AI agents are starting to book travel. What it means for you

A client messages you at 9 pm: "I asked ChatGPT to find me flights to Leh and it just... did it." No enquiry to you, no back-and-forth on WhatsApp, no quotation. That is agentic AI travel booking arriving in your market. It is not a rumour. It is a product feature that shipped in October 2025.

The honest version of this story is less dramatic than the LinkedIn posts about it, and more useful. AI agents can now search and shortlist flights and hotels inside a chat window. They still cannot stitch together a Kashmir houseboat, a shikara transfer, a taxi with a valid permit and a Ladakh Inner Line Permit into one booked, paid-for, WhatsApp-confirmed itinerary. That gap (between "AI can find it" and "AI can deliver it") is where your business either gets stronger or gets skipped. Which one happens depends on choices you make in the next twelve months, not on anything OpenAI or Google decides next.

This post lays out exactly what changed, what an AI agent can and cannot do with an Indian domestic or outbound trip today, and the specific, unglamorous defence that actually works: making sure that when a client's AI assembles a shortlist, your agency is on it.

The timeline, so far

Three moments matter, and the order tells you more than any single headline.

On October 6, 2025, OpenAI launched ChatGPT apps with Expedia and Booking.com as its first travel partners. A user could type something like "Expedia, find me flights to New York Nov 27–30" directly inside a ChatGPT conversation and get real search results back, without opening a browser tab (Skift). This is the moment "AI books travel" stopped being a demo and became a shipped feature used by millions of ChatGPT users.

Then, in March 2026, OpenAI walked part of it back. It shifted ChatGPT's role from completing transactions to product discovery: search and shortlisting stay inside the chat, but checkout moves to the partner's own app or website (Skift). Read that carefully: the booking still ends on the supplier's own rails. The intermediary between "I want to go somewhere" and "I have a confirmed ticket" has not been deleted. It has been reshuffled.

As of July 2026, that is where things stand: AI agents are strong at narrowing options and weak at owning the transaction end-to-end. Google and Expedia are both reportedly building more autonomous trip-planning agents on top of this, but the pattern of "AI finds, human or supplier system executes" has held for the better part of a year now.

For an Indian operator, the practical read is this: the part of the funnel that gets automated first is the part you already treat as low-value: comparing a Delhi–Goa fare across three airlines, or checking whether a Manali hotel has a room free in the last week of December. Nobody was paying you for that lookup. What clients paid you for was never in the "search" step. It was in the "make it actually happen, and fix it if it doesn't" steps, which is exactly where the agents stop.

Why Expedia is building the thing that scares it

The clearest signal of how seriously this is being taken comes from Expedia's own regulatory filings, not from a conference keynote. Expedia lists "generative and agentic AI" as a competitive threat, warning that travellers may shift toward AI-driven platforms where Expedia itself "may lack a significant presence." At the same time, it is partnering aggressively with those same AI platforms and reporting that AI-referred traffic converts at higher rates than its other channels (GeekWire).

That is not contradictory, it is the correct response to genuine uncertainty. Expedia does not know whether AI agents end up routing bookings through it or around it, so it is hedging by showing up everywhere the agents look, while also telling shareholders this could go badly. A small agency should copy the instinct, not the scale: you cannot out-build OpenAI or Expedia, but you can make sure you are one of the places an AI agent finds when it goes looking.

What an AI agent can actually book today

Strip away the hype and the current capability is narrow. Here is a realistic read on an Indian operator's typical work, task by task.

Task	Can an AI agent do this unassisted today?
Search and compare flight fares for a fixed route/date	Yes
Search and compare standalone hotel rooms	Yes
Complete checkout inside the chat	No, moved to partner app/site (Mar 2026)
Multi-city Indian domestic itinerary with connecting legs	Partially, needs human judgment on routing
Houseboat + shikara + taxi + permit as one bookable package	No
Visa document checklist and application paperwork	No
Group/FIT negotiated net rates with a hotel	No
Handling a mid-trip disruption (missed connection, landslide)	No

The pattern across every "no" is the same: anything that requires coordinating multiple independent suppliers who don't share a booking system, anything involving paperwork with a government office, and anything requiring judgment when a plan breaks mid-trip. That is most of what a tour operator actually sells.

Example: A client asks an AI agent to plan five days in Kashmir: Srinagar houseboat, Gulmarg day trip, Pahalgam stay. The agent can competently suggest an itinerary shape and even surface hotel options with prices. What it cannot do is call the houseboat owner to confirm availability for those exact dates, arrange the taxi with a driver who knows the Sonmarg road conditions in that week, or handle the Inner Line Permit paperwork if the client wants to push on toward Ladakh. Someone still books three separate phone calls and one government form. That someone is you, if the client can find you.

Will AI replace travel agents

This is the question every operator is actually asking, so it deserves a direct answer: not in the way the question assumes.

The tasks in the table above (permit paperwork, multi-supplier orchestration, judgment calls when something goes wrong) are not on any credible near-term roadmap for AI agents. They require relationships and local knowledge that live outside any bookable API. An AI agent can suggest Gulmarg. It cannot know that your driver's cousin runs the best breakfast dhaba on the way, or that a particular houseboat owner will hold a room without a deposit for a repeat client.

