An alert watches a restaurant and tells you when a table opens, and you book it yourself in your own name. A bot holds your booking credentials and completes the reservation as you, which is what the major platforms prohibit in their terms. The alert leaves nothing of yours exposed, and the bot puts your account at risk rather than the vendor's.
Sometimes. A bot can be faster than a human in the seconds after a release, which helps at the 30 rooms we measure that release between 7 and 11 in the morning. It cannot create availability that does not exist, and it cannot change what opens: Carbone logged 300 release events in 22 observed days, yet its most frequently open slot was 11:30 PM and no recorded open slot fell between 7 and 8:30 PM.
For your account, yes. An alert service never logs into your booking account, so there is no credential to leak and no automated booking for a platform to detect. SnagRes never books a table, never holds one, and never charges per reservation, which is also the line New York's 2024 reservation law draws.
Because you are paying a premium for a table booked in someone else's name, and restaurants increasingly cancel those. Free tables also open more often than people assume: we recorded 7,048 release events across our tracked rooms in at most 22 days of observation. The hard part is knowing which date and hour, which is what an alert answers.
Fast, and that is the honest weakness of alerts. A cancellation at a hard restaurant is usually claimed within minutes, so an alert only helps if you can act when it lands. If you cannot, an auto-booking service is the realistic alternative, with the account risk that comes with it.
More than most people expect. Of the rooms we measure, 31 are easier to book for four than for two and 26 are easier for two than for four. Monkey Bar never showed an open two-top in its window while a four-top came free on 6 of 30 dates, so the same restaurant can be fully booked for two and open for four at once.
The legal side of bots and resale is covered in are reservation bots legal, and the resale marketplaces specifically in the AppointmentTrader write-up. For the paid-membership version of the same promise, see what Dorsia is.
Every release hour quoted here is listed room by room on every release hour we have measured, the full study is at the full availability study, and the rooms that are genuinely bookable tonight are on the restaurants you can actually book.
An alert tells you a table exists and leaves the booking to you. A bot logs into your account and books for you. A resale marketplace sells you a reservation somebody else already made. Those are three different products with three different risks, and they get discussed as if they were one because the companies selling the riskiest two prefer it that way.
The table above sets out what each does and what it costs. The rest of this page is about the part the table cannot hold: what each approach is actually racing against, and what we can prove about that race.
We sell alerts, so treat this page the way you would treat any vendor's comparison. The difference worth knowing is that we have no exposure to the other two models, which means nothing here is written to protect a revenue line.
An alert needs you to be free when it fires. That is its whole weakness and there is no way around it.
A bot needs your booking credentials, because it has to act as you. The account that gets banned is yours.
A resale needs a stranger to hold a table in their name until you arrive at the door in yours.
Only one of the three leaves you with an ordinary reservation, made by you, in your name, at menu price.
Sometimes, and that is the honest answer. A bot wins by being faster than a human at the exact second a table appears, which is a real advantage at a restaurant that releases on a fixed schedule. What the sales pages tend not to say is that booking platforms actively work against them, so the speed advantage decays as the platform adapts, and the account being detected is the one you handed over.
The larger problem is that speed only helps during the seconds after a release, and most hard restaurants release less often than people assume. Of the rooms we measure, 30 have a release hour between 7 and 11 in the morning, 13 of them at 10 AM and 11 at 9 AM. For those, being early matters enormously. For the rest, the calendar advances on a schedule no bot can hurry.
And speed does not change what opens. Carbone recorded 300 separate release events across 22 observed days, which sounds like constant opportunity. On the day we measured it, 2 of the 31 dates in its window had a bookable two-top, its most frequently open slot was 11:30 PM, and not one of its recorded open slots fell between 7 and 8:30 PM. A faster finger there buys you a midnight table, not a Friday at eight.
Worth knowing who is telling you: the most-cited article on whether reservation bots work is published by a company that sells reservation bots. That does not make it wrong, but it is not a neutral source, and neither is this page. We sell alerts. The difference is that we can describe the alternative honestly because we are not exposed to its downside.
