Is it cheaper to book directly with the hotel? What 405 luxury hotels showed

Against the big booking sites, it is cheaper or the same more often than not. Read from each hotel’s location, the hotel’s own site beat every big site listed on 217 of 838 luxury stays and matched the cheapest of them on 279. That makes 496 stays, or 59%. On the other 342 at least one big site was cheaper. The 6 are Booking.com, Expedia, Hotels.com, Agoda, Priceline and Trip.com. Search from the United States and the same count falls to 243 of 652 (37%), so where you sit matters.

Bring in every seller on Google, small discount resellers included, and the answer turns round. The hotel’s site was lowest or level on 159 of 853 stays (19%), and on the other 694 somebody showed less. Across all 853 the cheapest seller of any kind stood a median 9.4% under the hotel, which means half the gaps were wider than that and half narrower.

Both of those counts take each seller’s lowest price, whatever room and terms came with it. A true match needs a seller that names its room, and on Google that in practice means the big sites: no discount reseller is in this comparison. On the same room with the same kind of cancellation, the hotel was lowest or level on 266 of 403 comparisons (66%).

If a friend asked me, I’d say this. Look at the hotel’s site first, and don’t pay a big site more than that. Anything cheaper than both is a lead to check, and it stays a lead until you’ve read what it sells.

These figures cover luxury hotels from late November to New Year, and they do not describe hotels in general. On 31 October 2026 we read Google Hotels for 1,125 stays at 405 luxury hotels in 38 countries and territories, for 2 adults, in US dollars, from each hotel’s location. Each hotel had up to 3 stays: a weeknight and a weekend night, mostly in early December, and 3 or 7 nights between Christmas and New Year. Check-ins run from 21 November to 28 December 2026 and the last check-out is 2 January 2027.

The Maldives weigh on the result, so here is the same count by place.

Outside the Maldives the hotel’s site was lowest or level against every seller on 56 of 525 stays, or 11%.

Other hotels give other answers. Altroconsumo, the Italian consumer group, published a survey on 27 April 2026 of 3-star hotels in 10 Italian destinations, for a family of 4 in the week of 22 June 2026. The hotel’s own price, online or by phone, was lower than the platforms on 74 of 90 bookings (82%). It is a different sample with a different question, and both results can be true.

Agoda, by a little: it was cheaper than the hotel on 146 of 587 stays (25%), and the other 5 on 18% to 23%. The hotel’s site matched each of them on more than a third of stays, and the typical gap to every one of the 6 was zero.

The cheapest of the 6 beat the hotel far more often than any one of them, on 342 of 838 stays, because the site that wins changes from stay to stay. That is the reason to open 2 or 3 of them and never only 1.

Matching the room changes who wins. Of the 403 comparisons, the hotel was cheaper than every other seller of that room on 143, level with the cheapest on 123 and dearer on 137. Hotels.com undercut it on 65 of 146 rooms and Agoda on 57 of 168, well above their showing in the table, while Expedia did so on 13 of 90 and Booking.com on 13 of 73. Those rooms come from 97 stays at 50 hotels, so the 66% is a soft number: the true figure could sit anywhere from 49% to 84%.

Here are 3 real stays, so you can see what these comparisons look like.

Per night for 2 adults. Prices from Google Hotels, 31 October 2026. The Paris prices include tax; the Coronado prices are before tax with the hotel’s compulsory fee included. A room’s title can leave out meals or other extras.

At Le Bristol the hotel’s site stood $371 a night under both big sites. At The Peninsula, Expedia stood $103 under the hotel and Hotels.com $36 over it, on the same stay, although one company runs both.

Free cancellation on the same room cost a median 12.0% more. We compared 2,419 pairs from one seller for one room, and the rate with no free cancellation stated was the cheaper one on 2,401 of them.

The Hotel del Coronado shows it in the table above. $670 buys the room with no free cancellation stated, $715 lets you cancel until mid-December, and $817 or $833 lets you wait until about 20 December. That last rate is also the one Coronado price where the hotel’s site was cheaper than Expedia.

