← live missions
analysis 2026-11-22

Best AI travel agent? What the measured record shows before you choose (2026)

Every travel platform now says it has an AI agent. On the public record, the usage figures are the companies' own, agent-booked volume is close to zero, and the one ruling from an outside adjudicator says the company owns every wrong answer its bot gives.

You are choosing an AI travel agent, and every vendor and platform in the category has a usage figure ready. Daily conversations, languages supported, a multiple on AI search traffic. None of them tells you how often the agent is wrong.

In travel that is the number that costs money. A wrong answer about a fare rule or a refund window is not a bad recommendation. It is a promise the company may have to honour.

The one outside ruling on the record says exactly that. A BC tribunal ordered Air Canada to pay C$812.02 after its website chatbot gave a passenger wrong bereavement-fare advice [source].

What an AI travel agent is

An AI travel agent is software that takes a traveller’s request in conversation and does part of the travel company’s work: it answers policy questions, searches and compares options, handles changes after booking, and in the fullest version completes the booking and payment.

The useful line is how far down that chain a product actually goes. Answering questions and handling post-booking support is the deployed reality. Booking whole trips on the traveller’s behalf is mostly still the pitch.

What is the best AI travel agent?

The best AI travel agent is the one whose accuracy you can audit, and on the public record nobody outside the companies has audited any of them. TIN cannot rank products it has not checked, so this post answers “best” with the test a buyer should apply.

That is not a dodge. Search the category and you will find ranked lists of AI travel agent apps. Look for the column that says who measured the error rate, and you will not find it.

What the case files show

TIN has two case files on AI agents in travel, and only one of them rests on a record from someone other than the company. This post says so plainly: the argument has one strong anchor and one company self-report.

Air Canada, the liability anchor. The tribunal found Air Canada liable for negligent misrepresentation and rejected the argument that the chatbot was “a separate legal entity that is responsible for its own actions” [source]. It held that “it makes no difference whether the information comes from a static page or a chatbot” ([source]).

The passenger paid C$1,630.36 for flights on the chatbot’s advice and was awarded C$650.88 in damages [source]. The Air Canada case file sets out the full decision.

MakeMyTrip, the scale claim. MakeMyTrip told the SEC that its Myra assistant “engaged in over 85,000 daily conversations in Q1 FY27” and resolved “over 50% of post-booking queries in relation to flights and hotels” [source]. Those are regulated disclosures, but they are the company counting its own assistant.

The MakeMyTrip case file caps below green for that reason: there is no measurement of Myra’s resolution rate from anyone other than MakeMyTrip.

What the platforms themselves report

The Q3 2026 earnings season added three more travel and mobility figures on AI agents. Every one is company-reported, and they are useful as disclosure, not as proof.

CompanyFigureWho counted
AirbnbAbout 40% of support issues resolved with no humanAirbnb 8-K
MakeMyTripOver 50% of post-booking queries resolvedMakeMyTrip 6-K
Booking.comAI search visitors up 10x, LLM referrals under 1% of room nightsBooking CFO
UberFour rides booked by outside AI agents in the quarterUber CFO
Air CanadaLiable for the bot’s wrong answer, C$812.02BC tribunal

Airbnb said “approximately 40% of all customer support issues” were resolved without a human agent in Q3, up from a third in Q2, and that support cost per booking fell more than 17% year over year [source]. That is the strongest efficiency claim in travel this year, and it is Airbnb’s own count.

Booking’s CFO said daily AI search visitors were “up 10 times” quarter on quarter, while LLM referrals “still account for significantly less than 1% of Booking.com’s Room Night” [source]. Uber’s CFO said outside agents fulfilled “a grand total of four trips” on rides ([source]).

Will AI replace travel agents?

AI will not replace travel agents on the current record, because the record shows almost no AI agent selling travel. Uber put the number at four trips, and Booking put LLM-referred room nights well under 1% [source].

