Travellers will hand AI the chores. They're keeping the choices.

A new Mews survey of 3,250 travellers across eight countries puts numbers on something the AI travel pitch keeps skipping. People will happily hand AI the chores. Most are not handing it the choices.

Ask what they'd actually use and the answers are functional. 63% would let AI monitor room-rate drops. 56% would use it for real-time translation. But itinerary planning drops to 44%, personalising the room gets 41%, and 57% say they'd rather stay in control of their trip than hand decisions to an AI acting on their behalf.

Mews frames it as chore versus choice, and that framing should pin to the wall of everyone building in this space. A price alert doesn't need to know who you are. It needs to be fast and right. Choosing the trip is personal by definition, and most travellers are saying they want to keep that part.

The data underneath is just as telling. People are far more comfortable with AI drawing on what they've knowingly shared - stated preferences at 61%, past stay history at 56% - than on what's been collected about them, with online behaviour at 43% and social data at 34%. Personalisation isn't dead. Assumed personalisation is on notice.

For mine, this is the clearest read yet on where the agentic travel wave meets the traveller. The winners won't be the products that decide the most. They'll be the ones that wait to be told, then do it perfectly. Watch the rate. Translate the menu. Book what I picked. The traveller stays the director. The AI stays the instrument.

One of the businesses I'm heavily involved with, Reely, is built from exactly this insight. Reely collects the choices humans are making themselves, saves them for when the planning phase hits, and uses AI to make the admin side simpler. Reely doesn’t snoop your social media, instead you tell it exactly what matters and what doesn’t.

The squeeze may land on anyone selling "AI plans your whole trip" to people who mostly never asked for that. The opportunity sits with whoever makes the chores disappear and leaves the choice exactly where most travellers want it.

With them.

YGO raises €8M to build the plumbing between travel and AI agents

The smart money is moving down the stack.

Berlin-based YGO has closed an €8 million Series A led by Ventech, with Sake Bosch, Caesar Capital and business angels, to build what it calls the AI infrastructure layer for the travel industry. In plain terms: making supplier inventory accessible, discoverable and bookable by AI agents, without suppliers rebuilding their technology stack or handing the customer relationship to a new intermediary.

The chat interfaces still get the headlines, but none of them work if the inventory underneath cannot be read, understood and booked by a machine. YGO's platform structures and enriches fragmented hotel and product data, plugs into the legacy systems the industry already runs on, and exposes the result wherever the customer starts: the supplier's own site today, ChatGPT, Claude and Gemini tomorrow.

It is not a slideware raise (don’t think those exist anymore, not in travel anyway). YGO says it already processes bookings for Condor, Bentour and GetAway Group, claims a 90% cut in manual operations, a 20% conversion lift for one airline client and a 2.4x lift in average order value for another, on more than €15 million in bookings so far. Co-founder Julian Kögel's announcement is here. Ralf Usbeck of TravelTainment and Peakwork fame is joining as well.

We have covered the protocol fights (UCP, PAP) in recent weeks. This is the same plumbing stack, except investors are now writing cheques for it.

And if that layer consolidates, it could become a toll booth of its own!

The AI discovery bump may be short lived

Activities operators could be forgiven for putting the champagne on ice. Fresh numbers are doing the rounds suggesting the direct booking tide is finally turning their way.

Alex Kremer's advice: keep the bottle on ice a little longer.

His point is that travel is late to this party. Other, non-travel sectors with much closer-in demand curves went through similar discovery booms earlier this year. The traffic surged, everyone declared a new normal, and the tide moved back out just as quickly as it arrived.

The lesson he lands on: do not take any gains in direct traffic for granted. Double down and capitalise on retention and cross-selling opportunities while the going is good.

I lived this at Urban Adventures when Trip Advisor was giving away free traffic and bookings whilst it concentrated on hotels. It was awesome and addictive and ultimately destabilising. I told that story in full recently on Mitch Bach’s Tourprenuer podcast. It’s in the first few minutes.

The lesson is a discovery bump you do not control is not a channel. The operators who actually win from this window will be the ones who turn a borrowed spike into an owned customer: the email captured, the second booking, the cross-sell into the next experience. Everyone else is just renting the wave.

AI is recommending products through videos brands didn't make

New research from Sundogs, covered by Net Influencer, compared 400 YouTube videos cited by AI shopping answers with 3,500 non-cited videos from the same creators, across US consumer packaged goods and auto.

Who made the cited videos:

• About 60% creators

• 30% editorial sources

• Only 10% brands

What didn't separate cited from non-cited:

• Views or recency (one cited car video was two years old, with 12,000 views and an AI voiceover)

• Production quality

• Creator fame

What showed up in cited videos:

• Formal reviews or physical demonstrations: 77%

• Framing the problem being solved: 76%

• Clear benefits: 90%

• Brand comparisons: 71%

• Product comparisons: 68%

• A practical, confident tone: 84%

• Length: usually eight minutes or more

Read that profile again. It looks a lot like good travel content: long-form, specific, honest comparison, showing the thing rather than selling the dream.

The caveat: this study covered CPG and auto. It doesn't establish these numbers for hotels, destinations or tours, because travel wasn't in the data. Whether travel answers work the same way is the open question.

Worth asking now, though. If the pattern holds for travel, AI won't recommend your property with your polished brand film. It will cite a creator's honest 15-minute walk-through, the one you didn't commission and may never have seen. Unless you did commission it - which is what we help smart brands with daily.

Worth knowing what the machines are watching and having a plan to get in front of it, the same way you did for Google + SEO in a past life.

