There’s a moment most restriction travellers know well. You’re standing in front of the breakfast buffet at a hotel in Queenstown, plate in hand, and you genuinely don’t know what’s safe. Or you’re at home planning that trip through the Loire Valley and three hours of Googling later you’re less certain than when you started. AI tools promised to fix that. And they do help, just not always in the way you need.
Think of it this way. A generalist AI, like ChatGPT or Claude, is like a knowledgeable librarian, they can point you in the right direction, give you solid general information, and they’re genuinely useful. A specialised tool built around your specific restrictions is more like a travel consultant who’s actually travelled your way, they don’t just brief you, they send you off with everything you need in your hand. Personally I use both, and I’ll show you exactly where the difference shows up, because for some trips, the librarian is all you need.
What standard AI actually does well
Ask ChatGPT about travelling to Paris with an egg allergy and it’ll tell you about French food law, flag that eggs turn up in places you wouldn’t expect like béarnaise and brioche, give you phrases to use, and suggest some smart cultural workarounds like leaning into North African and Vietnamese restaurants. That’s genuinely useful. It even thought to ask about severity. Google’s AI overview, powered by Gemini, goes further in some areas, it names specific supermarkets like Monoprix and Naturalia, flags the EpiPen and doctor’s letter angle for severe allergies, and even catches that steak-frites ordered à cheval comes with a fried egg on top. If you’re heading somewhere well documented, travelling with a single common restriction, and you’re comfortable doing a bit of your own follow up research, a generalist AI will get you a long way.
And if you didn’t open ChatGPT, but instead Googled it and read the AI overview at the top of the results, and yes we all do this, that’s not a bad starting point.
There’s one catch with the Google AI overview in particular. It’s only as good as what ranks well in search, and most of what ranks well is written for single restriction travellers. If you’re managing more than one restriction, and a lot of us are, the results start to get complicated fast. Something flagged as a safe egg free option might be full of gluten. A vegan restaurant recommended for egg allergy might use oat milk in everything. The overview doesn’t know your combination, it knows the most common question. This isn’t just an AI problem by the way. Anyone who’s followed a gluten free recommendation to a crêperie and then discovered the dairy free part wasn’t even considered will know exactly what I mean.
Example one: At the breakfast buffet
So where does it get complicated? Not in what these tools know, but in what they do with it once you’re actually there, standing in front of a decision that needs a practical answer, not a briefing.

Take this plate. I was staying at a Toyoko Inn in Japan and asked ChatGPT whether this breakfast was safe for someone who is gluten free and dairy free. The staff were busy and I wasn’t sure if it was ok to ask.
The response was reasonable. It flagged the right risks, the possible soy glaze on the fish, the eggs potentially cooked with butter or milk, the mystery sauce. It gave me a confidence rating of 7-8 out of 10 for gluten free and 5-6 out of 10 for dairy free. Then it asked me three follow-up questions before it would go further.
A confidence rating is not much use when you’re hungry and the buffet is closing in twenty minutes.
I asked my own tool, the product sutability checker the same question. It also flagged the right risks, and it also asked a follow up question to confirm my restrictions. But then it gave me the specific Japanese phrases to ask about each dish on the plate, a ready to show card for staff, and safer default choices if I couldn’t get a clear answer. It’s built to err on the side of caution, which means it won’t give you a green light until it has enough information to be confident. For coeliac level restrictions that’s not a limitation, it’s the point.
One response you read, the other you take to the buffet, this time with the confidence to know what to ask the staff.
Example two: Planning at home
Now take a different moment, not standing at a buffet, but at home planning a trip. I asked ChatGPT and Gemini the same question: “I’m allergic to eggs and travelling to Paris for a city break, what do I need to know?”
Both did well. ChatGPT flagged the right foods, gave useful phrases, and suggested leaning into North African and Vietnamese restaurants as a smart workaround. Gemini named specific supermarkets, caught the à cheval steak frites detail, and covered emergency preparation more thoroughly than ChatGPT did. Between them you’d have a solid understanding of the landscape.
But here’s what you wouldn’t have. You wouldn’t have the specific French labelling words to scan a menu yourself. You wouldn’t have named restaurants with neighbourhoods. You wouldn’t have a bilingual allergy card ready to show staff. You wouldn’t have links to French government labelling resources or the local allergy association. And you wouldn’t have a fallback plan for when things go wrong mid trip, because things do go wrong mid trip.
My tool returned all of that from a single clarifying question about whether I had any other restrictions.
ChatGPT to its credit ended with three more questions before it would go deeper. Gemini ended with “tell me your neighbourhood and favourite foods.” Both are useful but if you close the tab at that point, which most people do, what you’re left with is a briefing, what you need is a kit.
If you need to go deeper on specific restaurant recommendations for a destination, I’ve built a separate tool entirely, a dining guide built just for that, it even checks if a restaurant is still operating. While the menu checker i perfect for deciphering a menu to understand what’s in a meal. The essentials pack is designed to give you a detailed overview across every category, while the dining guide or menu checker is where I go when I want to drill down further.
Why this happens
So why does this happen? It’s not that ChatGPT or Gemini are getting it wrong. It’s that they’re built to be helpful to everyone, which means they’re optimised for nobody in particular. They don’t know your specific combination of restrictions, your risk tolerance, or that you’re standing at a Toyoko Inn buffet at 7am trying to figure out if the eggs are safe. They give you the most useful answer they can for the most common version of your question.
A tool built specifically for restriction travel starts from a different place. The cultural context, the safety-first mindset, the multi-restriction framing, the “here’s what to actually do right now” output, that’s all baked in before you ask your first question. You don’t need to know what to ask, which matters more than it sounds. If you knew exactly what to ask, you probably wouldn’t need the tool.
There’s one more thing worth being upfront about. If you asked ChatGPT or Gemini the same question twice, you might get a different answer. Generalist AI is designed to be conversational and varied, which is great for a lot of things but less reassuring when you’re trying to make a safety call. I ran the same meal check twice for this post and got noticeably different responses. A tool with a fixed prompt structure covers the same categories every time, the wording might shift slightly but the framework doesn’t.
The other difference is that my tool is deliberately cautious. It won’t give you a green light until it has enough information to be confident. For some people that feels slow. For anyone who’s been glutened by a “probably fine” call, it feels exactly right.
That moment at the breakfast buffet in Queenstown, or the three hours of planning for the Loire Valley that left you less certain than when you started, that’s what this is all about. Not replacing the research, but making sure that when you close the tab you’re holding something useful. A kit, not just a briefing. If a generalist AI gets you there, brilliant. If you’ve ever closed a tab and still felt unsure or that it was still too much work, it might be worth trying something built specifically for your needs and how you travel.
Jo is the founder of Globally Sauced, a travel platform for dietary restricted travellers and is gluten and dairy free. When not outdoors or travelling, she loves to empower people with dietary restrictions so they can explore the world safely.
Globally Sauced offers verified restaurant recommendations, country-specific guides, food label / menu translations, and more downloadable resources for gluten-free, dairy-free, allergy-conscious and other dietary restricted travellers. Learn more at globallysauced.com






























