AI Mode, the conversational surface Google bolted onto Search last year, can now track flight prices, help you find and book a hotel, and quote both in airline miles or hotel points. Google announced the three additions on 27 August 2026, and together they move AI Mode from a place where you research a trip to a place where you arrange one.
That is a smaller sentence than it sounds, and a bigger one. Smaller, because none of these features invents anything Google did not already have — the flight data has been in Google Flights for a decade and the hotel inventory has been in Search for almost as long. Bigger, because the point of entry has changed. You no longer start at a comparison grid. You start at a sentence.
Google has been assembling this inside AI Mode for nine months. Canvas, the side panel that builds an itinerary, arrived in November 2025 as a desktop Labs experiment. Hotel price tracking for individual properties followed in April 2026. What landed this week is the piece that was always missing: a way to finish. Booking now happens inside the conversation, with the transaction handed to a partner and paid through Google Pay.
This article covers exactly what shipped into AI Mode, where each feature actually works today, which ten travel companies signed up, and what the change does to the travel brands whose websites used to be the destination. If you run a hotel, an agency, a tour operator or any business whose customers find you through search, the last two sections are the ones that matter. The short version: the research traffic was already leaving, and this is the step that decides whether the booking follows it.
Table of contents
- What Google Actually Shipped to AI Mode This Week
- How Trip Planning in AI Mode Actually Works
- The Rollout Is Far More Uneven Than the Headline
- The Ten Partners, and What Signing Up Buys Them
- AI Mode Is Being Used Like a Planner Already
- What This Does to Travel Websites
- What Travel Brands Should Actually Do About AI Mode
- The Risks Nobody Puts on the Slide
- What to Watch Over the Next Six Months
- Frequently Asked Questions
- References
What Google Actually Shipped to AI Mode This Week
Three features went live, and they are not equally available. Reading the rollout carefully matters more than reading the announcement.
Flight price tracking inside the conversation
You describe when and where you want to fly, AI Mode returns the best options it can find, and then — this is the new part — you can say “track these flight prices” in the same thread. Google emails you when the price moves for that destination and those dates. No form, no saved search, no separate Flights tab. In AI Mode the tracking request is just another sentence in the conversation you were already having.
Hotel discovery and booking without leaving Search
The second feature is the one with commercial weight. You can now discover and book a hotel through a conversation with AI Mode, with reviews and the key decision factors surfaced for each listing. When you pick one, a “Continue on Google” option carries you to the partner’s booking flow, and the payment is processed through Google Pay. The conversation does not end at a link to somebody else’s website.
Prices quoted in miles and points
The third is the smallest feature and the clearest signal of intent. AI Mode can now show what a flight or a hotel costs in airline miles or hotel points rather than currency. Google’s own example prompt is unusually specific: “I want to travel from Atlanta to Miami using my AA miles. Help me find some options for nonstop flights departing Oct 9 and returning on Oct 12.” That is not a search query. That is a brief you would hand to a travel agent.
Where the underlying data comes from
Nothing in AI Mode is invented on the spot. The flight side draws on more than 300 airlines and travel sites, the same real-time feed that powers Google Flights. The hotel side pulls rates, room photos, amenities and reviews from partner inventory plus Google Maps. None of it is generated. The model is choosing and explaining, not inventing prices — a distinction worth holding onto when the failure modes come up later.
| Feature | What it does | Where it works today |
|---|---|---|
| Flight price tracking | Emails you when fares move for your dates and route | More than 180 countries |
| Hotel discovery and booking | Compare, choose, then finish via Continue on Google | United States, English only |
| Miles and points pricing | Shows cost in airline miles or hotel loyalty points | Global |
| Canvas itinerary panel | Side panel that assembles and saves a full trip plan | United States, all users |
| Flight Deals | Finds cheap destinations from a loose description | More than 200 countries and territories |
| Individual hotel price tracking | Watch one named property rather than a whole city | Global, English and Spanish, signed in |
How Trip Planning in AI Mode Actually Works
The mechanics matter because they explain why some of this is genuinely useful and some of it is a demo waiting for the rest of the product to catch up.
