Hacker News

I don’t think it’s hard to understand why an airline would get tired of getting more refund requests and customer service escalations for refusal of same according to agreed upon terms by entitled jerks who don’t want to accept the consequences of the decisions their ShitClaw made on their behalf.

Sure, the airline may not notice when it happens but it can cut those processes short when they have a full ban.

You can better understand the context here if you go read between the lines of the Amazon vs Perplexity case, and take a look at the appeal - who else was supporting Amazon? You will notice a trade group called Airlines for America.

Why is this? If you go look at the financial statements of airlines (for which Airlines for America is the industry trade group), you will understand that a large portion of their profits come from monetizing human friction and attention.

Imagine things like upsells for their branded credit cards, feeding into the points game, etc. These are all things that appeal to humans, but that agents are almost certainly going to skip and then relentlessly price-compare the flight at the end of the funnel.

Now, once Amazon’s appeal was denied by the court, it became obvious that they needed to resort to technical solutions if they could not block it by trying to leverage the CFAA. Therefore, these industries are now implementing technical solutions.

Ultimately, personal agents are a huge consumer benefit, but they challenge the business models underlying much of the internet. I expect to see more of this kind of behavior from incumbents, which ironically leaves them vulnerable to disruption by players who do support agents, as this market shift plays out.

It is possible to detect Claude controlling your Chrome browser, as well as many computer use agents. Give this a try: https://demo.switchfrog.com. Ultimately though the mainstream agents will (and mostly already do) self-identify and businesses will adapt.

Source: I am the founder of a company in the space, Switchfrog.

So this is why United airlines and all the other folks should open up an API or MCP server for agents. When you do that, you make it easy for the agent to do the right thing, and go through a known access point which you can control and rate limit.

The user also has more fine grained permission over the interaction. They could choose to give permissions just to read account details or flight info or to actually make a purchase. United knows it's an agent acting.

Of course, unbehaved agents will still go around, but by providing a golden path, the better behaved ones will be better citizens.

The alternative is, as parent comment suggests, a user can allow Claude to drive their browser. Then, apart from heuristics, United Airlines can't really tell the difference between Claude driving the browser and a user driving the browser. Therefore you have less granular permissions and more unbounded actions and risk. There is no way to audit what is done by a human being versus by an agent. And United Airlines still has agents interacting with their site.

For the big corporations (consumer facing), they are not necessarily able to adopt change because:

1) they are under tremendous amount compliance boundaries, liabilities ambiguities and customer support issues, when agents go rogue or caused monetary damages to both service providers or merchants and consumers.

2) we collectively haven't figured out a way how agenic commerce would work. They have been some ideas floating around from Google, Shopify Cloudflare, but it nowhere near settled as a consensus.

3) Fraud/spam/cyber - Agents AMPLIFY these malicious interactions with their existing Human User Interface (HCI). In other words, the current HCI is not designed for agentic interactions.

For example: If your Meta Muse agent buys a $5k business class ticket from US to Europe using your Chase credit card that you find out after the fact, and you call your credit card companies for non-authorized transaction chargeback, who's responsible to the damages? Did you authorize the agent to buy tickets not?

If Chase denies your chargeback and make you liable for $5k to be paid, you pay for it or no? If you don't pay for it, United sues you in court, do you tell the judge you didn't authorize your agent to buy that ticket? Or you sue Meta instead? You see where this is going??

> How do they plan on stopping me from using Claude code with the Claude in chrome plugin, or some other computer use agent?

They need to make a best-effort demonstration that they are complying with the federal regulations.

The federal regulations require that flyers see notices about banned luggage and other requirements. The article covers this.

The thing is, they don't need to go to extremes to detect and ban everything. They can state their policy and then they have some coverage if regulators come knocking.

I would guess they're also dealing with the same problems that occur in other spaces involving reservations: AI agents tend to operate greedily and optimistically by default. A common situation in other industries is that they'll read a cancellation policy, then proactively make a reservation before asking the user if they want it. The agent will try to hold a spot and then schedule a cancellation before the window ends (example, 23 hours into a 24 hour cancellation period) and wait for the user to confirm.

As a result, the churn rate for agent-created reservations is very different than when humans do it, which breaks a lot of assumptions and financial expectations.

Unregulated spaces like restaurants can charge a non-refundable reservation fee or threaten cancellation fees (which is enough to scare off agents, for now at least) but regulated spaces like airlines may be required to process cancellations according to some rules. So now if you have agents everywhere exhibiting reservation/cancellation behavior that nobody expected and it's changing suddenly with each new update of Meta Muse or other system, it gets messy.