The obvious reading of the table is the opposite: AI is taking over travel, look at the tenfold jump in AI search visitors. That reading is wrong. A multiple on a tiny channel says nothing about who sells the trip, and Booking disclosed both numbers in the same call [source].

Where the reported gains do sit is support after the sale. Airbnb’s 40% and MakeMyTrip’s 50% are both about resolving issues, not about choosing destinations [source]. That is the work a human agent is least paid for, and the work where a wrong answer becomes a liability.

How to evaluate an AI travel agent

You evaluate an AI travel agent on three things, and the vendor should be able to show each one from data you can see.

  1. 01
    Measured resolution, with a definition. “Resolved” must mean the issue stayed closed, not that the traveller stopped typing. MakeMyTrip’s own filing scopes its figure to post-booking flight and hotel queries [source].
  2. 02
    Error rate on policy answers. Fares, refunds and change fees are where the Air Canada ruling bites. Ask how often the agent contradicts the published policy.
  3. 03
    Who owns a wrong answer. The tribunal said the company does [source]. Your contract with the vendor should say who pays when that happens.

A vendor that offers only conversation volume is offering the number that measures demand for the bot, not its quality.

The bottom line

The best AI travel agent today is not the one with the most conversations. It is the one whose error rate you can check, because the only outside adjudicator to look at a travel bot made the airline pay for its mistake. The platforms reporting real gains report them in support, counted by themselves, while agent-booked travel is a rounding error.

The transferable rule: when software speaks for you, its accuracy is your liability, so buy the measurement before you buy the agent.

Sources

  1. 01
    CanLII (BC Civil Resolution Tribunal), “Moffatt v. Air Canada, 2024 BCCRT 149”, 14 February 2024. https://www.canlii.org/en/bc/bccrt/doc/2024/2024bccrt149/2024bccrt149.html
  2. 02
    CBC News, “Air Canada found liable for chatbot’s bad advice on bereavement rates”, 15 February 2024. https://www.cbc.ca/news/canada/british-columbia/air-canada-chatbot-lawsuit-1.7116416
  3. 03
    U.S. Securities and Exchange Commission, MakeMyTrip Limited Form 6-K exhibit 99.1, “MakeMyTrip Limited Reports Q1 FY27 Financial and Operating Results”, 3 August 2026. https://www.sec.gov/Archives/edgar/data/0001495153/000119312526329581/mmyt-ex99_1.htm
  4. 04
    U.S. Securities and Exchange Commission, Airbnb, Inc. Form 8-K, Exhibit 99.1 (Q3 2026 Shareholder Letter), 5 November 2026. https://www.sec.gov/Archives/edgar/data/0001559720/000119312526449727/d437988dex991.htm
  5. 05
    Earnings Whispers, “BKNG Q3 2026 Earnings Call Transcript”, 3 November 2026. https://www.earningswhispers.com/transcript/BKNG/Q32026
  6. 06
    Earnings Whispers, “UBER Q3 2026 Earnings Call Transcript”, 4 November 2026. https://www.earningswhispers.com/transcript/UBER/Q32026

Questions

What is the best AI travel agent?

The best AI travel agent is the one whose resolution rate and error handling you can check, not the one with the biggest usage figure. On the public record no outside party has measured any travel AI agent's accuracy, so no product can honestly be ranked first today.

Will AI replace travel agents?

AI will not replace travel agents on the current record. Uber said outside AI agents booked four rides in a quarter, Booking.com said LLM referrals are well under 1% of room nights, and the gains companies do report sit in post-booking support, not in selling trips.

Is a travel company liable for what its AI agent says?

A travel company is liable for what its AI agent says, on the one ruling in the record. In Moffatt v. Air Canada a BC tribunal held the airline responsible for its chatbot's wrong bereavement-fare advice and ordered it to pay C$812.02.

This is analysis, not a verified outcome. It carries no verification badge and never will. The proof lives in the case files, where every figure is checked against the public record and the method is printed on the page.