Reservation Fraud: What It Costs Restaurants and How to Stop It
Bot networks and scalpers can hold a restaurant's tables and never show up. Here is how reservation fraud works, what it costs, and the booking controls that stop it without adding friction for genuine guests.
A table can be booked, confirmed and never sat. Nobody calls it fraud at the time, because the booking looks like a booking: a name, a phone number, a party size, a slot the restaurant held and staffed for. Add up the ones that never arrive across a service and a restaurant is running a full dining room for a fraction of the guests it planned for. Reservation fraud is the name for the subset of those bookings that were made to take a table out of circulation rather than to fill it.
The mechanics are borrowed straight from concert ticketing. Sophisticated fraudsters use bot networks - AI-enabled software scripts that operate autonomously - to rapidly book multiple reservations at popular restaurants, creating artificial scarcity. Those coveted spots are then resold at inflated prices, in an underground market for sought-after dining experiences.
What has changed is that the phone is now part of the attack surface. Voice cloning and automated phone calling further exacerbate the problem: the restaurateur Slang AI spoke to had detected voice AI-powered bots calling restaurants and booking reservations directly with the host, bypassing online booking platforms and causing further disruption by tying up valuable staff time. A reservation line that rings a person mid-service is a line a bot can work.
What reservation fraud actually looks like
The first pattern is bulk reservation, and it is mostly invisible from inside the restaurant. A bot network rapidly books many tables at the places people want. What the bot network is building is scarcity, and the scarcity is artificial: manufactured a table at a time.
The second pattern is the phone call, and it costs labour rather than inventory. A bot calls the reservation line, talks to whoever picks up, and books. The staff time it consumes is real: the same restaurateur had detected voice AI-powered bots booking reservations directly with the host, bypassing online booking platforms, and tying up valuable staff time in the process.
The third pattern is the resale layer on top. The table was taken to be resold, not sat. This is the pattern that produces the most anger, because the guest in the seat has paid a resold price in an underground market the restaurant never sees.
What it actually costs
The first cost is the empty seat. Restaurants operate on thin margins, and wasted inventory - unsold food or empty tables - can be an existential threat. A table held for a booking nobody honours is inventory that was staffed and blocked and then produced nothing.
One high-profile restaurateur told Slang AI that up to 30% of their tables were being fraudulently reserved, a significant loss in revenue, and that they had responded by taking inspiration from airlines and overbooking. That is one operator's account rather than an industry average, and it should be read that way, but it shows the scale the problem reaches at a venue worth targeting.
Overbooking is a partial answer with a heavy price. The logistics of managing it are not trivial: airlines spend vast sums of money on predictive algorithms and logistics experts to perfect overbooking and still struggle to avoid bad customer experiences, and realistically restaurants do not have the resources to match airlines. A restaurant that overbooks to offset fraud takes on the heavy operational cost that strategy imposes on already stretched operators.
The second cost is staff time. A bot call that books a table is a call a host took. Some restaurants have responded by training hosts to ask qualifying questions on calls to check whether the caller is human - a clever strategy with an unclear window of effectiveness, because fraudsters constantly monitor their bot conversations and retrain them, giving the bots the ability to answer new questions.
The third cost is the hardest to measure, because it is not a loss the restaurant can see. When a guest cannot get a table and takes it somewhere else, that booking was never made, so it never shows up in the no-show numbers either. It is a lost cover that will not appear in the booking data at all.
Why the usual defences are not enough
Requiring a credit card to hold a table is one option restaurants weigh, and it is a genuine trade-off rather than a mistake on either side. Adding friction such as requiring credit cards to secure reservations risks alienating genuine guests by complicating their experience, while it may effectively drive off fraudsters and reduce no-shows - and today's guests want convenience and ease but expect their reservations to be honored without hassle. The card hold works on fraud and spends goodwill doing it.
Legislation is moving in the same direction, with limits. Legislators in New York proposed the Restaurant Reservation Anti-Piracy Act, which prohibits third parties from booking unauthorized restaurant reservations with food service establishments and specifies fines for those violations. The honest read is that it is hard to speculate how far penalties and fines will act as a deterrent, and that law enforcement is as expensive as mitigation. Legislative action alone has limitations and must be one part of a more collaborative approach - tech companies, reservation platforms and government agencies sharing information on fraud patterns, best practices and technological advances.
None of that changes what a single restaurant can set. Which is where the leverage is.
The booking controls that do the work
The controls that matter are policy settings rather than staff behaviour, because the AI host applies them when it checks availability and offers a table, without anyone having to notice something odd on the line. DineCall supports booking horizon, minimum notice, maximum party size, active-booking limits per phone number, table-lock duration, last seating, slot size and dining durations by party size.
Read them as bounds on what a single caller can take, rather than as blocks. A booking horizon bounds how far ahead a table can be taken. Minimum notice bounds how close to the slot a booking can still be made. Maximum party size bounds the size of the table a caller can hold. An active-booking limit per phone number bounds how many live bookings one number can carry at once. Each one narrows the range of bookings that are worth automating, and each one is applied by the booking flow rather than by a host's judgement.
Two of the settings are about the table rather than the caller. Table-lock duration sets how long a booking holds its table, and last seating and slot size bound the end of the book and the size of each opening. A table that is released sooner is available sooner.
