AI scams against hotels: phishing, voice deepfakes and account takeover

What AI has changed in scams against hotels

Fake guests, fake suppliers and urgent payment requests have always existed. Generative AI did not invent these scams, but it has removed the flaws that staff learned to recognise:

• Flawless writing: emails and messages in fluent Portuguese or English, in the right tone and with details of the hotel itself, such as the property name, plausible dates and references to real services.
• Scale: the same scam can be adapted for hundreds of hotels at once, each with a different message.
• Convincing conversation: AI tools keep up long exchanges by email or WhatsApp, answering questions the way a person would.
• Cloned voices: with current voice cloning tools, short samples of public audio, such as a corporate video, an interview or a webinar, can be enough for a convincing imitation of a director's voice.

Hotels are a natural target. The operation runs on scattered channels (email, WhatsApp, OTA chat), the front desk deals with strangers all day, and a service culture rewards solving the guest's request quickly. That is exactly the kind of environment in which a well-written, urgent request goes through without being questioned.

1. Takeover of the hotel's OTA account

This is a well-documented scam against hotels in recent years, and it hits guests directly. The sequence usually goes like this:

1. The contact: the front desk or reservations team receives an email from a supposed guest. In one case investigated by Secureworks in 2023, the "guest" said they had left their passport at the hotel and sent a link to "photos of the document". Other versions involve a complaint, a refund request or a special need for the stay.
2. The malware: the link or attachment installs an infostealer, a program that copies the passwords saved in the computer's browser, including the one for the OTA extranet.
3. The access: with the password, the criminal logs in to the hotel's dashboard on the platform and sees upcoming bookings, with names, dates and message history.
4. The scam on the guest: through the platform's official chat, in the hotel's name, the guest receives a message saying the payment failed and the booking will be cancelled unless they "confirm their card details" through a link. Because the message comes from the hotel's legitimate channel, many people pay.

In March 2025, Microsoft reported another variation, active since December 2024: emails imitating Booking.com itself, sent to hotel staff, leading to a page that asks them to copy and paste a command to "fix an error". That command installs the malware. The technique, known as ClickFix, works because the user runs the code themselves, which gets around some automatic protections.

Two points matter here. First, in the documented cases the platform was not breached: what was stolen was the hotel's access, from a front desk computer. Second, the reported campaigns targeted hotels in Europe, North America and Asia, but the mechanics do not depend on the country. Any hotel that uses an OTA extranet on the same computer where it opens emails from strangers is exposed.

The damage goes beyond the guest's money: negative reviews, hours of work to put things right and, because guests' personal data was accessed, a possible security incident under the LGPD, Brazil's data protection law (see our guide to hotel IT compliance in Brazil).

If you suspect the hotel's OTA account has been compromised:

1. Stop using the suspect computer to access the account.
2. From another trusted device, change the passwords, end active sessions and turn on two-step verification.
3. Contact the platform's support team.
4. Preserve the evidence: messages sent, times and the affected computer, without wiping it before it has been examined.
5. Identify the bookings that may have been affected and warn those guests through an official channel.
6. Assess the hotel's obligations under the LGPD with its lawyers.

2. Voice deepfakes: the "manager" who calls asking for a payment

The best-known case is not from hospitality, but it shows how far the technique has come. In 2024, the engineering firm Arup confirmed that an employee in its Hong Kong office transferred about US$25 million after joining a video call in which every other participant, including the chief financial officer, was a deepfake built from public videos.

In a hotel, the version is simpler and more likely:

• The "director" who calls at night: the night auditor or the front desk gets a call in the voice of the general manager or the owner, asking for an urgent Pix transfer (Brazil's instant payment system) to a supplier, "or there will be no delivery tomorrow". The voice is familiar, the pressure is real and the transfer seems to solve a problem.
• The change of bank details: the finance team receives an email and, shortly afterwards, a confirmation call from a "supplier" giving a new account for paying future invoices.
• Fake technical support: someone introduces themselves as a technician from the PMS or door lock supplier and asks the front desk to install or approve a remote access tool, such as AnyDesk or TeamViewer, for an "urgent update". The tools are legitimate and used by real suppliers; the problem is the unexpected request. Once connected, the criminal controls the computer.

The call usually comes with a request for secrecy or speed, precisely to stop the person from checking with anyone. That is the pattern to teach the team: urgency plus an exception to the normal process is a warning sign, no matter who is speaking.

