Direct Booking in the AI Era: Why the Direct Channel Is Strategic Again

A traveller with a suitcase crosses from a city square into a boutique-hotel courtyard while a host walks beside her and hands over a brass room key.

A traveller gives an AI assistant a brief: “Find a small hotel in central Lviv for three nights, with a quiet room, parking and late check-in. Keep the total below €700.”

The assistant compares hotel pages, maps, reviews and online travel agency offers. The traveller asks two follow-up questions, opens one hotel’s website and completes the booking there.

The hotel’s report may label this a direct booking. Yet the path was anything but purely direct. Demand may have been created by a friend’s recommendation, an OTA listing, an article, a map, a search engine and an AI answer. Attribute the entire outcome to the final visit and the hotel undervalues the channels that put it on the shortlist. Assume the direct site no longer matters and it gives another platform control of the final metres, the decision context and the opportunity to continue the relationship.

The direct channel is strategic again not because intermediaries are going away. Its value is that it gives the hotel a controlled place where intent becomes a verified offer, a reservation becomes a delivered promise, and a first stay can become a reason to return.

Direct booking is not the same as direct discovery

At least three events should be separated:

1. Discovery source: where the traveller first saw or remembered the hotel.

2. Influence source: what helped them compare, verify and trust the choice.

3. Completion point: where the reservation was confirmed and paid.

Those may be three different channels. A traveller finds a property on Booking.com, checks location and reviews on Google, asks an AI assistant a nuanced question, then books on the hotel site. The reverse also happens: a traveller discovers the hotel on Instagram and reads its site, but completes through an OTA because its rewards, payment method or trip management feel easier.

“Direct” therefore describes the commercial relationship and transaction point, not the whole demand journey. It can include the hotel’s website and app, phone, email, messaging and on-property sales, provided the hotel accepts the reservation itself—or through its chosen booking engine—rather than receiving the order from a third-party seller.

The evidence supports neither the death of direct nor the death of OTAs

SiteMinder’s 2026 Hotel Booking Trends, based on more than 135 million reservations across 20 markets, presents a restrained picture. In 2025, direct booking revenue share remained within 1.5 percentage points of the prior year in 95% of the markets studied. Direct again ranked among the top three revenue sources in 90% of markets. The same report nevertheless argues for a balanced mix and records the strength of OTAs and global distribution systems.

Hotel websites generated an average booking value of US$516 in the dataset, compared with US$312 for OTAs. That is not evidence that moving any booking to the direct site will increase its value by 65%. Direct guests may have booked different categories, longer stays or more extras; market and property composition also differ. The number describes the platform’s bookings, not a causal channel effect.

SiteMinder’s Changing Traveller Report 2026, a survey of almost 12,000 travellers across 14 countries, supplies the other half of the picture. Twenty-six per cent of respondents started accommodation research on an OTA, while 21% started with a search engine; 18% of those who began on an OTA ultimately booked directly with the hotel. This is vendor-sponsored survey research and stated behaviour, so it should not be projected mechanically onto every property. It does, however, illustrate a crucial point: research and purchase no longer belong to one channel.

What AI actually changes

Hotels once competed primarily for a place in a list of links or an OTA results page. Now part of the comparison can happen inside a system where the traveller describes the trip in natural language: “near the opera but not on a noisy street”, “suitable for a child with allergies”, or “arrival after midnight”.

Google’s own documentation says AI Overviews and AI Mode can break a complex request into related searches, compare sources and surface links in the response. It also makes an important corrective: there is no special AI-only optimisation requirement. A page must be indexable and useful to people, and its structured data must match what visitors can see.

A click is not guaranteed. Pew Research Center analysed 68,879 Google searches made by US users in March 2025. When an AI summary appeared, a traditional result was clicked in 8% of visits, compared with 15% without a summary; a source inside the summary itself was clicked in only 1% of visits. The study covers US Google use, not global hotel booking, but it demonstrates the risk: some comparison ends before any website visit.

AI can also route a traveller towards a direct transaction. In June 2026, IHG announced that its app in ChatGPT lets people explore more than 7,000 hotels with real-time availability, rates, maps and amenities, then guides the selected user to IHG’s direct channels to complete the reservation. This is one large chain’s implementation, not an off-the-shelf model for an independent hotel. The architecture is still instructive: discovery takes place in an external environment, while verified inventory and completion remain connected to the hotel’s own commercial system.