The real risk is narrower and more urgent than "replacement": it is invisibility. Some industry surveys report that roughly 44% of travel agents fear being replaced by AI, while the rest describe adapting their business around it (a directional claim from industry commentary, not audited data, so treat the exact figure with caution). The more useful framing than "will I be replaced" is "will I be found." If a client's AI agent builds a shortlist of Kashmir specialists and your agency has no structured, reviewed, crawlable presence, you don't get replaced by the AI. You simply never make the list it hands the client.

The actual risk: becoming invisible, not obsolete

AI agents don't invent recommendations from nothing. They pull from the same signals search engines already rank on: structured business listings, review volume and recency, and pages that clearly demonstrate expertise in a specific niche. Some industry analyses report AI-referred website traffic jumped sharply through 2025 and project a large majority of searches will end without a click by late 2026. Treat both as reported estimates rather than settled numbers, but the direction is worth planning around regardless of the exact percentage.

What that means in practice: the client asking an AI agent "who runs good Ladakh bike trips" is being shown a shortlist assembled from exactly the kind of public signals your agency either has or doesn't have. A Google Business Profile with accurate categories and photos, a steady stream of genuine Google reviews, and web pages that state plainly what you specialise in are not "nice to have" marketing extras anymore. They are the raw material an AI agent reads before it ever mentions your name to a client.

Careful: Don't chase becoming an official "ChatGPT travel partner" or wait for an API integration invitation. That path is built for Expedia-scale companies, not a five-person agency. The defence that actually works for you is the boring one: be well-described, well-reviewed, and specific enough online that any AI system crawling public data surfaces you for the exact niche you serve.

How to appear in ChatGPT and Google's AI recommendations

None of this requires new software or a developer. It requires treating your existing online presence as the dataset an AI agent will one day read.

1. Fix your Google Business Profile categories and attributes. Vague categorisation ("travel agency") loses to specific ones ("Ladakh tour operator," "Kerala houseboat specialist"). This is covered in detail in the profile setup guide.
2. Build a real, current review base. Reviews are one of the strongest signals both Google's ranking systems and AI summarisation tools draw on. A simple WhatsApp ask after every trip, done consistently, beats an occasional review-drive campaign. See the scripted approach.
3. Publish pages that state your niche in plain language. A page titled "Kashmir houseboat packages with permit assistance" is more citable than a generic "our services" page. AI systems, like search engines, reward specificity over breadth.
4. Keep your NAP (name, address, phone) consistent everywhere: website, GBP, Instagram bio, any directory listing. Inconsistency is one of the most common reasons a legitimate business gets under-surfaced.
5. Write plain-language answers to the questions clients actually type. A page or FAQ section that answers "how many days do you need in Ladakh in July" or "is a houseboat safe for kids in Kashmir" in direct, specific language is easy for an AI summarisation system to lift and attribute. Vague marketing copy is not.
6. Don't neglect the fundamentals of AI-era search visibility more broadly. The mechanics are covered in SEO for the AI Overviews era.

None of this guarantees a mention inside a ChatGPT answer. All of it improves the odds that when a human, or an AI acting for a human, goes looking for your specialty, you're findable. And none of it requires new software. It is the same groundwork that has always made a small business rank well, applied with the knowledge that an AI agent, not just a person, may be the one reading it.

Common questions

Will AI replace travel agents in India?

Not for the work that involves coordinating multiple suppliers, handling permits and visa paperwork, or fixing a broken itinerary mid-trip. None of that is close to automated. The risk isn't replacement, it's a shortlisting AI never surfacing you because your online presence doesn't clearly signal your niche.

Can my clients already book through ChatGPT?

They can search and compare flights and hotels through partner integrations like Expedia and Booking.com inside ChatGPT. As of the March 2026 change, actual checkout happens on the partner's own app or site, not inside the chat itself (Skift).

Should I try to get my agency listed as an AI chatbot partner?

Realistically, no. Those integrations (Expedia, Booking.com) are built for platforms with API infrastructure at a scale a small agency doesn't need to match. Your leverage is discoverability: a strong Google Business Profile, genuine reviews and specific niche content, which is what any AI agent's underlying search will actually be reading.

What should I stop worrying about?

Multi-supplier Indian domestic packages, visa and permit paperwork, and judgment calls during a live trip are not going to be automated by any agent shipping today. Spend your energy on the parts of the trip an AI genuinely cannot touch, and on being visible for the parts it recommends.

The short version

• AI agents can search and shortlist flights and hotels today (since October 2025); they cannot yet complete checkout inside the chat. That moved back to supplier apps in March 2026.
• Multi-supplier orchestration (houseboat, taxi, permits), visa paperwork, and mid-trip problem-solving remain human work with no credible near-term automation.
• Expedia itself lists agentic AI as a competitive threat in its own filings while simultaneously partnering with AI platforms: a sign of genuine uncertainty, not settled disruption.
• The real risk isn't replacement, it's invisibility: an AI agent building a client's shortlist from public signals you may not have optimised.
• Fix your Google Business Profile categories, build a steady stream of real reviews, and publish specific niche pages: that is the whole defence for an agency your size.
• Treat any "44% fear replacement" or "70% zero-click by 2026" statistic you see as directional industry commentary, not audited fact.
• Revisit this every few months: the timeline from October 2025 to March 2026 shows this is still moving fast.