Canvas is the workspace, not the answer
Canvas is a tool inside AI Mode that gives you a space to organise plans over time. You describe the trip you want, select “Create Canvas”, and a side panel assembles the plan: real-time flight and hotel data from Search, photos and reviews from Google Maps, and relevant material from sites across the web. Progress saves automatically, so a trip you started on Tuesday is still there on Thursday.
The AI Mode itinerary is assembled, not written
What Canvas produces inside AI Mode is closer to a working document than an essay. Hotel options are comparable on price and amenities. Restaurants and activities are ordered by travel time between them, which is the single most useful thing an itinerary tool can do and the thing most human plans get wrong. Attractions land on a map rather than in a list.
Flight Deals handles the traveller with no destination
Flight Deals is the front of the funnel and it works differently. You describe how you want to travel rather than where — a week somewhere warm in February, under a certain budget, direct if possible — and natural language processing turns that description into a shortlist of destinations, which is then priced against live Google Flights data. It now runs in more than 200 countries and territories and supports more than 60 languages.
What “Continue on Google” actually means
This is the part people misread. AI Mode is not becoming a merchant of record for hotels. The conversation narrows the choice, then hands the booking to the partner you selected, with Google Pay handling the payment step. Google keeps the discovery, the comparison and the moment of decision. The partner keeps the transaction, the guest data and the relationship afterwards — for now.
Agentic booking arrived first for restaurants
The booking machinery was tested somewhere lower-stakes. Agentic reservations for restaurants rolled out across the United States, then to the United Kingdom, India, Canada and Australia, working through OpenTable, Resy, Tock, Booksy, Fresha and Vagaro. Event tickets and local appointments followed through Ticketmaster, StubHub, SeatGeek and Vivid Seats. Hotels are the same pattern applied to a purchase worth a hundred times more.
The Rollout Is Far More Uneven Than the Headline
Every AI Mode feature in this announcement has a different footprint, and the gap between the widest and the narrowest is the story.
Global, nearly global, and one country
Miles and points pricing is global. Flight price tracking reaches more than 180 countries. Flight Deals reaches more than 200 countries and territories. Hotel booking — the feature with actual revenue attached — is live in one country, in one language, with the rest promised “over the coming weeks”.
Why the narrowest feature is the important one
Reach is not the same as weight. Price tracking in 180 countries generates email. Booking in one country generates commission, payment data and a habit. Google has scaled the cheap features fast and the expensive one slowly, which is exactly what a company does when it is confident in the demand and cautious about the plumbing.
The nine-month AI Mode build, in order
The sequencing tells you this was planned rather than reactive. Canvas in November 2025 as a desktop experiment behind a Labs opt-in. Canvas to every AI Mode user in the United States in April 2026, alongside individual hotel price tracking. Booking in August 2026. Each release added one missing verb: plan, watch, buy.
| Date | What Google added | Access at launch |
|---|---|---|
| November 2025 | Canvas travel planning; agentic restaurant, ticket and appointment booking | US desktop, Labs opt-in |
| November 2025 | Flight Deals expansion to 200+ countries, 60+ languages | Broad |
| April 2026 | Canvas travel planning to all US users; per-property hotel price tracking | US, then global tracking |
| April 2026 | Restaurant reservations extended to UK, India, Canada, Australia | 5 countries |
| August 2026 | Flight price tracking, hotel booking, miles and points pricing | Mixed — see the table above |
The Ten Partners, and What Signing Up Buys Them
Google named ten travel companies. The list is worth reading closely because of who is on it and what they gave up to be there.