There is a second lever, and it is a workflow rather than a limit. Restaurants can confirm eligible bookings instantly or hold new requests for manual approval. A booking a person sees before it is confirmed is a booking a person can decline.
The other half is knowing who is on the line. DineCall's No-Show & Fraud Protection automatically tags and warns restaurant staff about callers with a history of repeat no-shows or fraudulent ordering, permitting blacklist restrictions. Bookings also appear immediately in the owner dashboard, where staff can create bookings, seat guests, reassign tables, complete a visit or cancel a reservation. A tagged caller and a live board turn a policy from a setting into a decision.
The phone channel needs its own answer
The phone deserves its own answer because it is where a restaurant's guests actually are, and the channel fraud bots have already been caught using. A booking form is structured input, so rules can be attached to the form itself. A phone call is a conversation, and every control has to be carried out by the person on the line in a few seconds, without sounding hostile to a guest who did nothing wrong.
That rules some tactics out. Asking every caller to pass a test complicates the experience of everyone, genuine guests included, in order to protect against a minority, and it is exactly the tactic fraudsters monitor and retrain against. What works better is letting the booking flow enforce the rules automatically, so a caller behaving normally experiences nothing unusual at all.
There is a second phone-specific problem, and it is a staffing one. A bot call consumes a host. The way to reduce that cost is not to make each call shorter but to make fewer of them, which means the line is answered every time, including when it would otherwise ring out during service. DineCall's AI host answers in English, Cantonese and Mandarin, detecting the caller's language on the fly within one or two words and switching vocabulary and accent without requiring key presses.
One thing to be clear about is where a phone system sits relative to a reservation platform. DineCall takes the phone channel; it does not replace a reservation platform. OpenTable keeps managing your reservations while DineCall answers the phone - handling hours, directions and complete phone orders without touching your booking flow. If a restaurant already runs OpenTable, the phone channel is a layer alongside it, not a replacement.
What it costs to run it
DineCall prices per location, with a base monthly fee and zero order commission. Starter is CAD $99 per month per location with 200 AI call minutes included, then CAD $0.35 per minute. Standard is CAD $399 per month per location with 1,000 AI call minutes included, then CAD $0.20 per minute. Pro is CAD $599 per month per location with 3,000 minutes, then CAD $0.15 per minute. The first 100 partner restaurants joining by December 31, 2026 can lock the Standard base fee at CAD $199 while continuously subscribed. Multi-location plans are quoted by location count and call usage. There is no setup fee.
For a restaurant thinking about this as fraud control, the useful comparison is against the cost of the problem rather than the cost of the software. A table a bot took on a Saturday you could have filled, and a host's shift spent on a scripted call, both land in the same place as the tables that never got sat. Measuring them separately is how the size of the number becomes visible.
Setup takes 2 to 3 business days, including menu digitization, bilingual translation verification and routing tests. Restaurants keep their existing public phone numbers, and no new hardware has to be installed in the restaurant.
The part that is easy to miss
None of this fixes the problem on its own. Fraud will always persist and bad actors will always try to profit from that scarcity, which is why the response has to be collective rather than a set of isolated fixes: tech companies, reservation platforms and government agencies sharing information on fraud patterns and best practices. AI and biometrics hold promise for detecting suspicious patterns and adding a layer of verification, but they require continuous adaptation.
What a restaurant controls is the narrower question of what its own booking system allows, and that is worth doing on its own terms. Not because it ends reservation fraud, but because every table a bot cannot profitably take stays available for a guest who will actually show up.
FAQ
How do I know if I have a reservation fraud problem?
Start with bookings that were confirmed, never seated and never cancelled. The live reservation board shows them, because phone bookings appear immediately in the owner dashboard, where staff can seat guests, reassign tables, complete a visit or cancel a reservation. The harder-to-see case is the opposite one: times you were fully booked and did not serve, which appears nowhere in your booking data because the booking was never made.
Should my restaurant require a credit card for every reservation?
It is a trade-off, not a free win. Adding friction such as requiring credit cards to secure reservations risks alienating genuine guests by complicating their experience, while it may effectively drive off fraudsters and reduce no-shows - and guests today expect their reservations to be honored without hassle. Booking policy settings such as active-booking limits per phone number, minimum notice, maximum party size, booking horizon and table-lock duration are enforced by the booking flow before it offers a table, without putting that decision in front of a guest.
Does AI phone answering make reservation fraud easier?
Automated phone calling and voice cloning further exacerbate the problem, and they are part of why it has grown. The same automation also means a line that is always answered: a bot that reaches a line nobody picks up is a lost booking, and a caller with a history of repeat no-shows or fraudulent ordering is one staff can see flagged.
Can DineCall book into my existing reservation platform?
DineCall takes the phone channel, and the platform keeps doing what it does. OpenTable keeps managing your reservations while DineCall answers the phone - handling hours, directions and complete phone orders without touching your booking flow. You keep your phone number and your configuration.
How long does it take to set up?
Typically 2 to 3 business days, including menu digitization, bilingual translation verification and routing tests. Restaurants keep their existing public phone numbers, and no new hardware is required in the restaurant.
What happens if a caller needs a person?
If speech confidence falls below threshold, or the caller asks for a manager, the AI immediately alerts front-of-house staff by phone ring or dashboard notification, and logs a call-back ticket if lines are busy. Restaurants can also confirm eligible bookings instantly or hold new requests for manual approval.