Short answer

AI website builders let anyone create a polished site from a text prompt, and criminals use them too. Proofpoint reported in August 2025 that Lovable-hosted pages were used for credential phishing, malware delivery and card theft, with tens of thousands of malicious URLs detected per month in email data. A professional look is no longer a sign of a real business.

Key takeaways

  • Proofpoint says tens of thousands of Lovable URLs were detected as threats each month in email data since February 2025.
  • Examples include a fake UPS site that collected card details, and crypto-themed pages aimed at draining wallets.
  • Proofpoint's researchers built working phishing sites with one or two prompts as a test.
  • Lovable told Proofpoint it took down a cluster of hundreds of domains and added detection in July 2025.
  • Guardio's test shows the same type of tool can produce a convincing fake retail store.

What Proofpoint reported

On 20 August 2025, the security firm Proofpoint published research on criminals abusing Lovable, an AI website builder that creates sites from text prompts and hosts them on a lovable.app address. Proofpoint says tens of thousands of Lovable URLs were detected as threats each month in its email data since February 2025. This is vendor research from a security company, based on its own telemetry, and we cannot independently verify the counts.

The examples it lists show the range:

  • A February 2025 campaign aimed at Microsoft credentials, which Proofpoint attributes to the Tycoon phishing kit, reached over 5,000 organisations with hundreds of thousands of messages.
  • A June 2025 fake UPS site collected personal and card details, including SMS codes, and sent them to Telegram. Nearly 3,500 messages were involved.
  • A June 2025 campaign impersonated the Aave crypto platform and likely aimed to drain connected wallets, with nearly 10,000 messages.
  • A July 2025 German-language campaign used a fake invoice download site to deliver malware.
  • Lovable URLs also appeared in text messages for investment scams and banking phishing.

Researchers could build the attack in a prompt

Proofpoint says that with one or two prompts its researchers created working phishing sites with backend logic, plus a working ClickFix project, and that the service even added deceptive language to encourage visitors to act. Free accounts could make a limited number of prompts per day, and free sites can be remixed by other users, which is how a template can be reused.

The vendor response

Proofpoint says it reported the campaigns to Lovable, which matched them to a cluster of credential phishing its Trust and Safety team had found, and took down hundreds of domains the same week. Lovable added real-time detection during prompting and daily scanning of published projects in July 2025, and removed the UPS template after it was reported. Proofpoint’s advice to builders is to anticipate abuse and build safeguards in, and to organisations to consider allow-listing policies for frequently abused tools.

What this means for a shopper or account holder

Guardio, another security vendor, used the same tool to build a convincing fake Walmart store for its own test and reported that Google Safe Browsing did not block it. That was a lab test, but it shows how little effort a good-looking fake takes. The FBI’s December 2024 alert says AI-generated text supports fraudulent cryptocurrency investment websites and that AI-made images are used for celebrity endorsements of counterfeit products.

So the old tells, a clumsy layout or odd English, are weaker. Better checks:

  1. Look at the address bar. A real bank or retailer will not live on a builder’s hosting address.
  2. Start from your own bookmark or app, not from a text, email or ad.
  3. Treat a CAPTCHA that appears before any content on a message link as a warning. Proofpoint describes CAPTCHA steps in several of the campaigns.
  4. Never type a text-message code into a page you arrived at through a message.
  5. Search for the store name plus “reviews” on independent sites, remembering that fake reviews exist.

Our online shopping scams guide, phishing messages guide and crypto wallet drainer guide go through the standard checks.

If you entered details

Contact your card issuer or bank immediately, change passwords from a clean device, and report the page to the FTC at ReportFraud.ftc.gov or to your national fraud service.

What you can do

  1. 01Do not trust a site because it looks professional. Check the web address character by character.
  2. 02Go to retailers and banks by typing their address or using their official app, not through a link in a message.
  3. 03Be wary of pages that insist on a CAPTCHA before showing content from an email link.
  4. 04Never enter a one-time code from a text into a page you reached through a message.
  5. 05If you entered card details, call your card issuer at once and report to ReportFraud.ftc.gov.

Related reading: AI voice clone scams and the data breach tracker.

Frequently asked questions

Is Lovable a scam platform?

No. Proofpoint describes it as an AI website builder that criminals abuse, and says Lovable took down a cluster of phishing domains the week it was reported and added real-time detection in July 2025.

How many malicious pages were there?

Proofpoint says tens of thousands of Lovable URLs were detected as threats each month in email data since February 2025. It also cites individual campaigns, such as nearly 3,500 messages in a June 2025 UPS campaign.

Can an AI builder make a fake online shop?

Guardio used Lovable to build a convincing fake Walmart store for its own test, and Google Safe Browsing did not block it. That was a research exercise, not a report of a criminal shop.

How do I check a shopping site?

Read the address carefully, search for independent reviews outside the site, and pay by a method that offers dispute rights. Our online shopping scams guide has a longer checklist.

Sources

  1. Cybercriminals Abuse AI Website Creation App for Phishing, Proofpoint Threat Insight, 2025-08-20
  2. "Scamlexity": When Agentic AI Browsers Get Scammed (We Put Agentic AI Browsers to the Test - They Clicked, They Paid, They Failed), Guardio Labs, 2025-08-20
  3. Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud (I-120324-PSA), FBI Internet Crime Complaint Center, 2024-12-03

This article reflects public information as of 2026-11-19. AI tools and rules change quickly; we update articles when new facts are published. Spot an error? See our editorial and corrections policy.