AI agents are a real problem for a lot of interactions that previously assumed humans were making the decisions. When an agent running 24/7 with a goal of maximizing its owners' happiness starts doing things, some of these systems break down.

So I can see multiple reasons why they'd want to officially disallow it. I would expect that making it a formal rule is enough to start getting Claude, Meta Muse, and others to add a guardrail forbidding it too.

Not going to stop the motivated individual running OpenRouter tokens and some complex local agent setup, but that's such a small number of people it can be ignored.

The misunderstanding is that they're not selling flights, they're mainly selling insurances and credit cards. The valuation of some airlines' loyalty programs is higher than their company valuation. That is: operating flights comes at a loss. Flights are just the bait, and then the process is designed to extract maximum value from the caught human through phsychological tricks. Agents are not so suspectible to those tricks, so the airline doesn't make enough money.

Must see by Wendover Productions: https://www.youtube.com/watch?v=ggUduBmvQ_4

It’s pretty easy to stop Claude or ChatGPT since they won’t fill CAPTCHAS or get around any step designed to stop them. They play nice. 99% of the population couldn’t figure out how to run any model that doesn’t.

So while it’s not an impenetrable barrier, it will almost completely solve the problem they’re trying to.

If you look at how airlines make money, you'll see that an agent skipping over the airline website's internal ads for the loyalty program, credit card partners, pushy upsells of seating/refundability, etc...destroys their revenue model. Not endorsing this approach of just banning agents, but it's easy to see why they don't know what to do yet.

They might not be able to stop you, but they can cancel your flight if they determine later that you violated the ToS or contract of carriage.

I don't think they're actually trying to stop you. Rather it's a public stance to protect them from customers' misaligned agents. They don't want to be liable for actions your agent took on your behalf; e.g. refunding a ticket when your agent bought the wrong one.

The "ban" doesn't have to be an actual technical ban, but a "ban" in the sense that you automatically forfeit your rights to customer service or a refund if a human doesn't do the booking.

>How do they plan on stopping me from using Claude code with the Claude in chrome plugin, or some other computer use agent?

Just like other companies that have similar low alternative services. They let you do it. They log and flag and as soon as you have a problem, instead of working to solve it, they see you are on the flagged list and you are now banned from the service. People who find themselves banned from an airline when they miss a connection or report lost luggage have done some other activity that got them to be flagged as a potential unwanted customer.

I made a Chrome plugin for VRBO sites that parsed pet policy on vacation homes as the info was all over the listing and not consistent for visual parsing. In spite of aggressive backoff, a weekend's worth of usage by me and vrbo put the kibosh on my IP.

The problem is that AI agents play all sorts of games to get better prices, and this guzzles resources that are supposed to be for dealing with legitimate problems. e.g. If an airline is dealing with an AI caller that's rescheduling a flight it intentionally booked at the wrong time to shave a few bucks off an airfare, somebody who needs to cancel a flight to attend a funeral will wait longer in line. If a human does this to save a few bucks, you let them. It's not a big problem because most people place more value on their own time. If everybody uses AI agents to do this for them, then you have to do something.

Airlines and other companies are faced with a choice: Let everything go to shit for people who don't use AI agents, or start discouraging the use of AI agents. It will be interesting to see whether companies embrace anti-AI AI agents or try to weight the penalties such that deterrence is achieved without them. e.g. Would you continue using an AI agent to book flights if you faced a potential lifetime ban from that airline?

I personally like the idea of AI surcharges. The airline could put it in their ToS that any use of AI booking agents will result in a surcharge. If your account gets flagged for AI agent games, you get an AI surcharge slapped on your flight. In the case of a false positive, if a human calls up and demands the surcharge be dropped, you do it. If an AI agent calls to complain, you double it. Most users of AI agents don't want to deal with customer service to get surcharges dropped every time they book a flight. This would make it cheaper and/or easier to just book a flight yourself... or use a very well behaved AI agent.

Why would they need to stop you? Why can't you voluntarily respect their wishes? What's with this entitlement that AI users have that they have a right to unleash their automation on services that have made it clear that they do not want that.

Ideally the wish to not have your service accessed by automation should also be legally enforceable. And we need clear precident that users and operators of automation are 100% liable for the damages they cause.

