Eighteen percent of the travellers who start their search on an OTA end up booking with the hotel directly anyway. That number went up 3.3 percentage points last year, according to SiteMinder's Changing Traveller Report 2026, which surveyed 12,000 travellers across 14 countries.
Read it the right way round. Those travellers are not won by a marketing campaign. They are already looking for you, deliberately, after seeing you somewhere else. Whether they finish the booking on your site or give up and go back to the OTA comes down to what happens in the next four minutes.
Most independent hotels spend their direct booking budget on getting more people to look. The bigger number is sitting in the people already looking.
Globally, 28% of travellers now book direct. In the US it is 40%, up from 36%. Both figures come from the same SiteMinder research, and both are a long way above the 36.6% direct share that independent hotels actually capture, per Cloudbeds' 2026 State of Independent Hotels Report. Travellers want to book direct more than independents are set up to let them.
This piece is about closing that distance. We covered where the revenue leaks in the seven handoffs, including the full commission arithmetic for a 20-room property. This is the other half: the five things that move share, and why running them separately is what keeps them from working.
Lever 1: your website has to be a shop, not a brochure
The traveller arriving from an OTA listing has already seen your photos, your rating and your rate. They came to your site for one of three reasons: to check you are real, to see whether direct is cheaper, or to find something the listing did not tell them.
A brochure site answers none of those. It has a hero image, an About page, a Rooms page that lists room types without prices, and a "Book Now" button that opens a booking engine in a different font.
A site built to convert answers all three above the fold: live rates, live availability, and a visible reason to book here instead of there. SiteMinder found 58% of travellers now plan to choose superior or luxury rooms, up four points year on year. If your site does not show the difference between your standard double and your corner room with the balcony, you are selling the cheapest thing you have to people who arrived willing to spend more.
The uncomfortable version: for most independents, the booking engine is the weakest page on the site, and it is the only page that takes money.
Lever 2: make sure what travellers find about you is current, wherever they find it
Hotel discovery fragmented fast. SiteMinder's data on where travellers start their research:
| Starting point | 2026 | Prior year |
|---|---|---|
| OTA | 26% | 18% |
| Search engine | 21% | 36% |
| Word of mouth | 14% | 7% |
| Familiar hotel brand | 7% | 3% |
| AI assistant | 4% | 1% |
Search engines lost 15 points in a year. AI quadrupled off a small base, faster among Gen Z and Millennial travellers.
Be careful what you conclude from that. Four percent is not a channel worth restructuring around, and we are not going to tell you it is. The useful reading is the pattern across the whole table: there is no longer one place where travellers form their impression of your hotel, and you control almost none of them.
What you do control is whether the information they find is right. Breakfast at €18 when your site still says €12. A pool listed as closed that reopened in March. Two renovated rooms that exist in the PMS and nowhere else. Every one of those sources, the OTA listing, the directory page, the AI assistant, the friend who looked you up, is reading something you published once and never went back to.
This is a smaller claim than "get found by AI," and it is the honest one. Being accurate everywhere will not fix an awareness problem. It fixes the version of you that travellers meet before they meet you.
Lever 3: the pre-arrival window is the longest stretch you get, and most hotels leave it empty
Travellers book around 40 days ahead globally, 47 in EMEA. That is six or seven weeks with someone who has already committed money to you, and in most independent hotels it contains one automated email: the confirmation.
The reason is mechanical. The email tool has a name and a date. It does not know they booked the small double, so it cannot offer the suite. It does not know they arrive at 22:00, so it cannot send late-arrival instructions. So it sends everyone the same message, everyone ignores it, and nobody writes a second one.
Give the same tool the reservation and the window changes character. At C-Hotels Silt, a 73-room seafront property in Middelkerke, fewer than 6% of arriving bookings received any message across an eight-month window before launch. Breakfast and parking were sold at the desk, in the ninety seconds of check-in, if the person on shift remembered to ask. After launch, more than 65% of arrivals were reached, and the property booked over €24,000 of upsell revenue in three months: €15,000+ breakfast, €8,000+ parking, 450+ orders written straight back into the PMS. That is €12 or more per arrival, spread across every arrival rather than concentrated in the few who got asked.
Nothing about that requires more staff. It requires the message to know what the person booked.
Lever 4: messaging is where the direct relationship is actually built
Two-thirds of bookings are one or two nights. For most of your guests, the in-stay window is a single evening. There is no second attempt at it.
At C-Hotels Andromeda in Ostend, more than 90% of guest messages have been answered without a person over eleven months, 6,500+ of 7,500+, at a median under ten seconds. Over 500 parking reservations were placed by guests themselves, with no staff involvement. Parking is the single most common question at a coastal hotel, and it had been answered at the desk one guest at a time.
At Yalo Ghent, 92 design rooms with a 24/7 brand promise and a normal-sized team, 750+ guests are now in conversation with the hotel in a month against roughly 30 before. Same team, no new hires.
The point is not the deflection rate. It is that a guest who got a real answer at 23:40 in their own language has a relationship with your hotel rather than with the app they booked through. That is the thing you are buying back from the OTA, and it is worth more than the commission line.
Lever 5: the stay has to end somewhere other than a review request
Most post-stay automation is one email to everyone on the departures list, including the guest whose shower broke on Tuesday.
Yalo runs the post-stay flow harder than either other property, and asks every guest while there is still time to act on the answer. More than 85% of rated stays score 9 or 10, 650+ of 750+ responses since May. Roughly one in ten score 6 to 8, and barely one in a hundred score 5 or below. Happy guests get prompted to say so publicly. Unhappy ones get found before the review site does.
Then there is the return. When that guest comes back in September, they open whichever route they remember. If the only thing that survived check-out was a row in the PMS, that route is the OTA, and you pay commission a second time to reacquire someone you already had.
The levers are not five projects. That is the whole argument.
Here is what happens when you buy them separately, which is what most independents have done.
| Moment | Five tools | One record |
|---|---|---|
| Breakfast price changes | Updated in the CMS, the email platform, the upsell tool and the chatbot, on four different days | Changed once |
| Pre-arrival offer | Generic, because the email tool has a name and a date | Based on what they actually booked |
| Guest asks at 23:40 | Answer depends on who is on shift | Same answer as the website |
| Upsell accepted online | Logged in the upsell tool, invisible at the desk | In the PMS, visible to the person checking them in |
| Post-stay | Review request to everyone | Informed by what happened during the stay |
Every lever above depends on knowing something the previous stage learned. The pre-arrival offer needs the booking. The in-stay answer needs the website's version of the truth. The post-stay message needs the stay. Run them in five systems and each one starts from a name and a date, which is why each one underperforms on its own and why the combined result is worse than the sum.
The fix is not more integrations. Integrations move a fact between two places that both still hold it, and one of them is always stale. The fix is that the fact lives once, and everything a traveller sees reads from it.
What this does not do
It does not get you found. Accurate, current information everywhere makes sure what people find is right; it does not create demand that was not there. If nobody has heard of you, this is not the fix.
It does not remove OTAs, and it should not. They are a demand channel, they work, and 26% of travellers now start there. The goal is that they stop being the only route in, and that the 18% who want to come to you directly can actually finish.
You do not need five tools to run five levers. Describe your hotel, and AskWhisper builds the website, then runs the messages, the upsells and the post-stay follow-up from the same record. Change a price once and it is right everywhere a traveller looks.
Turn lookers into bookers. Just Ask Whisper.