3. "Guest" messages written by AI

WhatsApp, Instagram and email are common guest service channels in hotels, and they are where everyday social engineering arrives:

• The fake VIP guest who, after a long and polite conversation, asks the hotel to buy gift cards, pay for a third-party service "that they will refund at check-in" or send a refund to a different account.
• Forged receipts and documents, easier and easier to produce with image tools. For Pix payments, the rule is to check the bank statement, never the screenshot, as we explain in our guide to Pix scams and booking fraud (in Portuguese).
• Fake hotel websites and profiles, with copied photos and AI-generated text, that take bookings and payments from travellers who believe they are dealing with the property.

4. Leaked passwords and automated login attacks

Infostealers and breaches of other websites feed huge lists of email addresses and passwords. Bots test these combinations automatically against OTA extranets, cloud PMS, company email and loyalty programmes, a technique known as credential stuffing. Anyone who reuses the same password across several services, or shares a single front desk password across every shift, hands over access without realising it.

Here AI plays less of a part in the break-in and more in what comes after: once inside, it becomes easier to read conversations, understand the operation and write messages that look as if they came from the hotel.

AI in defence: what actually helps

AI is also used on the protection side, and it is worth using. Modern email filters analyse how messages behave, not just keywords. Detection and response tools on computers and servers (EDR, or MDR when a team is monitoring) can spot an infostealer running or an unusual login and block it before damage is done.

But it pays to be realistic: in many of these scams, the victim hands over the access or authorises the payment. Technology can detect malware, compromised credentials and unusual behaviour. It cannot, on its own, stop an authorised employee from making a payment because they believed they were following the manager's instructions. When assessing a tool "with AI", ask the supplier exactly what it detects, who is alerted when something happens and how quickly someone acts, including outside business hours.

The controls that reduce the risk most

1. Two-step verification on everything that matters. OTA extranets, PMS, email, banking and remote access tools. Prefer an authenticator app to SMS. A password stolen by an infostealer loses much of its value when it is not enough on its own.
2. A call-back rule. No payment, change of bank details, password reset or remote access is authorised on the strength of a call, voice note or message. The person hangs up and calls back on a number already on file. For urgent requests from management, as well as calling back the contact on file, amounts or exceptions above a set limit need a second internal confirmation.
3. Independent approval for changes to bank details. A change to a supplier's payment account only takes effect once it is confirmed with the contact already on file, never through the phone number or email that came with the request, and approved by a second person.
4. Keep the extranet away from everyday email. Ideally, access OTAs and administrative systems on a separate computer or browser profile, without opening attachments from strangers. Avoid saving extranet credentials in the browser of a shared front desk computer; for credentials with privileged access, use a password manager approved by the hotel.
5. Remote access only through approved channels. Suppliers connect through tools chosen by the hotel, in sessions started by the hotel. Any unexpected request to install or approve a remote access tool is treated as suspicious until it is confirmed with the supplier's known contact.
6. Named accounts, no shared passwords. Every employee has their own login on each platform, and access is removed as soon as someone leaves the team.
7. Tell the guest. In the booking confirmation and at check-in, say that the hotel never asks for card details or a new payment through a link in the chat. That turns the guest into a line of defence.
8. Short, recurring training. Real examples, a few minutes long, repeated through the year and including the night shift, which is often left out. Drop the old rule of "watch out for spelling and grammar mistakes": it no longer works.
9. A plan for when it happens. Who changes the passwords, who contacts the platform, who warns the guests and who assesses notification to the ANPD, Brazil's data protection authority. Agreed in advance, not during the incident.

These controls add to those that protect against ransomware in hotels (in Portuguese), since a computer compromised by phishing is also a way in for bigger attacks. To check where your hotel stands across the wider list of threats, use our free hotel cybersecurity checklist.

Next steps

AI has made scams more convincing, but the defence is still largely the same: two-step verification, processes that allow no exceptions over the phone and a team that knows what to do when something seems wrong. For a hotel that depends on OTAs, digital payments and online guest channels, these controls should no longer be treated as optional. A good place to start is checking today who accesses each OTA extranet, from which computer, and whether two-step verification is turned on.

Our team can review access to your hotel's platforms, set the verification rules for payments and remote access, and organise training for the front desk and reservations teams. Talk to our team.

Sources

• Secureworks Counter Threat Unit, Vidar Infostealer Steals Booking.com Credentials in Fraud Scam (2023), sophos.com
• Microsoft Threat Intelligence, Phishing campaign impersonates Booking.com, delivers a suite of credential-stealing malware (March 2025), microsoft.com
• CNN, British engineering giant Arup revealed as $25 million deepfake scam victim (May 2024), cnn.com