They won't with perfect accuracy, and I'm sure you're right that some methods will be a lot more difficult to detect reliably (e.g., a Muse agent that runs in a specific type of VM and is hosted in known data centers using specific network providers vs an LLM moving your laptop's cursor to navigate the website).

But pretty much every company that's involved with payment risk assessments, identity verification, or bot detection (Forter, Incode, Akamai, etc.) now has services they're trying to sell to merchants so that they can identify agentic traffic. Time will tell how effective they are with the different varieties of automation.

Some merchants want that capability to make sure legitimately authorized AI agents don't get unduly tangled in fraud prevention systems, but others, like United it seems, will use it to try to block them outright.

> I wish the simple “you have a product, I have money” relationship was possible

It is still but you're not shopping around prices based off "you have a product, I have money". You're shopping around because "you have a product that others also sell and I will pay any one of you the least amount of money I can get away with".

> How do they plan on stopping me from using Claude code with the Claude in chrome plugin, or some other computer use agent?

I proactively answered this very question in my own comment. Initially they will try and block your chrome plugin and if they fail, then those $50 flights will just be gone from the website. More context in my own comment but this is a very reasonable question to ask.

I don't know but given the fact we now live in a post-truth era and heavy AI agent/assistants users now embrace heavily hallucinations and have stopped checking output quality and sources anymore like they used to in 2024 and early 2025, I would do the opposite. Detect AI use as much as possible and charge them more automatically as the customer behind becomes the easiest to trick.

I wish the simple “you have a product, I have money” relationship was possible - right, and then your child spends 100k in amazon orders and we blame the merchants for not having better checks. Or the state fines the merchant for not ensuring that a person did indeed check in or attest to something personally. How do you prove the LLM collected your TOS agreement? I used to want this too, but having engineered multiple systems in regulated industries, automation will never be as simple as we want it to be.

Because the product is often not what we think it is anymore.

Surely the airline company makes money with the actual flight... but also they might start to make money with selling private data. Maybe they even start to make MORE money due to a commission on car rental, hotel, tourist attraction, etc than the actual flight.

The "product", due to pressure from competition, enshittification, novel alternatives, is changing and we as customers don't necessarily notice it. In fact the company is probably doing its very best so that we don't and can thus benefit with side effects that we are not aware of, otherwise we would ask for compensation.

So... I'm not justifying the practice but also business wise, it makes sense that they try to keep control of everything around their product.

> How do they plan on stopping me from using Claude code with the Claude in chrome plugin, or some other computer use agent?

They don't have to. The policy itself allows them to take consequential actions to address various issues in this regard.

With Captchas? But I do understand your sentiment

> How do they plan on stopping me from using Claude code with the Claude in chrome plugin, or some other computer use agent?

The endgame is just to raise prices for everyone. It's not like airlines are making a ton of money. Everything affects their profits and small margins.

The relationship is unfortunately different nowadays. «You have money, we are going to manipulate you». When the profit demands on them being able to manipulate you, an AI is bad news.

> Claude code with the Claude in chrome plugin

Chrome has this architectural problem where the webpage can see what extensions you have installed. United could just refuse to run if you have that installed.

> How do they plan on stopping me from using Claude

The righteous Claude would refuse to book on that airline. It already refuses to search amazon.com.

But I agree that at that point I'll just change the agent, or use local model that is less principled.

This is relevant: https://cdn.ca9.uscourts.gov/datastore/opinions/2026/08/04/2...

You don't seem to understand, it's fine for them to use AI to use "dynamic pricing" but for you to navigate their ridiculous refund/claim processes, using AI, that's "unfair"

You see how it's unfair when you do it... right?

> you have a product, I have money

> I prefer and will do my best to avoid detection or blocking

Looks like saga about exporting GPU to China :)

I have booked several flights using cdp browsing with my local agent. The more friction they add for agents, the more there will be for humans and the market will vote with their feet (or agents).

They're going to put it in their terms of carriage that if you use an AI agent to book a flight they get to dump you in the middle of the Pacific. Hope you're a good swimmer.

"like I have an adversarial relationship with many merchants" which ones?

From a technical perspective, does anyone know if Claude in Chrome is detectable beyond artificial movement?

I have been exploring know your agent solutions and nobody can give me an honest answer.

Yes, but how much do you value that money, and how much do you value our product?

:^)

Would you have a problem if instead of stopping you, they charged you more because you're booking with an agent?

