Your Loyalty Program Is Invisible to AI Travel Planners — and That's Where Direct Bookings Die

The short answer

Only 38% of AI-assisted hotel bookings complete on the hotel’s own channel. 44% land on an OTA, and 8% complete inside the AI tool itself — according to an NPS Prism and Dynata study of more than 2,300 US travelers who used AI to plan or book a trip (2026). The mechanism is structural, not accidental: AI travel planners build their shortlist of two or three hotels from what they can see from the outside — price and availability — and your loyalty program is precisely the kind of asset that pays off only inside your app, after the choice is already made. Four in ten travelers who booked through AI say it changed which hotel they picked. The scale is no longer hypothetical: Phocuswright (March 2026) puts 56% of active US travelers using AI for planning, booking or in-trip assistance in the previous twelve months, and Adobe Analytics measured AI-referred traffic to US travel sites up 194% year over year in May 2026. The fix is not abandoning loyalty — it is making loyalty legible to machines. With Google’s agentic hotel booking now live in AI Mode (Skift, August 27, 2026) and loyalty pricing entering the answer layer, hotels have a narrow window to do exactly that.

Where AI-assisted bookings actually complete

The NPS Prism/Dynata study is the cleanest look yet at what happens after the recommendation. The distribution tells the story:

Source: NPS Prism + Dynata, survey of 2,300+ US travelers who used AI to plan or book travel, 2026.

Read that middle row again. The traveler asks an assistant for a hotel, the assistant names properties and the transaction settles on an OTA more often than on the hotel’s own website. The recommendation happens in AI, the transaction happens on a platform, and only the stay happens with the hotel. Every step away from your channel is a step away from the enrolled member, their first-party data, and their reason to return — which is why the 8% in-tool figure deserves attention too. It is the smallest slice today, and it is the one Google is actively trying to grow: after confirming tests on August 7, 2026, Google launched agentic hotel booking inside AI Mode in the US on August 27, 2026 — travelers chat to find and compare rooms, select, review cancellation terms, and pay with Google Pay in one place (Skift). The hotel or OTA still charges the card and sends the confirmation; the launch partner list — Booking.com, Expedia, Hilton, IHG, Marriott, Choice, Wyndham, Priceline, Hotels.com, Trip.com — is a roll call of chains and OTAs.

Why loyalty loses when the AI builds the shortlist

Hotels and OTAs have fought over the booking journey since the mid-2000s, and loyalty tiers, app experiences, and direct-booking perks were the weapons of that fight. All of them share one assumption: the guest arrives already choosing between you and a short list of rivals. For the growing share of travelers planning with AI, that assumption breaks upstream — the assistant assembles the shortlist before your loyalty value proposition ever gets to compete.

The reason is visibility, plain and simple:

• AI selects on comparability. Price and availability are structured, crawlable, and comparable across every property at once. Loyalty benefits are not.
• Loyalty pays off too late in the journey. Member rates, tier perks, and points only register inside your logged-in environment — after the hotel has already been chosen. NPS Prism calls it the industry’s invisible advantage.
• Trust accrues to the recommender. Three-quarters of AI users in the study said results met their expectations. That trust accrues to the AI and the platforms it surfaces — not to the property that delivers the stay. A guest who trusts AI to pick a good hotel has one less reason to stay loyal to any single brand.

None of this runs against traveler preferences — it exploits them. More than half of AI travel users cite comparing options and saving time as their motivation; personalization ranks far lower, cited by only about a third. The AI layer is an efficiency machine, and efficiency machines route to wherever the comparison is cheapest. Kantar (2026) already counts 40% of global travelers turning to ChatGPT, Claude, Gemini, and Perplexity for travel answers. OTAs line up inventory on exactly AI’s terms; your strongest differentiation is structurally unreadable to it.

