Can AI agents book your hotel? What I see running a bot army, and what to fix now

By Alexander van den Berg, founder of Atelier ATTN., Hanoi. Published 13 October 2026.

The short answer: Yes, and sooner than most hotels expect. I run teams of AI agents daily and see how fast they improve. On 8 October 2026, Google warned that if an AI agent cannot read your room facilities and cancellation policy, your hotel “will be invisible.” But agents read far more than Google: OTAs, reviews, Reddit, video, maps, Wikipedia and your website. AI bookings are still under 1% of Booking Holdings’ room nights. So act now, while it is cheap: make your facts correct and consistent everywhere agents look, in your guests’ languages.

Why do I take Google’s warning seriously?

Because I watch agents work every day. For some time I have run what I call bot armies: many AI agents doing real work. In my own setup, agents research, draft content and outreach, compile reports, and handle booking-style tasks: finding, comparing and preparing options for me to approve. What they can do changes month by month.

Agents do not browse like people. They look for facts they can check. If a fact is missing, they move on.

That is what Google described at the WTTC Global Summit in Malta, on a panel called “AI Unlocked: The New Era of Travel”. Christie Travers-Smith, Google’s Director of EMEA Strategic Partnerships, said agents check room facilities, cancellation policies and prices. As reported by Destination KSA:

“If it can’t read your room facilities, your cancellation policy… then it simply won’t show. And it will be invisible.”

Microsoft’s Julie Shainock added that agents need identity controls, access permissions and limits on their actions (RusTourismNews).

To be fair, the panel announced no booking system and no timeline. It described a direction. From what I see daily, agents will book rooms, and soon.

The agent explosion goes far beyond ChatGPT

A year ago, “AI” meant a chatbot. Now every big platform ships agents that act for you.

• Meta Muse. A personal agent app launched in the US on 8 September 2026 that books dinners and manages email, with approval. It hit 5 million US downloads in 22 days, faster than ChatGPT, Grok or Claude, per Sensor Tower data reported by Forbes. Meta plans to take small fees on bookings. On 12 October 2026 it launched Muse Wonder, its strongest agent model, reserved for paid plans of $20 or $100 a month (CNBC). It is not yet available everywhere (Meta).
• OpenAI. ChatGPT has 1.2 billion weekly users (The Verge). July brought ChatGPT Work, an agent that acts across apps and files (OpenAI). On 29 September came dots: always-on agents with their own cloud computer that can connect to over 4,000 apps (OpenAI).
• Grok and Grok Bot. xAI launched Grok Bot in beta on 11 August 2026. Its agents have their own computer, sign into websites and finish jobs end to end (xAI). Thousands of organisations adopted it within weeks, xAI says (xAI).
• Kimi. Moonshot AI’s Kimi Work, launched in June 2026, is a desktop agent that drives your browser and can split a task across up to 300 sub-agents (Decrypt).
• OpenClaw. An open-source agent that runs on your own device and takes orders through WhatsApp or Telegram. GitHub calls it the fastest-growing project in its history (GitHub).
• Manus. A general-purpose agent. After splitting from Meta, it went independent again in August 2026, serving millions of users (TechNode).
• Perplexity Comet. An AI browser that browses on your behalf, free worldwide since October 2025 (Business Insider).
• Gemini. Over 1 billion monthly users. On Android it acts in 40+ apps, booking rides or reserving tables (Google).

How fast is this growing?

Exponentially. Three markers:

• OpenClaw began as a weekend project in November 2025. By 26 August 2026 it had about 388,000 GitHub stars (GitHub). On 13 October the repository showed about 391,800. Stars are not users, but the curve is steep.
• 56% of US leisure travellers used AI for a trip in the past year, up from 33% in the first half of 2025 (Phocuswright).
• Traffic from AI sources to US travel sites grew 194% year on year in May 2026 (Search Engine Land). By July they converted only 1% below non-AI traffic (TechInformed).

Is AI hotel booking actually happening yet?

Yes, in early form.

• Google AI Mode started hotel booking on 27 August 2026, in the US and in English (Google). Partners include Booking.com, Expedia, Trip.com, Marriott and IHG. Right before payment, it shows the room type and cancellation policy: the two facts Google named in Malta.
• ChatGPT runs Booking.com and Expedia apps with live prices; checkout finishes on their sites (PhocusWire, Skift). Radisson’s ChatGPT app converts about 1.5 times better than its organic search (OpenAI).
• Perplexity has offered hotel booking since March 2025 (Skift).

Still, less than 1% of Booking Holdings’ room nights come from AI channels (Skift). In the same report, Stiplo analysed 27,360 AI answers about Barcelona hotels. Independents were 8 times more likely than chains to be unknown to AI. Once known, they were recommended at about the same rate.