"if you were me, then I'd be you, and use your body to get to the top." - ace venture

Because Claude can be legally compelled to advertise its "claude" acting on behalf of a human? Such a policy isn't that crazy.

Now sure you could bypass this by running your own harness/tooling, but then you'd be in breach of laws. There's various policies like this today around web scraping

> I wish the simple “you have a product, I have money” relationship was possible.

It is - these restrictions are part of the product. How you allow people to purchase your offerings is part of the business and has its own costs associated with it. How much more would you pay in order to have a LLM purchase a ticket for you?

> How do they plan on stopping me from using Claude code with the Claude in chrome plugin, or some other computer use agent?

Maybe they won't be able to stop you, but the average Joe won't be able to cover his tracks as effectively, or most likely, Claude Code and others will by default be required to assert the browser for the current session is being operated by an agent.

You always had an adversarial relationship. You just didn't see it until you tried to push back on the information asymmetry they wield.

If they notice you doing it, you won't be allowed to board the flight, and won't get a refund. This is the usual way things like this are enforced.

They can ban you by name and birthdate combo, united mileage number, and/or your debit/credit card pan if detected. They have no obligation to allow you to purchase air travel online, it is their right to refuse you service.

(know folks inside United, but not speaking in any way for the company, I just know how the process works, as a curious security person)

>(Once the corporate capture of the frontier cloud AI models is complete.)

By that time, the open obliterated models will have caught up and will be able to do this just as well as the current frontier models, if not better.

They can ban you.

> How do they plan on stopping me from using Claude code with the Claude in chrome plugin, or some other computer use agent?

They don’t.

When a dispute that either party(you or United airlines in this theoretical case) decides to litigate an argument far enough that it becomes clear you used an AI agent like Claude then you are going to run afoul of the CFAA[1] and are going to get fucked legally because they stated who could access their site and you violated it with your agents against their stated will.

This is a legal ward, not a technical one. Violate it at your own peril.

[1] https://en.wikipedia.org/wiki/Computer_Fraud_and_Abuse_Act

they don't care if you do that, because as far as they're concerned it's you that booked that ticket. you can't turn around and demand a refund or sue them because it "wasn't you, it was your agent".

they just don't want the possible liability of dealing with whatever legalities that might occur from autonomous agent usage. and i can't really blame em for that.

I've been saying this same thing. It used to feel like businesses actually appreciated your custom to some extent. Not all of them, not every time, but in general it seemed like they wanted to give you something in exchange for your money and saw you as a partner in their success.

Now it feels like a lot of businesses don't see the customer as any sort of partner. They're an obstacle that stands between the business and the customer's bank account, to be removed by any means possible. Providing goods or services in exchange for part of that bank account is merely an unfortunate and undesirable necessity that they haven't yet figured out how to avoid.

Trigger Passkey step.

Some of us have an adversarial problem with AI. I create things for humans. I don't want your AI on my website or reading my documents.

How do they plan to stop you? I don't know but the punishment would be cancelling your tickets if you get caught

There's nothing streamlined about any of these.

Booking a flight is obfuscated and you'll have to comparison shop across multiple websites or even devices to get the better deals.

Same with hotels - they have tons of rate obfuscation and you have to cross shop their individual websites and the aggregators as well as third party discounts to figure out best deals.

Concerts are similar but different. Sometimes they should be selling tickets via lottery but make everyone fight over them. Other times there are presales. Other times you have to shop through scalpers. Arguably tickets are less painful than the other two, but they're still a manipulative mess where you waste a lot of time.

The mismatch here is that you don't have a personal assistant, but the executives in charge of developing these personal AI assistants do, so they expect everybody to want one if only the AI gets good enough at it and the price is right.

Having never had a PA myself, I have no strong prediction here either way, but I think this can explain for a lot of the confusion of "who'd want this"/"who would not want this".

Hrm. Flight booking is ok, but the search optimisation problem is very tedious for anything more complex than a simple return between two fixed places on fixed dates.

Hotel bookings are mostly fine. Airbnb, bookings.com, fine.

But concerts? Concert booking is fucked. It's the main reason I don't go to concerts, I feel like I'm being absolutely screwed over by ticketing middlemen at every step of a horrible torturous process.

> it was always odd to me that "booking a [flight|hotel|concert tickets]" is the canonical example

Yeah, the weird part is how the people promoting the idea seem to believe the average (?) consumer will tolerate a particular risk/reward profile: It implies over-spending or cancellation fees or return-restocking fees are beyond reasonable concern, trivial compared to the minutes it saves the user who never does any verification. ("It's one banana, Michael. What could it cost, $10?")