Two categories, one strategy
| Type | Named partners |
|---|---|
| Booking platforms | Booking.com, Expedia, Hotels.com, Priceline, Trip.com |
| Hotel groups | Choice Hotels International, Hilton, IHG Hotels & Resorts, Marriott International, Wyndham Hotels & Resorts |
The hotel groups had the most to gain
Five of the ten are hotel companies with their own direct-booking apps and loyalty programmes, and those are precisely the businesses that have spent fifteen years trying to claw guests back from intermediaries. Appearing natively in AI Mode gives them a route to the traveller that does not pass through a booking platform first. The miles-and-points feature is aimed straight at them: a loyalty member who can see the points price without leaving the conversation is a member who never opens a comparison site.
The booking platforms had the most to lose
The five platforms are in a harder position. They joined because the alternative — being invisible at the exact moment a traveller decides — is worse than being one option among several. It is the same calculation every publisher made about search a decade ago, and it produced the same result: participate, and hope the traffic that arrives is worth the position you surrendered.
Nobody is being disintermediated yet
Worth saying plainly, because the coverage keeps implying otherwise. AI Mode is not selling hotel rooms. It is not holding inventory, not setting rates and not taking the booking on its own balance sheet. What it has taken is the top of the funnel — and in travel distribution, the top of the funnel is where the margin has always been decided.
AI Mode Is Being Used Like a Planner Already
Google’s own AI Mode usage data explains why travel was the obvious place to push. The behaviour was already there.
The scale of AI Mode is not in question
AI Mode passed one billion monthly users within a year of launch, and Google says queries have more than doubled every quarter since. For comparison, AI Overviews — the summaries that appear above ordinary results — passed 2.5 billion monthly users by May 2026. Trip planning is now landing on one of the largest artificial intelligence deployments in consumer software, not on a side experiment with an audience of enthusiasts.
Planning queries are the fastest-growing category
The number that matters most for travel: planning-related queries in AI Mode grew 80% faster than overall AI Mode queries over six months. Brainstorming queries grew 30% faster since launch. Decision questions beginning with “which” rose 40%. Google groups the behaviour into five buckets — Explore, Decide, Learn, Create and Do — and trip planning touches every one of them.
People type differently here
The average AI Mode search is three times the length of a traditional Search query, and follow-up queries rose more than 40% month on month in the United States. More than one in six searches use voice, images or video rather than typed text. Long, conversational, iterative, multimodal: that is a description of how somebody plans a holiday, not how somebody looks up a fact.
The model underneath changed too
Google made Gemini 3.5 Flash the default model in AI Mode globally, which is the unglamorous reason a side panel can now assemble a week-long itinerary in the time it used to take to render a summary. Latency, not intelligence, was the constraint on this kind of feature.
What This Does to Travel Websites
Here is where the announcement stops being a product story and starts being a distribution problem for everyone else.
The click was disappearing before AI Mode
By early 2026, roughly 68% of United States Google searches ended without a click through to any website, up from about 60% in 2024. That trend predates this week’s features and was driven by AI Overviews answering on the results page. Trip planning in AI Mode extends the pattern into a category where the research phase used to involve visiting a dozen sites.
One publisher’s number, as a warning
Tripadvisor has been reported to have suffered roughly a 33% decline in search-driven visits as AI search products answer travel questions inside the interface. Treat the exact figure carefully — it is an external estimate, not a company disclosure — but the direction is not really disputed by anyone in the sector.
And yet the booking share is still tiny
The counter-evidence deserves equal billing. Lighthouse, measuring across a network of around 80,000 hotels, put AI referral traffic at just under 1% of total hotel website visits in mid-2026. Booking Holdings has separately said bookings arriving through AI chatbots remain below 1% of its volume. Visits to AI platforms grew 76% year on year in the second half of 2025 while onward referrals stayed roughly flat.
The gap between those two facts is the whole risk
Research is moving. Booking has not moved yet. That gap is where travel brands are currently losing measurable traffic without losing measurable revenue, which is the most dangerous shape a problem can take — it looks survivable right up until the booking step closes too. Hotel booking in AI Mode is Google testing exactly that step.
What Travel Brands Should Actually Do About AI Mode
None of the standard advice changes much. What changes is which parts of it stopped being optional.