The gap is not favouritism. It is missing information.

Where do AI agents get their facts about your hotel?

Not only from Google. An agent pulls from many sources and cross-checks them:

• OTAs. In a Nokumo study of 3,600 AI hotel answers in five European markets, Booking.com appeared in 95.3% of queries. OTAs took 22.9% of citations; independent hotel sites 11.8% (Nokumo).
• Review sites. TripAdvisor was among the three most-cited domains in that study. In the Stiplo research, Google review volume was the factor most linked to being discovered.
• Reddit, forums and blogs. A Peec AI analysis of 30 million sources found Reddit was the most-cited domain across ChatGPT, Gemini, Perplexity and Google’s AI features (Search Engine Land). Travel blogs and editorial mentions tell AI what makes you different.
• Social and video. YouTube ranked second in that analysis. Chinese travellers plan on Xiaohongshu, with over 400 million monthly users, 130 million of them outbound travellers (Xiaohongshu), and Douyin. TikTok plays a similar role elsewhere.
• Korean platforms. Naver’s AI Tab links answers to Naver Map and real-time reservations (Naver). Kakao has put ChatGPT inside KakaoTalk chats (Kakao).
• Maps. Google Maps and other map apps hold your location and reviews, and answer “near” questions.
• Wikipedia and Wikidata. Wikipedia is among the five most-cited domains (Peec AI). Wikidata holds structured facts about places.
• Your own website. The source of truth. In the Nokumo audit, only 7% of hotel websites used Hotel schema, and only 5.7% were cited by any AI model. The cited sites were the clearest, not the flashiest.

When these sources disagree, the agent cannot verify you. It picks someone else.

Why should hotels in Vietnam act now?

Because your guests already plan with AI, and each market uses different tools.

Vietnam welcomed a record 21.2 million international visitors in 2025 (Government of Vietnam). From January to September 2026 it welcomed 17.7 million, up 14.5%; European arrivals rose 55% (VOV).

What this means for you:

• One language is not enough. A Chinese guest may ask Doubao and check Xiaohongshu. A Korean guest may ask Naver. Facts only in English give those tools little to work with.
• Booking inside AI is US-only for now, and the first partners are chains and OTAs. Use the time. If an agent cannot read your facts, it sends the guest to an OTA, or to a competitor.
• It is cheap now. Most fixes take hours, and pay off today in search, OTAs and maps.

How can you check whether AI recommends your hotel?

Test it yourself in about 30 minutes in ChatGPT, Gemini, Perplexity and Google AI Mode; add Doubao and Naver for Chinese or Korean guests. Use a logged-out window and run each prompt twice, because answers vary.

Prompts to try in English

1. “What room types does [Hotel] in [city] have, and how many beds are in each?”
2. “What is the cancellation policy at [Hotel], [city]?”
3. “Does [Hotel] have a room for 2 adults and 2 children? Is breakfast included?”
4. “Best hotels in [district] under [price] per night with free cancellation.”
5. “Hotels near [landmark] with a pool and an airport transfer.”
6. “Can I book [Hotel] directly for [dates]? Give me the link.”

Prompts to try in Vietnamese

1. “Khách sạn [Tên] ở [thành phố] có những loại phòng nào, mỗi phòng có mấy giường?”
2. “Chính sách hủy phòng của khách sạn [Tên] là gì?”
3. “Khách sạn [Tên] có phòng cho 2 người lớn và 2 trẻ em không? Có bao gồm bữa sáng không?”
4. “Gợi ý khách sạn tốt ở [quận] giá dưới [giá] một đêm, được hủy miễn phí.”
5. “Khách sạn gần [địa danh] có hồ bơi và xe đưa đón sân bay.”
6. “Tôi có thể đặt phòng trực tiếp tại khách sạn [Tên] từ [ngày] đến [ngày] không? Cho tôi đường link.”

How do you read the answers?

Mark every answer: correct, wrong, missing or sent to an OTA. Write down which competitors the AI names instead of you, and which sources it cites.

A wrong answer is worse than none: a guest who books on a wrong cancellation policy becomes a complaint. Missing from prompts 4 and 5 means you are not on the shortlist at all.

Rather have us run it? Book the free “Can AI agents book you?” check at atelierattn.com/ai-check. We test your hotel in four AI tools, in English and Vietnamese, and send you a score out of 100 with your top 5 fixes.

What does AI need to recommend and book your hotel?

Facts it can read, verify and trust, everywhere it looks.

