMA has signalled that fines in the first 12 months are likely to be lower.

In the European Union the pressure is aimed at intermediaries. In May 2026 BEUC and 29 member organisations filed complaints under the Digital Services Act against Meta, TikTok and Google over scam adverts, asking the Commission and national coordinators to investigate and, if necessary, fine the platforms. vzbv argues that fake shop advertising should be treated as a systemic risk under Article 34 of the DSA and that a ten-day response to a trusted flagger report fails the requirement to act without undue delay.

What are platforms doing about scam ads, and is it enough?

The most uncomfortable numbers came from Meta itself. In November 2025 Reuters reported internal documents in which Meta estimated that about 10% of its 2024 revenue, roughly USD 16 billion, came from ads for scams and banned goods. A December 2024 document put the average number of higher-risk scam ads shown to users at 15 billion a day, and Reuters reported that Meta would only act against a likely scammer when its systems were 95% certain. Meta disputed the reporting, said the documents were not indicative of the full picture and stated that user reports of scam ads fell 58% globally in 2025 thanks to improved detection.

TikTok has published large enforcement totals for its shop. According to Allure Security figures cited by PYMNTS, TikTok rejected more than 1.4 million seller applications, blocked 70 million products before listing and removed about 700,000 sellers for policy violations in the first half of 2025. Such figures show effort, not outcome: they do not reveal how many fraudulent listings got through.

Independent testing points to a weak notice-and-action system. Between December 2025 and March 2026 BEUC and its members reported nearly 900 suspected unlawful ads in 13 countries. The campaign focused on financial scams rather than fashion, but it measured the same mechanism that ghost stores exploit.

Only 27% of the reported ads were taken down; 52% of reports were rejected or ignored. "Meta, TikTok and Google not only fail to pro-actively remove fraudulent ads but also do little when being notified," said Agustín Reyna, director general of BEUC. ÖIAT has proposed a review of an entire page once a threshold of blocked ads is reached, stronger automated blocking of recreated pages and a fixed response time, such as seven business days, for trusted flagger reports.

What do fake shops and AI scams cost consumers and brands?

Consumer losses are measured, though not by product category. The FTC recorded 3 million fraud reports in 2025 and reported losses of USD 15.9 billion, up from USD 12.5 billion the year before, PYMNTS reported. Americans lost USD 2.1 billion to scams that began on social media, an eightfold increase since 2020, and impersonation was the most reported category. Most scams are never reported, so these are floors, not totals.

The cost to legitimate brands is harder to count but visible in the cases. Real independent boutiques compete against operators that copy their storytelling and undercut their prices. The Canadian Federation of Independent Business, which represents 103,000 small firms, told The Canadian Press it was unaware of the fake Canadian boutiques. Designers whose likeness is cloned face enforcement work that drains budgets: "It's hard to put all your money into fighting when you could be innovating," Popflex's Ho told Forbes.

What are the limits of the crackdown, and of the evidence?

Three caveats matter. First, the AI share of the problem is usually inferred rather than measured. The Canadian Press judged that images appeared AI-generated from their glossy, unnatural look; ÖIAT notes AI-generated storefronts but does not quantify them; vzbv does not mention AI at all. Much of the ghost-store model is old-fashioned drop-shipping with better visuals, and treating every fake shop as an AI story risks misdirecting the response.

Second, the official statistics are complaint-based. The FTC, ACCC and vzbv count what victims report, not what happened, and none isolates fashion. Platform numbers such as Meta's 58% fall in user reports or TikTok's 700,000 removed sellers are company claims that cannot be checked independently. Third, enforcement remains a game of whack-a-mole: Austrian researchers watched nine new account variations appear after a single page was removed.

There is also a counter-argument that brands should take seriously. The same rules that target fraudsters apply to them. The FTC's December warning letters went to businesses, not scam networks, and the rule covers incentivised sentiment, undisclosed employee reviews and bought followers. Brands that use AI to draft review responses, produce synthetic testimonials or inflate social proof are now inside the regulated zone. Heavy-handed platform filtering could also catch genuine small boutiques that use emotional storytelling in their ads.

What should fashion companies do now?

1. Audit your own review practices against the FTC rule and the UK banned practices list, including incentives, employee reviews, AI-drafted testimonials and purchased followers.
2. Monitor ad libraries for impersonation by searching Meta's Ad Library and Google's ad transparency tools for your brand name, designer names and campaign imagery every week.
3. Make your provenance verifiable with a named team, a real address, company registration details and a clear statement of which channels and shops are official.
4. Build a takedown playbook with evidence templates, platform reporting routes and marketplace brand-protection contacts, so that clones are reported within hours rather than weeks.
5. Work with trusted flaggers and associations such as national consumer bodies and trade federations, whose reports carry formal weight under the Digital Services Act.
6. Label your own AI imagery consistently, so that customers learn to distinguish your disclosed synthetic content from undisclosed fakes.
7. Brief customer service teams to recognise victims of clone shops, record cases and feed patterns back to legal and marketing.