Isn't it obvious to try to replace human travel agent with an AI agent?

Came here to say the same thing. It's probably because it's hyped towards people that already don't make these decisions, they have people they outsource it to book flights, reserve table at restaurants etc.

But for most people, we don't do it frequently enough that it's actually a big problem to solve. And when we do, it's a significant enough investment that we want a say.

It honestly just goes to show what a stress inducing mess airline sites are. People are desperate to offload it and avoid it, even at the risk of trusting a robot with a credit card.

If this approach of using AI to get around intentionally hostile platforms continues, it would be better for everyone to just regulate. Especially the environment, which air travel is already extremely bad for.

In the last millennium I worked for a newspaper and what we were promising advertising clients is that if they paid us more, readers could buy advertised products off our web site.

The example the publisher demanded we launch with was the ability to buy a car. Nobody bought a car off our web site.

We must be opposites, I despise booking flights and find the websites often fail on payments, are full of way too many steps, and require that you restart if you ever make a mistake.

It's truly weird that my 70yo mother can buy plane tickets or book a hotel without a hassle but someone claiming to be 10x programmer can't figure out how to use a website to do this and thinks that an agent is the best option. As you've noted, many airlines and hotels have made the process extremely streamlined (unless you're trying to get a "deal" in which case yeah, it's going to be more complicated).

Really, the problem is that a lot of people that are all-in on AIs aren't very good at a lot of life's little details so they think that AIs will solve everything and don't understand why nobody else agrees. If they got out and lived a little they would quickly realize that AIs are overkill or just the wrong tool for most things, but they have to actually experience life to learn that.

Amen to this, but it‘s not odd. It‘s a sign of how detached Silicon Valley people and more broadly tech people are from the rest of the world’s population. This is what they think a common problem really is. There might be another reason, i.e. building visual cases for these examples for demos is easier and marketing thinks that leveraging leisure elective activities is a good way to associate the product with a positive experience and perception.

i mean for most people, there's a real pain point. having to sit down and spend an hour or more booking flights, accoms, etc for a trip is a pain in the ass

same with reservations, ride shares and hotels and anything else. the K shaped economy in action. Rich people want meta glasses so they don't need to interact with the poors, even their own servants. I book flights rarely enough its kind of fun.

Maybe I have Stockholm syndrome, but what about the booking pages is super complicated?

I often use Kayak or google flights to screen multiple airlines, but once I've found a flight I like, I generally book it on the airline's site, and the UI is basically identical to Kayak. Put in some dates, some locations, click search, browse through the options, pick the one you want. If you are flexible on dates, there is a little +/- 3 days checkbox that will show a table of flight prices.

I'm probably stupid, but what are the improvements that would help customers find the best deals?

> I guess the best option is, to use a good reseller/travel agency that provides an API/MCP for booking and managing flights.

But do these actually exist?

I forget if it was in an official airline or an aggregator. But the page had a built in time for deal expiration that was a standalone javacript timer.

Imagine you're an airline. You have to follow the following regulations/standards:

- 24 hours to cancel any flight (or some airlines let you reserve with a 24 hour window to pay, which also fits the regulation)

- Seats that are sold cannot be sold to another passenger during this window (it's removed from inventory in GDS)

Now imagine you have people using agents to do the following:

- Book and cancel flights regularly to try to save a few bucks

- Set up an agent "if prices hit $xxx, book immediately and then cancel within 24 hours unless I tell you otherwise"

You suddenly have an inventory problem that can effectively DoS an airline.

>Somebody once said, that airline booking pages itself are one of the most sophisticated CAPTCHAs.

I think this is only true now that the AI's can interpret the pages correctly. Failing to decern the correct information is, I believe, the more human response.

It seems like the upside down turtle again.

> No idea if this is true, but they obviously don't want their customers to easily find the best deals.

CBS being, probably the most reliable and spin-free reporting, I believe the article.

FAA regulations and security are equally as good of reasons as "out to hide deals".

that's good.

I do not understand the airline's side of that argument, sorry. It seems like what you are describing is: 1. The airline is willing to sell a few tickets for $50 instead of $250, but only to the folks who it deems "worth it" and advertises accordingly. 2. Those $50 tickets are below cost, so each $50 ticket sold means everyone else will have to pay a little more. 3. If enough "general population" people find and buy those $50 tickets instead, the airline will discontinue the practice.