We watch 284 rooms in New York, San Francisco, Los Angeles and Chicago, and as of 19 October 2026 we have logged 61,530,416 checks of their live booking calendars and recorded 465,235,816 open time slots. That is a small list watched closely rather than a large one watched loosely, and it is what lets us describe the behaviour of a specific dining room instead of the category.
That record answers the question that decides whether any of these tools is worth buying: what does this restaurant actually do? Not whether it is famously hard, but at what hour it releases, how far ahead, for which party size, and at what time of night the seats that come free tend to sit.
Monkey Bar is the clearest example. Its calendar reaches 20 days ahead, its releases land at 9 AM, and across 550,180 checks of that room a two-top never came free on a single date in the window while a four-top came free on 6 of 30. Ninety-five percent of everything that opened there was before 6 PM, and its single most frequently open slot was 11:45 AM. It logged 483 release events in 22 days.
Read that as a buying decision. A bot pointed at Monkey Bar for a Saturday two-top at eight is racing for something our data says almost never appears. Knowing that is worth more than being fast, and no service watching tens of thousands of restaurants can tell you either way.
39 of the 284 rooms with a full measurement record had no bookable two-top on any date in their window, and 16 of those had none for a four-top either.
31 rooms are easier to book for four than for two. Twenty-six are easier for two than for four. Party size, not luck, explains a lot of failed attempts.
12 rooms logged open slots but never one between 7 and 8:30 PM, which is the window most people mean by dinner.
Booking windows range from 7 days to 31. Thirty rooms never let you book more than 14 days ahead, so a bot has nothing to race for on a date two months out.
Three cases, and we would rather name them than let you find out after paying.
You need a specific date that is months out. If a restaurant releases on the 1st for the following month, the useful thing is knowing the date rather than being told the moment it happens. The release table covers that and you do not need to pay anyone for it.
You genuinely cannot act on the alert. An alert is worth nothing if you are in meetings all day and the table is gone in four minutes. That is the honest case for an auto-booking service, and pretending otherwise would be a lie of convenience.
A large party. We measure parties of two and four, so a table for eight is usually a phone call to the restaurant rather than a platform problem.
The restaurant is not one of the 284 we watch. Our list is deliberately short, and a short list has an edge you can fall off.
A resale marketplace is not a faster way to get a table, it is a different transaction. Somebody books a reservation, then sells it, and you show up to a booking held in a name that is not yours. Restaurants increasingly cancel these on sight, and New York passed a law in 2024 aimed at paid third-party reservation resale.
That law is the cleanest line in this whole category, because it separates services that arrange or sell bookings from services that only watch and report. TableSwap publishes a price of $25 a month plus $75 per booking. We charge nothing per reservation, because we never make one.
The practical argument against paying is in the data rather than the ethics. Tables do come free at hard restaurants, often repeatedly: 7,048 separate release events across our list in at most 22 days of observation. What varies is which date and which hour, and that is exactly what an alert is for.
We check 284 restaurants across Resy, OpenTable, Tock and SevenRooms continuously, measure when each one releases tables, and tell you when something opens. We never ask for your booking login, never hold a reservation, and never complete a booking on your behalf. That is a real limitation, not modesty: it means you do the last step. It also means there is no account of yours at risk, which is the trade.
Browsing live availability is free and needs no account. A free account watches one restaurant and emails you when it opens. Plus is $14.99 a month, watches 30, shows the exact time slots that came free rather than a status, and extends the calendar view from 14 days to 30.
The measurements are not a marketing layer on top of the alerts, they are the same system. Every check that decides whether to notify you is also a data point about how that room behaves, which is why the numbers on this page can be checked against the restaurant's own booking page today.
Alerts tell you a table opened, bots book it with your account, and resale sells a table booked under another name. An honest comparison.
Reservation alerts, auto-booking bots and resale marketplaces compared for hard-to-book restaurants, including the account and legal risk of each.
Keep reading: Alert services, side by side · Are bots and resale legal? · When restaurants actually release
Browse all restaurant reservation guides
SnagRes checks 284 hard-to-book restaurants across four cities and alerts you when one opens. Watching one is free and no card is needed: watch a table free in New York City, or last-minute tables in NYC tonight, or see the hardest tables right now.