The AI-readiness checklist for hotels

• Your website as the source of truth. Schema.org Hotel or LodgingBusiness markup, including rooms and offers. Room facilities in text, not only photos. The cancellation policy in crawlable HTML, not a PDF. Rates that load without a JavaScript-only widget, plus Google’s free booking links.
• Complete OTA listings. Fill every field on Booking.com, Agoda, Trip.com, Expedia and Traveloka.
• Reviews on every platform. Google, TripAdvisor and Booking.com: volume, recency and replies, in your guests’ languages.
• Maps. A complete Google Business Profile, and correct listings in the other map apps your guests use.
• Reddit, forums and blogs. Know what is said about you. Answer openly as the hotel. Never fake reviews.
• Social and video. Room tours on YouTube and TikTok, plus Xiaohongshu or Douyin for Chinese guests. Put the facts in speech and captions.
• Wikipedia and Wikidata. If your hotel has an entry, check the facts. Follow their rules; no promotional copy.
• Consistency everywhere. Name, address, phone, room names, policies and amenities must match across all of the above.
• AI crawler and agent access. Make sure your robots.txt, firewall or CDN does not block AI search crawlers such as OAI-SearchBot and PerplexityBot by accident. Decide on training crawlers such as GPTBot deliberately.
• Fast multilingual answers. Pages in English and Vietnamese, plus Chinese or Korean for those markets, linked with hreflang. A real FAQ. Quick replies on email, WhatsApp, Zalo, KakaoTalk, WeChat and OTA inboxes.

How does Atelier ATTN. check whether AI agents can book you?

We built a free check called “Can AI agents book you?”

What you get:

• 10 real guest questions in ChatGPT, Gemini, Perplexity and Google AI Mode, in English and Vietnamese.
• What AI gets right, wrong or cannot find about rooms, prices and cancellation, and which competitors it recommends instead.
• A website check: structured data, crawler access, how readable your rates and policies are.
• A consistency check across Google, Booking.com, Agoda and your site.
• A score out of 100, with a clear traffic light per item.
• Your top 5 fixes, ranked by impact, with the effort in hours.
• A 2-page report in English or Vietnamese, plus a 15-minute walkthrough. Results within 5 working days.

We log the date, the tool and every prompt. “Not found” is a valid result. We never invent figures.

How do we fix it, and keep it fixed?

Hotel Data & Visibility

Our core service makes your hotel’s information correct, consistent and machine-readable everywhere guests and AI look.

• Fix: schema for your hotel, rooms and offers; crawlable policies and rates; AI crawler access; matching data on your website, Google, the OTAs, review sites and maps.
• Measure, every month: the same guest questions in all four AI tools, in both languages. We report what AI gets right, wrong or misses, and how often you appear against three competitors. AI answers vary, so we run each question several times and report ranges.
• Price: setup 4.9M VND (EUR 295), then 7.9M VND (EUR 449) per month.

We guarantee the work and the measurement, not rankings or bookings.

A guest AI that sounds like your team

Next comes answering guests faster. Our guest AI replies at any hour in English and Vietnamese. It learns from your hotel’s own history, past conversations, reviews and booking data, so it sounds like your team, not a generic chatbot. It works with your current booking system and hands over to staff when a person should take over.

Frequently asked questions

Will AI agents really book hotel rooms?

They already do, in early form. Google AI Mode books hotels for US users, and Perplexity has since March 2025. Agents such as Meta Muse, OpenAI’s dots and Grok Bot act on the web for their users. Volumes are still under 1% of Booking Holdings’ room nights.

Is Google still the main source?

It is one of many. Agents also read OTAs, reviews, Reddit, YouTube, maps, Wikipedia and your website. Chinese travellers use Doubao, DeepSeek and Xiaohongshu; Koreans use Naver.

Does this matter for hotels in Vietnam?

Yes. Your two largest markets plan with their own tools: Doubao and Xiaohongshu for China, Naver and KakaoTalk for Korea. Most other guests use global tools such as ChatGPT and Gemini. Domestic guests have Google AI Mode in Vietnamese.

Is this the same as SEO?

It overlaps. Good SEO helps. But agents also need verifiable facts, such as rooms, rates and cancellation terms, that match across your website, OTAs, reviews and maps.

Do beautiful photos help AI recommend my hotel?

Photos help guests. AI needs text and structured data. A room shown only in a photo is a room an AI agent cannot describe.

Should I block AI crawlers?

Decide deliberately. Blocking AI search crawlers can keep your hotel out of AI answers. Training crawlers are a separate choice.

How long does it take to fix?

Most top fixes take hours to days. Keeping data consistent is ongoing, because rooms, rates and policies change. That is why we measure monthly.

Is the “Can AI agents book you?” check really free?

Yes. No obligation. You get the score, the top 5 fixes and a 15-minute walkthrough.

Fix it now, while it is cheap

AI booking is small today. From what I see every day, it will not stay small. The hotels AI can read, verify and trust will be the ones it recommends, and later the ones it books.

Book your free “Can AI agents book you?” check at atelierattn.com/ai-check. Or email me directly at [email protected].