If this is indeed the case -- good riddance to that practice! Why should I be sorry that it goes?

This is true of a whole lot of stuff. Many special offers are loss leaders that largely rely on human behavior to make up for the losses, like a credit card that gives you $500 cash back if you spend $2500 in the first 3 months. This only works when a fair number of the humans stick with that card long term, if there are agents running around creditmaxxing then the offers don't work any more.

I don't follow why an mcp server buying a 50 dollar flight would make costs to up? Surely an mcp server booking flights doesn't mean there is more demand for flights overall? I'm not about to take a trip somewhere just because I have an mcp server to book it.

Yes, the question about AI agents is whether they are actually increasing productivity, or just allowing individuals with more compute to outcompete for a larger share of a finite resource.

If AI agents were more effectively managing flight paths, determining pricing, managing communication channels on the supply side, it would technically decrease cost per flight, but on the demand side, if AI agents just find the cheapest flight and book it, all it does is beat someone else who likely would have gotten to that flight a bit later.

I do think eventually it's going to be supply agents communicating with demand agents that lead to a coordination equilibrium. But until then it's just a race to who has the faster, more integrated models, and now, who has models which can bypass the restrictions providers place on their use.

> The airlines are trying to reduce prices for everyone without giving specific people (atleast those who they don't deem "worthy") an edge. Here's a great, short comment: https://news.ycombinator.com/item?id=50459444

Your sentence and the comment you linked say completely different things. You say airlines are trying to reduce prices for everyone, the comment says they're trying to achieve perfect price discrimination to maximize revenue. Those are not the same thing.

Your argument doesn't make any sense. Why is the flight even offered if it's such an issue for the airline?

Indeed - not sure if it is the case here, but businesses sometimes want to change richer people more and poorer people less for the same thing, if they can make more money doing that. One of the ways of doing that is by making the cheaper prices a hassle, a classic example being that you might have discount coupon vouchers at grocery stores that richer people won't regard as worth the hassle. If AI can get rid of that hassle for lots of people then the business case for doing it stops working.

I am copying my comment below here because there is a fundamental misunderstanding of consumer surplus here:

In your example ($50/$100), the airline is extracting all consumer surplus. This is called perfect price discrimination, and it is bad for consumers.

There is nothing stopping the airline offering Bob $50 per seat in the second scenario, other than profit preservation. If they were making a profit on the $50 ticket in the first scenario, they are making it in the second. They just happen to also be making less profit on the $100 ticket.

In effect you are making corporate profit a constraint on the market, which is not the case.

Your example actually shows that the welfare extractive pricing algos companies use today can be undermined by AI, which is good for consumers in this narrow example.

> For example, if you had an MCP connector to United and found a flight from SFO to DEN for $50, the rest of the seats have to go up in price to compensate.

Just to be clear, your argument is that it's "counterproductive" and a "local optima" for people to have complete and accurate price information available for purchasing their tickets? And that making this information harder for them to find is better for them? Can you point to any real-world examples of an asymmetry in pricing information ending up saving the side with less information money rather than the side with more information to use that in a way that disadvantages the other side?

> Well, eventually they will try and block you and then those $50 flights will just be gone from the website.

Since you explain that the rest of us are effectively paying for that subsidised seat, I’m okay with this.

So… please excuse my ignorance if I’m being ignorant…

But isn’t this the same argument, more or less, that police make for not wanting to have body cameras on all the time, because they would then be “forced” to give everyone who was breaking the law a ticket or to arrest them instead of letting them go if it “feels like they should”?

The analogy being that the police could pick and choose who they should arrest versus who they should not and let off with a warning - who is “worth it” (i.e. generally, people of, umm, not color, it turns out).

Anyways… it does seem that some kinda of preferential treatment could be lost here. But preferential good treatment for some is kind preferential bad treatment for others, so theres that.

Feel free to ridicule my argument.

If the 50$ flights are gone from the site, won't another site sell those instead?

Later on, assuming all of the carriers are affected by this, flights will get more expensive and maybe we will invest in other modes of transport that is cheaper

> Personal AI agents cause a local optima while overall increasing cost for rest of the travelers.

This has nothing to do with AI. Any sale pricing cause the average non-sale price to go up! This is why many companies outright refuse to participate in sales outside of damaged / opened goods.

