The AI World Of Commercial Air Travel Sucks…

I use generative AI fairly regularly in day-to-day life, and in many ways I’m so grateful for how far it’s come. Whether I’m using ChatGPT or just using Google’s AI summary, there’s an incredibly high accuracy hit-rate. Whether it’s trying to figure out a travel destination, write some code, or just helping reformat a (non-confidential) document, I’m very happy using AI to generate some possibilities, or to figure out how something works (in the latter case I would always recommend using AI to uncover some terminology you don’t know, but sense checking the facts with a reputable website using the new terminology you’ve learned through AI).

However, the same accuracy and rigour doesn’t seem to have applied to anything related to commercial aviation. While I never trust an AI-generated source 100% (and you shouldn’t either), I have a much higher level of trust towards something AI-generated in most fields of knowledge compared to anything air travel related. This applies not just to when I’m searching something up, but also when I see an article that’s quite obviously AI-generated.

I thought I’d chat a bit about what I think has happened, and finish up with some recommendations I have for those hoping to leverage AI to plan their air travel. Just to be upfront, at The Alviator I personally write every single post, and will never AI-generate a post unless I’m specifically quoting something to prove a point. The only time I ever use AI is to collate research, and I’ll always ask it to cite sources (and check the sources firsthand).

(Yes, the cover photo is AI generated. Hopefully this proves my point. That must be the widest A220 I’ve ever seen, and sure, I’ll replace my headrest with another TV screen please!)

Current AI technology is really good for lots of things

I think that we’ve come an impressive way when it comes to how good AI is at collating information. Whether it’s trying to figure out how to do something (e.g. how do I unclog the U-bend in my sink? / how do I restart the Spotlight function on my laptop?) to figuring out recommendations (e.g. what app would you recommend for this coding project?) I find modern generative AI to be useful, and reasonably accurate in most cases.

I think that this extends to travel as well. ChatGPT is, unsurprisingly, pretty good at making itineraries. For example, when I was in the French Riviera with my family last week and wanted to do a day trip to Italy, ChatGPT recommended Sanremo, and specifically the old town of La Pigna. This ended up being a very easy journey from Nice, and we really enjoyed our wander – and I even found some of the best gelato I’ve had so far.

La Pigna, Italy

Air travel-related AI sucks

While I think that modern cutting-edge AI technology is great for figuring out travel plans, this doesn’t extend to its knowledge of the air travel industry. Maybe I’m just more knowledgeable in air travel compared to most industries, though if I’m just trying to figure out the best fifth-freedom flight to fly within Europe or how to redeem miles for a certain cabin product, generative AI just doesn’t give the most accurate or helpful picture in most cases.

That’s not to say that AI is completely useless when it comes to the air travel industry, though I have to spoon-feed the prompts way more significantly and sense-check results against sources much more rigorously than I have to with virtually any other field I interact with on a daily basis. For example, if I just try and search for the best transatlantic business class product, ChatGPT 5.2 recommends that I fly Qatar Airways, ANA, or Singapore Airlines. Singapore Airlines may have one transatlantic route, but Qatar Airways and ANA sure don’t:

What if I just want to find a way to fly Star Alliance (e.g. Air Canada) business class in the UK, where there aren’t many major opportunities to transfer points to Star Alliance carriers? According to ChatGPT, get an AmEx card and earn Membership Rewards points…..that can’t be transferred to Aeroplan. (While I also learned through ChatGPT that Flying Blue partners with Air Canada, I can’t find a reward seat on Flying Blue for an Air Canada flight.)

Maybe I’m over-relying on ChatGPT and other LLMs such as Perplexity are better at this, though I’m using ChatGPT as the industry-standard model, and there’s a significant difference in its understanding of the air travel industry compared to most other fields.

Here’s what I think happened

There’s a lot of knowledge sharing on the internet, and that’s a great thing. The same amount of information existing on the internet from the dawn of civilisation to 2003 is now produced every 48 hours, and there’s huge net gain for society in that.

There’s also been the growth of information that has been twisted and sensationalised to fit society’s narratives. I think that air travel has been highly impacted by this, and that’s mainly because society generally doesn’t have a great relationship with air travel, especially in longhaul economy class. Any “free” or cheap way to fly first or business class, any airline with a “tolerable” economy class, or any airport security hack becomes social media bait. Given that the average social media user doesn’t travel that many times per year (and the most frequent travellers are unlikely to be the most fervent social media users), you end up with some hugely overblown travel stories that shape society, and therefore AI’s view of commercial aviation.