Not “direct versus OTAs”, but a portfolio of roles

OTAs are not merely a cost line. They create reach in unfamiliar markets, simplify comparison, provide familiar payment and support, run their own rewards programmes and invest heavily in demand generation. In its official “How we work” disclosure, Booking.com explains that it enables comparison, earns commission after the stay, and may give visibility to advertising and programmes that carry higher commission. This is a platform’s self-description, not an independent performance audit, but it clarifies what the hotel is buying.

A healthy strategy assigns each channel a job:

The objective is not to force direct share towards an arbitrary 100%. It is to build a mix that reaches the right markets, avoids buying the same demand twice and preserves an economically healthy share of relationships the hotel can manage itself.

Four reasons the direct channel is strategic

1. Control of the final promise

In its own journey, a hotel can present live availability, the final total, cancellation terms, the actual room configuration and relevant extras. It can carry dates and party details from the conversation into the booking engine instead of returning the traveller to the beginning.

That does not make the hotel site inherently more accurate. If rates, images and rules are manually maintained in several places, the direct channel may be the weakest one. Strategic control exists only when the page is backed by synchronised sources and accountable operations.

2. Visible economics rather than “zero-commission” theatre

A direct booking is not free. Its total acquisition cost may include:

• paid search, metasearch and social media;

• content, organic search work, photography and video;

• the booking engine, property system, guest relationship platform and analytics;

• payment processing, fraud prevention and refunds;

• a member discount, breakfast, upgrade or other direct benefit;

• staff time for messages, calls, changes and support;

• an allocated share of the repeat-guest or loyalty programme.

An OTA calculation should also go beyond its headline commission. Visibility programmes, hotel-funded discounts, payment conditions, cancellations and operational work belong in the comparison.

The more useful unit is contribution from a completed stay:

Collected room and ancillary revenue, less refunds, variable stay costs, all acquisition and distribution costs, and the cost of included benefits.

A paid direct click in a competitive market can cost more than an OTA commission. A repeat guest who responds to a lawful first-party message may have a very low incremental acquisition cost. Averaging those two journeys into one attractive number hides the decision the hotel needs to make.

3. First-party relationship context—within lawful boundaries

When the traveller interacts directly, the hotel can, with an appropriate legal basis, connect the enquiry, choice, reservation, pre-arrival request, stay and feedback. That makes it possible to avoid asking for known dates twice, understand abandonment reasons and prepare a more relevant next offer.

But “first-party data” does not mean “unlimited use”. The European Commission’s GDPR guidance requires a defined purpose, data minimisation, limited retention and compatible further use. Data required to fulfil a booking does not automatically authorise unlimited marketing surveillance or the training of any model.

The strategic advantage is not the largest possible profile. It is high-quality, explained and governed context that the guest can trust.

4. The ability to build repeat business—not a guarantee of loyalty

A direct relationship makes it easier to invite a return, offer something relevant or recover an unfinished choice. The hotel can recognise a returning guest and measure relationship value across time.

The booking point itself does not create loyalty. Guests return because the stay was right, the price felt fair, the process was dependable and recognition was useful rather than intrusive. OTAs also operate powerful rewards programmes. A poor direct experience will simply teach the traveller to return to an intermediary. Direct creates the opportunity to build a relationship; operational quality realises it.

When an intermediary booking can be better for the hotel

Direct does not need to win every transaction. An OTA reservation may be rational when:

• the hotel is entering a foreign market where it has little brand demand;

• the platform brings genuinely incremental guests into a weak period;

• its trust, support, currency or payment method removes a traveller barrier;

• the hotel’s own media cost for that segment exceeds the full distribution cost;

• the direct site cannot match the OTA’s availability, clarity or convenience;

• the traveller values platform rewards, a flight package or one place to manage the trip.

Closing OTAs before replacing their demand is poor strategy. So is repurchasing the hotel’s own repeat guests through an intermediary because the hotel never gave them a simple direct route back.

Rate parity: the rules have not changed in the same way everywhere

The EU Digital Markets Act changed the position in the European Economic Area. The European Commission’s April 2026 staff working document states that Article 5(3) enables business users to differentiate price and conditions on their direct channels. It also records that Booking Holdings reported waiving most-favoured-nation clauses and removing parity references from Booking.com’s EEA partner terms.