> Now United isn't against offering a $50 flight to customers who they deem "worth it" - they just don't want it to be anyone who can run an MCP connector.

I don't understand the premise you're painting here.

Airlines lower the price because they have unfilled seats that are not selling at the current price. The price is lowered as an intentional choice to reduce margins on that seat because making any margin over the fuel cost of that passenger's additional weight is worth more than getting nothing at all.

They aren't thinking about the customer and how lil' Timmy wants to visit his dad in Idaho, they're maximizing profit by rapidly adjusting their service to the supply-demand curve.

> For example, if you had an MCP connector to United and found a flight from SFO to DEN for $50, the rest of the seats have to go up in price to compensate and not just by a few cents because the nominal cost of an economy seat is likely $250, not $50.

Your premise is false, as it rests on the logic that there is a zero sum game between passengers. Airlines offer tickets at the level and price they feel like they can sell them in the marketplace. There's not some invisible total allocated to passengers on each flight, where if one person "gets a deal" others pay for it.

Is there some issue with pricing being out in the open and the same for everyone. I understand discounts here and there to fill a flight, but why not offer them to everyone? Why do you care who is in the seat?

If I need to fill a seat last minute at a ridiculous low price, I'm glad an MCP client will fill it for me.

If the concern is that everyone will wait until last minute, that sounds like a problem that solves itself, last minute tickets will increase in price until the minimum price for air time is achieved.

Economics and all that...

That would make sense if not everyone was using comparators like Google Flights and SkyScanner. It doesn't make any sense in the real world.

I'm not sure if this was your point, but AI will eventually just flow to the competitor that wants business and has the best price.

Want to block bots? Okay, watch your customers just seamlessly move to a competitor without requiring even the slightest thought or action from the customer.

Of course, if--somehow--customers moving to a competitor doesn't actually hurt the company they're moving away from... I might have to start questioning exactly how this whole "free market capitalism" experiment is working out. I think companies want to compete to win customers, but I'm not 100% sure anymore.

It's funny. For a few days I had my own personal bot. It didn't browse many pages, but was enough to have me blacklisted from so many sites.

A lot of online retailers refuse to load for me as a human. But Amazon does. I used to avoid them in the past but guess who's getting my money now.

It's not the search's processing cost, it's the lost opportunities to make you pay more for the flight.

Flight reservation systems are some of the most complicated on the planet. They are specifically designed to maximize profit while meeting constraints imposed by airports (such as monthly minimums for seats sold per month to keep their gates). Prices change every few minutes and depend on the utilized capacity of all segments that a route touches. There's a reason most discount sites use the backend operated by Google instead of writing their own.

Can be a few cents for Sabre api calls. This is the gds fee on booking and excess querying, that should be rolled up into doing business but it depends of course, for some airlines every query costs them a fraction of a cent and if every employee agent is doing them alongside customers doing self service, yeah

But at the same time they should’ve just made it more accessible and just raised the prices anyway.

About 1 cent per search once users hit GDS "look to book" limits and start incurring penalties.

Inflation adjusted, airline prices are just a bit higher than historic lows.

Nothing stopping you from starting an airline if you think you could do it better.

You could expose all prices as add-ons, that way it's totally transparent. Ryanair does that and people love them…

/s

Both things can be true. Airlines can be miserly rip-off artists, and we can start seeing lengthy waits at screening points because people had agents make their bookings and weren’t reminded about the things they aren’t allowed to pack.

Claude: This form requires me to agree I won't bring explosives on board. If I reject this form, the booking won't go through. I should accept this to complete the booking.

I think it may be worthwhile to not compare basket sizes but maybe revenue over time? Would be interesting to see if the same agents (maybe bound to the same user or same project) now default or bias towards your storefront in the future, or if they always go out and do a search for "best deal".

Good, I hate this world where planning and micro optimizing my life has so much advantage over living in the moment, I want my AI to deal with all that (and restaurant reservations, credit card churning, etc) I can live my life with humans

> There's lots of speculation that mortgage rates will have to increase if people use AI agents to start optimally refinancing their mortgages

What does “optimally refinancing” mean?? Really. This sounds like spreading FUD.

A perfect decision about financing would need to know future interest rates and how long you will stay in the house.

From someone who's worked in the business travel industry - more than you think. Business travelers are a disproportionate amount of airline revenue (possibly over 50%+ of revenue while being around 10-20% of passengers). Those who are doing the same trip over and over again probably don't care, but salespeople who are doing different trips every week get some value out of this kind of thing. There have been piles of startups trying to solve this (and failing).