Set that 12.0% beside the 9.4% by which the cheapest seller undercut the hotel across those 853 stays. The 2 sizes are close. That proves nothing about any one price, because a reseller’s row on Google rarely states its terms. It does mean a lower figure can be a different promise, and you can’t call it a saving until you’ve read the promise.

On Google’s list, yes, and what you would actually get for that price is the part the list leaves out. We took the 10 resellers with at least 100 comparable prices that typically stood 5% or more under the big sites and the hotel. Of the 5,540 prices they showed in our reads from each hotel’s location and from the United States, 52 named a room. The 6 big sites named one on 2,101 of 11,145.

That gap comes partly from how Google’s list works. Rooms appear only where a seller pays for an ad: on the 1,125 stays, all 2,100 rows naming a room were ads. So a bare price on Google does not prove that the seller’s own page hides the room. It means the list can’t tell you, and we did not open those pages.

Low prices also move. We read 805 stays a second time, about 70 minutes later, as a searcher in the United States. They were not a random sample. Those 10 resellers had shown the lowest price on 611 occasions, and 120 of those rows had gone from the list. For the 6 big sites it was 11 of 203, and for hotels’ own sites 2 of 46.

Nothing in this read was booked, so we can’t say whether any low price can be bought at that figure, and we can’t say it can’t. I’d treat a low price from a name I don’t know as a lead and open it before I believed it.

Google’s big price is usually the lowest on its list, whoever is selling it. It equalled the lowest row on 1,002 of 1,110 stays and was never below it. On 494 of 1,070 stays the cheapest of the 6 big sites or the hotel’s own site stood more than 5% above that headline, and across all 1,070 the typical gap was 3.3%. For the Hotel del Coronado on 28–31 December the headline was $618, while the hotel’s site and all 6 big sites showed $670.

Tax adds a second trap. In our read, prices for hotels in the United States and Canada came before tax with the compulsory hotel fee included, and prices everywhere else came with tax in, although a few sellers broke that rule. In the United States the FTC’s rule on fees, in force since 12 May 2025, makes the total price include the fees the business knows about, and lets taxes be shown later, before payment. So the only figure worth comparing is the total on the last page before you pay. We have written separately about how hotels appear on Google’s results page.

Compare totals on the last page before payment, room against room and term against term. These 5 steps cover it.

A booking site is the better buy when it is cheaper on the same room and the same terms, and the tables above show that happens. Three other cases can tip it.

Hotel points pull the other way. Marriott’s rules class a room booked through Expedia, Priceline or Booking.com as non-qualifying, with no points and no elite benefits. Hilton’s terms exclude third-party website bookings “irrespective of rate paid”. If you hold status at a chain, count that before you count the dollars.

Asking worked at Italian 3-star hotels, the one place we have seen it measured, and it costs you a phone call. In Altroconsumo’s survey of 90 of those hotels, 53 gave a price by phone. Of those, 32 came in below the hotel’s own website, 15 matched it and 6 were higher. The other 37 didn’t answer, sent the caller to the website or would only quote by email.

We did not phone anyone, so I can’t tell you how a luxury resort would respond. I would still ask whenever the stay is long, when there are 2 rooms, or when I want something a website can’t sell, such as connecting rooms. Have the numbers from the steps above in front of you and ask 3 things: the total for that room, the last day to cancel for free, and whether they can do better than the price you see on a big site. A hotel doesn’t have to say yes. Get the answer in writing before you cancel anything.

The limits are real, and each one changes what you can take from the numbers.

In this luxury read, hotels’ own sites stood closer to the big sites than Google’s list makes it look, and the gap that hurts is the one a guest sees first. Guests meet the headline, and on 694 of 853 stays somebody showed less than the hotel. Look at your own Google list as a guest does, for 3 dates, with who sells below you and on what terms in mind.

Start with the 342 stays of 838 where a big site undercut the hotel, because that is a gap you can look into, and it costs you commission on top. Then write on your own site what a guest will compare: the room’s name, the total, the last day to cancel, what comes with it. On 258 of 1,125 stays Google listed no row from the hotel at all, so a guest comparing there had no hotel price to look at.