Make the facts machine-readable before anything else
AI Mode rewards structure long before it rewards prose, so this is where the work starts.
AI Mode assembles answers from structured data, Maps entries, partner feeds and page content. A hotel whose rates, amenities, cancellation terms and accessibility details exist only as prose inside a JavaScript-rendered page is a hotel the system has to guess about. Schema markup, a complete and current Google Business Profile and clean, crawlable pages are the entry fee. This is unglamorous technical SEO work and it is now the difference between being quoted and being skipped.
Optimise for being the answer, not the result
Ranking and being cited are no longer the same job. Answer engine optimisation targets the second one: writing content that resolves a specific question completely, in a form a model can lift and attribute. For a travel business that means real answers to the questions people actually ask a planner — what a room costs in points, whether the pool is heated in March, how far the airport transfer really is.
Treat generative surfaces as their own channel
The broader discipline of generative engine optimisation extends the same logic across every assistant that now sits between you and a customer. AI Mode is the largest of them, but it is not the only one, and the work of making a brand legible to generative AI systems is portable across all of them. Answer engine optimisation and its generative equivalents are far cheaper to start now than to retrofit in a year.
Give people a reason to leave the conversation
AI Mode is happy to hand off a booking. It is not going to explain why a guest should book direct. Rate parity, loyalty benefits, room selection, late checkout, a package that no intermediary lists — those have to exist and be visible in the material the assistant reads. A property with nothing distinctive is a property that will be chosen on price, by a machine, every time.
Fix your measurement now, because Google will not
The hardest part of this shift is that it is very difficult to see. Google does not currently provide analytics or attribution for what happens inside AI Mode, so a booking that began there and finished on a partner platform is invisible to you. Start tagging what you can, watch branded search volume and direct traffic as proxies, and stop treating flat organic sessions as a neutral result.
| Action | Why it matters now | Effort |
|---|---|---|
| Structured data on rates, rooms and policies | Determines whether you can be quoted at all | Low, one-off |
| Complete Google Business Profile with current photos | Feeds the Maps layer the planner reads from | Low, ongoing |
| Question-shaped content with direct answers | Makes your pages liftable and citable | Medium, ongoing |
| Loyalty and points availability published clearly | Directly feeds the new miles and points feature | Medium |
| Direct-booking incentives an assistant can see | The only defence against price-only selection | Medium |
| Proxy measurement for assistant-driven demand | No native attribution exists yet | High |
The Risks Nobody Puts on the Slide
The optimistic reading of trip planning in AI Mode is easy to write. The failure modes deserve the same space.
A wrong itinerary costs more than a wrong summary
If a model gets a recipe slightly wrong you eat a bad dinner. If it gets a connection time, a visa requirement or a check-in cut-off wrong, somebody misses a flight. Travel is one of the few consumer categories where a confidently phrased mistake has an immediate, expensive, non-refundable consequence, and the interface gives no signal about which parts of a plan are verified data and which are inference.
Freshness is a genuine engineering problem
Fares and room rates change constantly. A conversation that stretches across three days — which Canvas explicitly encourages, since it saves your progress — is a conversation where the numbers you were shown on Monday may not exist on Thursday. Price tracking mitigates this for the items you asked about. It does nothing for the twelve options you looked at and left in the panel.
Attribution and commercial terms are unresolved
Nobody has said publicly how a booking that originates in AI Mode is attributed, what it costs a partner, or whether visibility in the panel will eventually be purchasable. Ads already appear in a meaningful share of AI results. Assuming travel stays a purely organic surface would be an unusual bet to make about Google.
Regulators are already interested
A search engine that ranks travel options, then routes the booking through its own payment product, is a shape that competition authorities have examined before in this exact vertical. The hotel groups joining the panel are the same firms that spent years complaining about search self-preferencing. That tension has not gone away because the interface got conversational.
