Direct bookings fell to 51.3%. The phone is why.

European hotels booked 51.3% of their overnights directly in 2025, down from 57.6% in 2013, according to the HOTREC European Hotel Distribution Study 2026. That is the figure the trade press will quote. The more useful number is one row further down the same table: over those twelve years, real-time booking through a hotel's own website went from 7.4% of overnights to 13.9%, while the telephone fell from 20.5% to 14.1%. The booking engine did not win share from the OTAs. It replaced the hotel's own switchboard, almost point for point.

That matters because it changes what the decline is evidence of. Direct distribution did not lose because hotels failed to modernise; by the study's own measure they modernised exactly as advised, and the reward was to stand still. What direct lost, and did not replace, was every channel where a guest could ask something before booking.

The study was produced by Professor Roland Schegg of HES-SO Valais-Wallis with HOTREC, the European umbrella body for hotels, restaurants and cafes. Fieldwork ran February to May 2026 for the reference year 2025, covering 2,713 hotels: 1,882 individual responses from 28 countries, plus aggregated data from 831 chain properties. The European figures below are weighted by Eurostat overnights and, unless stated otherwise, exclude the chain data.

The booking engine replaced the phone, almost point for point

Between 2013 and 2025 the six direct channels moved in opposite directions. Real-time booking with availability check on the hotel's own site rose from 7.4% to 13.9%. Telephone fell from 20.5% to 14.1%. Walk-ins almost halved, from 5.8% to 3.0%. Email drifted from 14.9% to 13.5%, mail and fax from 3.2% to 1.6%, and the website contact form without an availability check from 5.8% to 5.2%.

Add them up and direct goes from 57.6% to 51.3%, a loss of 6.3 points. The booking engine gained 6.5 over the same period. Every point the engine won was spent covering the phone's retreat: the two net out at a rounding error, and the 6.3-point shortfall is what the remaining four channels gave away between them.

What direct lost was the conversation, not the transaction

Group the channels by what a guest can do on them rather than by who owns them, and the pattern sharpens. The three channels where a guest can ask a question in their own words - phone, email, and mail or fax - fell from 38.6% of overnights in 2013 to 29.2% in 2025, a loss of 9.4 points. Over the same twelve years, OTA share rose 10.2 points, from 19.7% to 29.9%.

A 9.4-point fall in the channels that carry a question and a 10.2-point rise in OTA share are close enough to be suggestive and not close enough to be proof. They are different guests, different trips and different years, and the study measures where a stay was booked, not why. But the shape is consistent with something hoteliers describe without needing a study: the guest with a question used to call the hotel, and now searches instead, and the search ends on a platform that answers in a schema rather than a sentence.

The study has no row for WhatsApp

The channel table lists sixteen routes to a hotel room. None of them is a messaging app. The nearest column, social media channels, was 0.4% in 2013 and 0.5% in 2025 - flat for twelve years, which does not match what properties running a WhatsApp number actually see. Across the properties vGuest serves, 97% of guest conversations arrive on WhatsApp.

The reason for the gap is structural, not an error in the study: it counts the channel a booking was completed on. A guest who asks about a cot on WhatsApp, gets an answer in forty seconds and then books on the hotel's own site is recorded, correctly, as a booking-engine booking. The conversation that decided it leaves no trace in the data. So the 13.9% booking-engine figure is real, but reading it as proof that direct demand has gone self-service is a conclusion the numbers cannot carry - and a hotel that draws it will keep investing in the form while neglecting the questions guests actually send before they book.

Chains run a different distribution business

The chain sample - ten European chains, 831 hotels, 128,154 rooms across twelve countries - looks nothing like the independents. Direct is 45.1%, lower than the 51.3% European figure. OTAs are 29.4%, roughly the same. The difference is GDS at 11.2% against 1.9% in the overall sample, reflecting corporate and business travel distribution, and a phone share of 17.5% that is higher than the independents', not lower.

It is worth holding that next to the usual advice. The operators with the largest technology budgets in European hospitality did not answer OTA pressure by removing the telephone. They added a third distribution system and kept it.

Half of hotels report the OTAs undercutting their rates

The pressure the study documents is not only about share. 51% of hotels report OTA rate undercutting and 44% report multi-sourcing, where their rooms appear through channels they did not contract with. Concentration makes that hard to answer: Booking Holdings takes 68.8% of European OTA bookings and Expedia Group 16.6%, so 85.4% of the market sits with two companies, and Booking.com alone is 66.1%. The movement that remains is between them rather than towards anyone new: Expedia Group has risen from 12.5% in 2021 to 16.6% in 2025.

Country spread is wide. Italy has the highest direct share at 64% and Finland 62.7%, while Hungary is the most OTA-dependent at 39.8% and Spain at 35.2%. On the technology side, 66% of hotels use a channel manager and 58% connect to meta-search, but only 26% use AI in any form - the one adoption figure in the study with real room left in it.

Where this reading breaks

Four caveats, because reading a falling phone line as a lost conversation is an interpretation of a table, not a law.

Overnights are not money. The study measures share of overnights, so a direct point and an OTA point count the same here even though they differ by the commission. A hotel that loses share while raising net ADR may be doing well by the only number that pays wages.

The sample is self-selected. 1,882 hotels across 28 countries is a substantial response, but the countries are not equally represented and hotels that care about distribution are likelier to answer a distribution survey.

The phone-to-engine swap is an aggregate, not a traced path. Nobody followed a guest from a call in 2013 to a booking form in 2025. Two lines crossing on a chart is a starting point for a question, not an answer to one.

And answering faster does not, by itself, move share. A property with no booking engine and an excellent WhatsApp reply still loses the reservation to Booking.com, because the guest has to be able to finish. Conversation is what the study shows direct losing; it is not a substitute for the transaction it also has to win, and the levers that do move the mix are a shorter list than most advice suggests.

What to do first

For the next two weeks, log where each direct booking's first contact happened, not where it closed. A booking that started as a message, a call or an email and finished on your site is a conversation your engine merely recorded, and if you count it as self-service demand you will fund the wrong thing. Once you know the real ratio, you can decide whether the gap in your own mix looks more like the phone line the European sample lost, or the part that happens after a guest's message arrives.