The quiet margin problem nobody prices in

There is a second, less-discussed cost beyond channel share: AI-compressed booking paths skip the merchandising steps. The upgrade, the package, the dining reservation, the spa slot — those are the moments where a hotel sells its margin. When the assistant compresses search-to-reservation into a single conversational flow, the in-between steps get dropped. In the NPS Prism study, 23% of travelers who booked hotels through AI reported feeling they missed out on extras and add-ons — and a hotel with no presence in that flow has no way to put the offers back in front of the guest. So even the direct slice of AI-era bookings arrives thinner than the equivalent booking from your own website, where you control the attachment path. A loyalty program, surfaced at the right moment, is the natural merchandising vehicle — but only if the AI layer can see it exists.

Loyalty is entering the AI layer — without most hotels in it

Two developments in the past weeks changed the loyalty question from theoretical to urgent.

First, Google’s AI Mode booking now carries loyalty economics. The August 27, 2026 rollout isn’t just a payments flow — Google’s travel stack update includes live loyalty-based points-and-miles pricing for hotels, surfacing member-relevant pricing directly inside AI answers (Skift, August 2026). For chains already in the launch partner set, that means their loyalty programs are becoming machine-visible through Google. For everyone else — independent hotels, small groups, brands outside the partner list — the gap is widening: the largest players’ loyalty value props are entering the recommendation layer while yours stays in the app.

Second, loyalty data is becoming portable into assistants. Tools like Gondola, an MCP integration now live with ChatGPT, pull a traveler’s loyalty accounts, past trips, and preferences directly into the assistant’s context (The Points Guy, 2026). When the guest’s status lives inside ChatGPT, the assistant can reason about “your IHG points” or “your usual Marriott in Milan” mid-conversation — which is powerful for enrolled members of big programs, and another layer of invisibility for properties with no machine-readable loyalty presence at all.

The strategic irony is sharp. NPS Prism frames loyalty as “the one asset agents can’t take” — the data an agent has to come to you to get. That’s true only if your loyalty program is structured as a data and channel asset. If it’s a PDF of tier benefits behind a login page, it’s not an asset in the AI era. It’s a cost center with better branding.

What Marriott and IHG are doing — and what it means for everyone else

The strongest chains are already reframing loyalty as AI infrastructure rather than a discount ladder:

• Marriott is using AI to pull travelers deeper into Bonvoy, rolling out natural-language search across its own site and app rather than ceding the conversational interface to third parties (Hospitality Net, 2026). The logic: if the conversation is where the choice happens, own a conversation.
• IHG is consolidating loyalty and customer data into a single platform with an AI-powered content system designed to anticipate guest needs rather than react to them (Forbes, February 2026). The logic: the more an external agent needs your first-party preference data to personalize, the more leverage your channel retains.

Both moves share the same read of the market NPS Prism’s data supports: an AI agent that wants to book a genuinely suitable stay has to source the guest’s preferences from somewhere. A hotel that makes the agent come to it for that data keeps the agent dependent. A hotel that lets the data leak to platforms lets the agent learn the guest once and never come back.

You do not need Marriott’s budget to apply the principle. You need loyalty content that machines can read.

The playbook: make loyalty legible to AI

1. Publish your member-rate logic as crawlable content. A plain-language page — “what booking direct gets you” — with the actual benefits: best available rate, room selection priority, late checkout, the welcome amenity. Not behind a login, not as a PDF. This is the single highest-leverage fix, because it converts your invisible advantage into answer-layer content. The mechanics overlap heavily with the homepage conversion fixes in “AI Referral Traffic to Hotels Doubled in 11 Weeks — Here’s How to Convert It into Direct Bookings”.

2. Get your program into structured data. Schema for offers, loyalty program membership benefits, and rates gives assistants and AI Overviews a machine-readable version of your value prop. Structure is what makes you comparable on dimensions beyond price.

3. Keep your direct-rate story consistent everywhere. Assistant answers are assembled from your site, your Google Business Profile, and OTA listings. If the member rate story only exists on one surface, the model treats it as noise. We covered the same consistency principle — and the cost of stale rates — in “AI Hotel Price Accuracy and Rate Parity”.

4. Turn first-party data into the thing agents must request. Stay history, stated preferences, tier benefits — these are the elements an external agent cannot route around. Fund the parts of your program that generate them, not the parts any OTA can replicate with a coupon.