There’s some credibility from blogs and travel sites that do what we do here, and provide firsthand experiences. However, I think that for every reputable air travel site, there are loads of other sources that get paid way more to churn out content related to air travel. That’s shaped the industry a bit, even before AI existed – for example, that’s where I feel Emirates gets their “halo” effect from, even though they were one of the first carriers to introduce a 3-4-3 product on their 777s.

Does this look like “the world’s best economy class” to you?

I think that generative AI has exacerbated this sensationalised part of the media, as opposed to learning the basics – for example, what people should expect from a business class/premium economy product, what defines a transatlantic flight, or what separates a good intra-European business class from a mediocre one. Instead, random junk information gets picked up by AI and publicised on social media – for example, the myth that using a VPN gets you a cheaper airline ticket, or that people get “free” flights by using their hard-earned miles, etc.

We have awesome resources out there on air travel, so this is frustrating

The thing that makes the AI junk so frustrating is that in an alternate world, the air travel industry is so well set up to use AI effectively. There’s so, so much data out there on aviation that will tell you exactly what to expect from a flight:

• FlightRadar24 provides up to 7 free days showing exactly when a flight took off and when it landed, which plane operated the flight, as well as where exactly the plane was at every point in time (and 90+ days for paid subscriptions)
• Planespotters.net will tell you not just how old the plane is, but also which airline owned the plane, and when the plane was reconfigured
• AeroLOPA will tell you not just how many seats the plane has in each cabin, but where exactly in the plane which seat is, which is helpful if you’re trying to figure out whether your seat will have a window or whether it’ll jut out into the aisle
• ExpertFlyer.com will tell you live how much availability your flight has, or which seats in the cabin (including the seat next to yours) are free
• Google Flights will tell you, on one page, how much every single airline selling a flight from your origin to destination will cost in a particular cabin class, in ascending order
• Google itself will tell you which terminal and gate your flight is departing from/arriving at
• Flighty collates most of the above information for your specific flight, and even tells you which baggage carousel your bag will arrive at
• That’s not to mention sites such as ours, and others on BoardingArea, that provide some of the most detailed firsthand reviews out of any product you can buy on the planet

The air travel industry has extensive data, possibly more than most other consumer industries out there

There’s so much data that a reasonably configured MCP should be able to pull from each of these sites, and tell you exactly what plane you’re flying, how old the plane is, what seat you’re flying in, what to expect, and provide reviews of other travellers that have flown a similar flight before. But frankly, that’s not the shape of the internet when it comes to commercial air travel right now.

Instead, the media has heavily sensationalised airlines based on branding and random hacks (many of which aren’t even true), leading to warped responses from LLMs whenever people try and find something related to air travel.

Okay, so as an air traveller how should I use generative AI?

Here are a few recommendations I’d make, as someone who’s been writing about the industry and also uses AI regularly:

• I would never trust any source that looks quite obviously AI-written – there are a few great sources out there, and ChatGPT itself can tell whether something was written using AI
• I think that when looking up really general questions (e.g. what the best premium economy class products are), AI does an okay job, as there are enough credible resources on this that feed into the training data; I would still recommend reading sites that do firsthand reviews for this, as AI also picks up on sites written by people who have never flown these products firsthand before
• When looking up specific point-to-point flights, I would always recommend using Google Flights over any AI site, as Google Flights has the best real-time pricing and also links directly to airline websites
• Again, I’d consider AI to be decently good just at listing options for airlines I could fly, but would use Google Flights or an airline’s reward search to check the price and availability for my specific day, and would book directly via the airline
• When trying to look up what to expect onboard a flight, I would first look at the seatmap (using the airline website, ExpertFlyer or otherwise) and AeroLOPA, as well as previous reviews of flights on the same aircraft on a similar route
• If I were to look for very specific pieces of information (e.g. does the last row in this premium economy cabin recline?) I would explicitly ask the LLM to cite sources

Conclusion

While AI has come a long way in day-to-day life, I’m not thrilled by the way that AI has shaped itself when it comes to the air travel industry. My hypothesis is that most of this comes down to a not-great relationship between air travel and the general public, and this has led to random (and in many cases untrue) travel hacks being prioritised over robust industry basics. I find that to be a shame, since the air travel industry has an incredible amount of data – more than most other consumer industries out there.

If you’re going to use AI for air travel reasons, always prioritise using trusted firsthand sources if you’re just finding out what to expect, and I wouldn’t ever recommend using it to book a flight. There are just much better resources out there at the moment.

How do you use AI if you’re a frequent air traveller? Do you have the same frustrations that I do?