Concentration risk is the quiet one
If the planning layer for global travel consolidates into a single assistant, every hotel, airline and operator inherits a dependency on one company’s ranking logic and one company’s product roadmap. Businesses that spent a decade reducing exposure to booking platforms should notice they are being offered the same arrangement with a different owner.
What to Watch Over the Next Six Months
Three things will tell you whether this becomes the default way people plan travel or stays a well-built feature people use occasionally.
Whether booking finishes inside the conversation
Right now the handoff is real: you leave for the partner’s flow. If Google closes that loop and completes the transaction in the panel, the relationship changes from referral to distribution, and every commercial term in the sector gets renegotiated.
Whether advertising arrives in the travel panel
Watch for sponsored placements in itinerary results. The moment position in a Canvas plan can be bought, the optimisation work described above becomes a bidding problem as well as a content problem.
Whether the booking platforms build their own agents
Expedia and Booking.com both have assistants of their own, and both are building AI agents intended to keep a traveller inside their own app rather than inside somebody else’s. Whether they can hold direct demand — or end up as inventory suppliers to another company’s interface — is the defining strategic question in travel distribution for the rest of the decade. AI Mode did not create that question, but this week made it urgent.
Whether anyone outside the United States gets the booking flow
The most concrete thing to watch is also the simplest. Hotel booking in AI Mode is live in one country. If it reaches Europe and Asia before the end of the year, the timeline in this article compresses by a lot.
Frequently Asked Questions
Is trip planning in AI Mode available outside the United States?
Partly. Flight price tracking works in more than 180 countries and the miles-and-points display is global. Hotel booking and the Canvas itinerary panel are United States only at launch, with wider availability promised over the coming weeks.
Does Google take the booking itself?
No. AI Mode surfaces options and narrows the choice, then hands you to the partner you picked through a “Continue on Google” step, with payment processed through Google Pay. The partner remains the merchant.
Which travel companies are partners?
Ten were named: Booking.com, Choice Hotels International, Expedia, Hilton, Hotels.com, IHG Hotels & Resorts, Marriott International, Priceline, Trip.com and Wyndham Hotels & Resorts.
Are the prices in AI Mode real or generated?
Real. Flight data comes from more than 300 airlines and travel sites through the same feed as Google Flights, and hotel rates come from partner inventory. The model selects and explains rather than inventing figures — though rates can still go stale between the moment you look and the moment you book.
Will this reduce traffic to my travel website?
For research queries, almost certainly, and that was already happening before this release. For bookings, not yet — measured assistant referrals are still around 1% of hotel site visits. The sensible planning assumption is that the research decline continues and the booking share starts climbing from a very low base.
What should a small hotel or tour operator do first?
Structured data and a complete, current Google Business Profile, in that order. Both are cheap, both feed the layer AI Mode reads from, and neither requires you to predict how the product develops.
Does any of this apply to AI agents beyond Google?
Yes. The same structured, answerable, machine-legible content is what every assistant reads. Work done for AI Mode is not wasted if a traveller asks a different tool instead.
References
Google’s AI Mode can now track flight prices, help book hotels, and more — TechCrunch
Explore new ways to plan and book travel with AI in Search — Google
7 ways to travel smarter this summer, with help from Google
How AI Mode is changing and expanding the way people search — Google
Flight Deals is our new, AI-powered flight search tool — Google
Google Search’s I/O 2026 updates: AI agents and more
Google Reveals First AI Mode Usage Numbers After One Year — Search Engine Journal
Google Extends AI Travel Planning And Agentic Booking In Search — Search Engine Journal
Google pushes deeper into the travel funnel with new AI capabilities — eMarketer
Google’s New “AI Mode” for Search: The Impact on Trip Planning — Skift
Hotel Bookings via AI Chatbots Stay Below 1%, Booking Holdings Confirms
Tripadvisor’s Collapse Is the AI Warning the Travel Industry Can’t Ignore — eTurboNews
5 OTA Trends Shaping Hotel Distribution in 2026 — Cloudbeds