There's not a ton of money in skimming commissions off of flights (probably < 1%), but there is if you can attach a hotel booking to it (upwards of 10%), so if you can become the preferred vendor for a business traveler, there's money to be made.

Hmm I definitely think having an ai agent find me a flight is something useful because it’s very difficult to compare flights, there’s intentionally a lot of friction. The way I’d use it is search for a flight myself and have it search independently. If I find a cheaper flight that suits my exact needs then ask it to beat it. Whoever finds the best flight, that’s the one I book. I’ll repeat that, that’s the one I book.

Unless you have money to throw away I can’t imagine not checking the details and comparing prices. Hell, even if you do have money to throw away you still want to double check everything just before the last purchase decision. Is it going to the right city? Is it at a reasonable hour? Is baggage included or purchased? Etc

I definitely do. Astrophotography is a hobby of mine and I've been working on a kind of personal app to find me locations in the US that will yield good shots. The variables are moon phase (close to new as possible), weather, cloud cover, humidity, and plane ticket/car rental prices. All things considered, if it's all the same in nevada, new mexico, and utah for a given week, I want to find the cheapest flight to any airport in that area. It's a huge pain in the ass to trawl through all the different flights/airports/dates to minimize my spend.

I would like to be able to have an AI agent book my flight. Airline websites are often horrible to use and it's annoying having to fill out all the details. On more than one occasion I've had to book through a third party or choose a different flight because the airline's website was broken and I couldn't complete the transaction.

Google flights is great and I use it every time I want to fly, but it's very rare that you can purchase the actual flight on google flights too.

Booking a flight is a typical task some big wig will ask their company or secretary. I think that's why all these demo focus on it.

At this point so much hot tech stuff is incredibly weird niche things to solve the rare problems of overscheduled Silicon Valley tech VPs. They're not real problems that regular people have.

I mean you could also take the pure market economy conclusion that a competitor might allow agents and you buy the cheaper flight there.

I was having an agent search for a set of skincare products to the lowest price within a certain budget, and a lot of sites just blocked it, so I ended up buying it from a site that didn't.

Accountability

What are you going to do when muse books the wrong flight non-refundable? Will you go to Meta support or airline support?

They can look at things like the path the mouse takes, or the distribution in delays of actions.

Probably many other things too. But it's an arms race.

They can't. Maybe they can today, but it's a cat-and-mouse game that heavily favors the agent.

How is Cupertino any different to AppleScript (and enabling AppleScript in chrome), out of curiosity?

Absurd if they think so. It’s just going to have the AI companies build an official browser plugin.

I think it’s as simple as agents are hard to upsell on credit cards and seat upgrades.

Heh have you seen how slow airline sites are to get results? Now imagine them buckling under the load of an agent launching 20 simultaneous browser sessions to price-compare all minute variations of a specific trip request

I run a small e-signature service and went the opposite way, I gave agents their own door (an API and an MCP server. I never made money from the pages they skip. No ads, no upsells in the flow. People use it free and software pays per document past a free allowance, so an agent is a customer.

United makes a real share of its money on the screens between search and pay. An agent moves right past all of that. Idk, I think eventually everyone will accept the AI agents.

Airline margins are incredibly thin.

They want a person to go through the checkout flow because people might click on the high margin ad-ons they offer. United has a bunch.

> but is not realized is that on a competitive market with multiple players you will get better prices when others pay worse.

I'd rather have uniform pricing than have to hunt to not get fucked

It's likely more trying to block services from starting up offering ways to book the absolute cheapest with auto repricing.

How many flights do you book per month for all this hustle to be worth the time?

By adding a message on the website telling agents they are forbidden by ToS to use it.

I guess that would be a great opportunity for non-tech people to discover local LLMs.

"What if the airlines instead focused on delivering great service at a competitive price?"

Germania, IIRC, tried that in the late 90s, offering a flat fare for every seat.

They had to roll it back very quickly because their bookings dropped unsustainably - nobody wanted to pay a fair price for a seat, even if it was cheaper than they might realistically have had to pay under dynamic pricing.

All that potential customers fixated on was that the flat fares were more expensive than the few cheap loss-leader seats advertised on rival airlines.

"The market can remain irrational longer than you can remain solvent" etc etc

Those used to exist, nobody flew on them and they went bankrupt. The market has decisively decided. In airlines, price beats all else, even time in most cases.