5. Compete where AI is weak. Value and speed are a contest the platforms are structured to win. Loyalty-aware service, recognized preferences, and human recovery when things go wrong are the ground AI does not own. Make sure your direct channel demonstrably delivers them — that’s the retention story the shortlist can’t match.

6. For independents: build the loyalty-equivalent. No Bonvoy? Your version is a direct-booking perks program — fifth night free, room-upgrade priority, a named host — documented in plain text on your site. Twenty-three percent of AI bookers already feel they missed extras; a visible perks program positions you as the property where booking direct restores them. This is also how you appear in answers shaped by loyalty pricing, where every chain is suddenly louder.

7. Measure the loyalty question in your AI visibility work. When you audit which prompts surface your property, audit which value proposition surfaces with it. If assistants describe your hotel in price-only terms while describing competitors with their programs, you’ve found the exact content gap to close. A travel AI audit answers this prompt by prompt, and the full AI visibility program for hotels builds the content system that fixes it.

What tour operators and DMCs should take from this

The loyalty problem has a direct analog for operators: your repeat-guest relationship is invisible at exactly the moment AI builds the shortlist. Assistants compare operators on price, duration, and review volume — structured fields. Your return-client rates, local partnerships, and guide tenure — the reasons a traveler rebooks you — live in relationship land, unreadable by comparison engines. The fix is the same shape: publish the proof as citable content (repeat-client percentage, years operating, named local partners), one page per high-intent experience, so the assistant shortlisting “private tours in Lisbon” has your differentiators in front of it rather than only your price. Our travel GEO agency program builds this page system for operators and DMCs.

FAQ

Do AI travel planners actually change which hotel gets booked?

Yes — and measurably. Four in ten travelers who used AI to book a hotel say it changed their choice of property, not just saved them time (NPS Prism/Dynata, 2,300+ US travelers, 2026). The assistant typically narrows the field to two or three options selected on price and availability, which is why the booking then completes on an OTA 44% of the time.

Can guests book hotels inside AI tools yet?

Yes, in the US. Google launched agentic hotel booking in AI Mode on August 27, 2026 — travelers compare rooms, select, review cancellation terms, and pay with Google Pay in one flow, with the hotel or OTA still charging the card (Skift). Only 8% of AI-assisted bookings complete inside AI tools today (NPS Prism/Dynata, 2026), but that share is the one every platform is now engineering to grow.

Why can’t AI assistants see my loyalty program?

Because most loyalty value lives behind a login or in app-only context. AI assistants assemble recommendations from publicly crawlable, structured information — rates, availability, reviews. Member pricing, tier benefits, and points redemption don’t exist in that layer, so they can’t factor into the shortlist. Publishing them as crawlable, structured content is the fix.

Should independent hotels build a loyalty program for the AI era?

Build the lightweight version: a documented direct-booking perks program on your public site. It gives assistants a non-price reason to recommend you, gives the 23% of AI bookers who miss add-ons a place to land, and costs a fraction of a points economy. Pair it with the direct-booking conversion tactics that keep the referral on your channel.

How do I check whether AI presents my loyalty value proposition?

Audit the prompts a traveler would actually use — “best boutique hotel in [city],” “where to stay near [district] with breakfast” — and read what the assistant says about you versus competitors. If your program never appears while a competitor’s member rate does, that’s the gap. A travel AI audit does this systematically across assistants and prompts; programs and pricing are on the site.

Loyalty was never the problem. Its invisibility is.

The direct-booking fight was always going to be decided upstream — in whatever interface assembles the shortlist. For fifteen years that interface was search and metasearch, and loyalty programs were built to win the click after the list. Now the list is assembled by assistants that can only compare what they can see, and 62% of AI-assisted bookings already settle off-property (NPS Prism/Dynata, 2026). The window where fixing this is cheap — publishing loyalty and perk content in machine-readable form before agentic booking hardens platform loyalties — is open now. Google’s August 27 launch put chain loyalty pricing inside the answer layer. Independents and small groups get the same leverage not by matching Marriott’s budget, but by making their own version of the invisible advantage visible. If you want to know what assistants currently say about your property — and what to publish to change it — start with the AI visibility program for hotels or an audit that measures it prompt by prompt. The commissions you save are the ones the shortlist was about to give away.