72% of voters clicked nothing on Google last week

Jared Alster ran a quick LinkedIn poll this week that is worth your two minutes: "In the past week, how many times have you conducted a search on Google and clicked on either a paid ad or organic (blue) link?"

When I grabbed this, 36 votes were in:

• Zero: 72%

• A handful of times: 16%

• Once or twice: 11%

Not science, but interesting nonetheless.

The careful read: zero clicks is not zero Google use. A lot of searches now get answered right there on the results page, so people can Google plenty and never click through to anyone's website.

And 36 votes from a LinkedIn network is nobody's representative sample.

But if you make money from search traffic or paid clicks, that 72% is the number that might chill (but you already know that right? You can see it in your numbers too!).

Whether it is AI answers doing more of the work or that consumers are now mainly using the search bars in social, not Google or plain habit change across to LLM’s, the click through to your site is the thing quietly going missing.

The poll closes today. Add your vote or see where it landed.

If you think someone (or everyone) you know or work with could grow from being more informed on the topic of ai + travel (or could use the training above) then please forward this email to them and they can click the button below:

The travel manager secretly rebuilding her dashboard in ChatGPT

Daniel Senyard, SVP at FCM Travel, opens his Business Travel News piece this week with a story that should make every travel tech product person wince.

A travel manager at one of his enterprise clients exports her spend report once a month, uploads it into an AI chat tool, and asks it the questions her actual dashboard was never built to answer. She rebuilds the whole workflow by hand every month - explaining each column, telling the AI how to calculate each metric - because the tool she wants doesn't exist inside her travel platform. It exists in pieces everywhere else.

His answer is what the industry is starting to call "headless". The plumbing is Model Context Protocol (MCP), which Senyard describes plainly as a standard way to plug a business' own systems into an AI assistant so the user stops doing the export-and-re-explain dance. The value stops living in the interface and starts living underneath it: the data, workflows and content.

What FCM says it looks like in practice (his account, so hold it accordingly):

• Ask for hotel statistics with no date range and it defaults sensibly to the last year

• Ask for a three-tier nightly rate policy from actual spend, and it returns three priced tiers grounded in that client's real booking data

• Ask it to draft the traveller memo, the CFO summary and the review email, and one conversation goes from question to policy to communication

What is actually the system of record? Salesforce owns customer data. Workday owns HR data. Senyard's pitch is that a TMC - where company policy, traveller preference, real booking behaviour and supplier terms converge - can be that record for the business traveller.

If a traveller can get a policy-compliant answer or book a trip inside Copilot, Teams or Claude without opening a dedicated app, the app was never the point. And that dear reader is THE point.

Worth noting where he deliberately stops: around 70% of FCM clients' reporting is financial, which is where they've focused first, and traveller safety and duty of care stay outside the AI loop on purpose. An instant wrong answer about spend is a nuisance. An instant wrong answer about where your traveller is in a crisis is something else.

The winners won't be whoever has the prettiest app. They'll be whoever owns the governed data the AI is allowed to touch.

AI books almost no travel yet. That's the wrong number to watch.

The comfort stat doing the rounds: large language models drive well under 1% of Booking Holdings' room nights, and AI referrals are apparently below 2% at Booking and Marriott.

Skift's commerce piece this week argues the volume is the wrong thing to watch. The commercial terms the agents will run on are being written now, while volume is small.

What happens when Meta's Muse agent books, per the piece:

• Asked for a flight, it runs through Duffel's direct airline connection, with loyalty numbers and cancellations supported

• Asked for a New York hotel room, it opens its own browser and shops Expedia and Hotels.com unprompted

  So the airlines hold a defined commercial relationship with the agent, while hotels got shopped by software they never signed with

What Google's agentic hotel booking passes the hotel:

• Name, payment, occupancy, property

• The hotel stays merchant of record, but the conversation that produced the booking stays with Google. In the transaction, not in the relationship

Where the money sits:

• The two checkouts Skift describes both settle through a wallet the agent platform chose: Muse via Link by Stripe (one-time virtual cards), Google's tool via Google Pay

• Whoever's wallet an agent defaults to has a say in which card funds the trip. Delta collected $8.2 billion from American Express in 2025, about 14% of adjusted operating revenue

How loyalty holds up:

• Half of 1,000 US travellers in a McKinsey and Skift Research study changed a confirmed stay after spotting a better option

• Skift Research separately found 55% would follow an AI tool to a better deal, and loyalty membership ranked last as a reason to trust an AI recommendation

• Hilton already expects agents to reshop bookings until departure

Put those together and points start behaving like one more price an agent can compare. What's left is the part no algorithm can comparison-shop: how you're treated when you arrive.

Booking volume is genuinely small. But the integrations going in now will set the terms sellers trade on. Sitting out doesn't pause that. It just means someone else writes your part.

Honestly, this is not the time to be sitting out. I’d have every spare pair of hands, eyes and brains working out how you fit with personal agents right now.

Slack Group!

Podcasts and Sponsors

Podcasts now on Spotify and Apple Podcasts:

All podcasts are on Spotify and Apple Podcasts. Now is a great time to binge those you may have missed.

The latest released episode is my conversation with Michael Culhane on an agentic travel marketplace: better search isn't better judgement, and how traveller and supplier agents could negotiate the trip.

Listen to the episode here.

Partner with Us - sponsorship info for the newsletter and podcast can be found here.

Most clicked last week was Amy Wei's write-up of three and a half years with TripGenie.

That is it - you have made it to the end of this edition. If you are thankful for this newsletter, you can buy me a coffee.

If I have missed something, you have a tip or any feedback at all, simply reply to this email and it will come straight to me.