This is not a universal licence to ignore contracts. The rules vary by country, platform, visibility programme, rate type and contract date. A hotel should:

1. review current agreements and local law rather than rely on a generic post;

2. define public, member-only, packaged and personalised offers separately;

3. synchronise availability, taxes, charges and material terms so the comparison is fair;

4. promise a “best rate” only if it has a process that can keep checking the claim.

A direct advantage does not have to be a lower headline rate. It can be the right room category, a flexible change, an included service, a better-confirmed requirement or a shorter route to help.

What the direct channel needs in the AI era

One operational source of truth

Rates, availability, restrictions, taxes, policies and room attributes need defined systems and last-verification times. AI does not repair a discrepancy between the site, booking engine and front desk; it distributes the discrepancy faster.

Pages that machines and people can verify

Material facts should be available as text, not trapped only in an image, PDF or widget. Pages should be indexable and structured data should match visible content. This is ordinary technical and editorial discipline, not a secret optimisation trick for answer engines.

Completion without repetition

The mobile journey needs a final total, currency, material terms, secure payment and an intelligible confirmation. Dates, party, category and rate already established in a conversation should carry forward when technically and lawfully possible.

A person at a defined boundary

Connecting rooms, accessibility, allergies, group stays, refund exceptions and payment failures need an accountable staff member. The handoff should contain the context, unresolved question and response deadline—not merely a notification in a shared inbox.

Post-booking delivery

If a direct reservation is difficult to change, nobody answers the guest and the promised benefit never reaches operations, the hotel trains that guest to choose an intermediary next time.

How to measure the channel without fooling yourself

A single “direct booking share” metric is dangerous. It can rise because a discount destroyed contribution or because OTA demand fell and rooms remained empty.

A minimum scorecard should include:

• share of stayed room revenue by channel;

• total acquisition cost per completed booking;

• contribution per occupied room night after variable and distribution costs;

• booking value, length of stay, lead time, cancellation and refund by channel;

• progression from verified offer to booking start, confirmation and stay;

• new and returning guests separately;

• 90-, 180- and 365-day repeat booking by cohort;

• the share of bookings whose discovery source is unknown.

Google Analytics’ official attribution paths report can show channels that initiate, assist and close the route to a defined event. That is more useful than last-click reporting, but it is not complete truth: cross-device behaviour, consent choices, calls and AI answers leave gaps. direct / none often means “source not identified”, not “the brand generated this demand unaided”.

Joining analytics to the property system requires a unique transaction identifier. Google’s event reference recommends transaction_id to prevent duplicate purchases and provides a new-versus-returning customer field. The revenue source of record should still be confirmed, uncancelled stays in the hotel’s operational and financial systems—not only a browser event.

A practical 90-day plan

Days 1–30: establish reality. Reconcile channel reservations to completed stays. Add media, technology, discounts, payment cost and labour—not only commission. Separate new and returning guests. Review parity and current agreements.

Days 31–60: repair five high-value journeys. Examples include exact dates and total price; family or pet requirements; late arrival; two-room requests; and reservation changes. Each needs a verified source, a substantive answer, one next action and an accepted human handoff.

Days 61–90: run controlled tests. Test a specific hypothesis, not a blanket “book direct” discount: a pre-filled page; a value-add without a lower rate; a lawful message to former guests; better factual content for one source market. Evaluate completed stays, contribution and repeat behaviour—not clicks alone.

Where Greetio fits

Greetio should not manufacture direct demand or replace OTAs. Its role is to remove the break between intent and the next action: identify the decision condition, retain dates and requirements already given, prepare an answer from verified rules, carry context into the booking engine or to an accountable staff member, and record the sequence to a confirmed outcome.

If the system cannot access a live price, availability or guarantee, it should not invent one. The correct response is to state what requires verification, collect only what is necessary and route the case to someone who can take responsibility.

Conclusion

AI does not make the direct channel obsolete. It makes the route to it less linear.

A traveller may discover a hotel on a platform, validate it in search, ask a system to compare conditions and only then move to booking. In that environment, the hotel does not need to own every discovery moment. It does need to be an authoritative source, provide a strong completion path and preserve the ability to learn from a relationship that has already begun.

The strategic question is not “How do we get rid of intermediaries?” It is:

Which channel brought an incremental guest, what did the completed stay cost to acquire, who controlled the promise—and did the stay make it easier to return?

When a hotel can answer all four parts, direct booking stops being a supposedly cheaper cell in a distribution report. It becomes a governed part of distribution, service and long-term guest value.