I normally don't like plugging on an AI chat into my projects, because it's seems too easy and basic I think you should instead rebuild entire projects from the ground up to be AI first, not just add some AI button But in this case is already AI from... the ground up (it collects data and rates hotels with AI), so here it makes sense I think So you can now can talk to the hotel assistant and ask it to find hotels with a 🏋️‍♀️weightlifting gym, ✨newly built, 🔝highly rated or ones that have 🥯 cinnamon rolls for breakfast 😋 And it controls the site, moves the map, opens hotels for you! Hope you like it 😊😊😊 P.S. it doesn't make money, it's just my contribution for now to fight the enshittification and get you a good hotel for a good price!show more

I have a fitbit Air since 3 days and stayed in a hotel in Antwerp last night where.. guess what. The AC turns off after 12; environment yadiddi. Since it was like 34 degrees the whole day, which heated the walls, the temperature of the room instantly increased and I sweated my ass off; impacting my HF and sleep performance like crazy. (Normally it instantly goes to 52 and stays there; see pic 2) Tldr. How to flag hotels. It currently doesnt show "woke" xD B&B Hotel Antwerp; everything else was amazing but this on a hot day was totally rekt

I can’t click or search properly on mobile. Unsure if it’s the X browser or something else but really wanted to try 😂

I genuinely do not believe the revenue you post

X AI?

Ja alles is xAI op mijn sites!

Think I'll try that too, maybe bit cheaper than OpenAI

Some hotel information is occasionally incorrect. In this case, the hotel is in the centre of Tarragona (as stated in the text), but the details incorrectly show the city centre as 82 km away. It also wrongly lists the hotel as being in Barcelona.

Thank you, fixing!

Really this and the other tools that you built for nomads. I'm always using them when traveling. Something that I wrote about enshittification:

Do you think this will be the new user interface or will everything turn into something like a backend plus an MCP server and we all use our own clients?

love ! my to-go website to get honest hotel feedback thanks @levelsio

Getting something went wrong errors. Also on mobile, it auto reframes the chat screen to hide the submit button.

Badass!

Bug: Help modal doesn't show properly. I was genuinely interested in understanding the background and how this works, but modal doesn't show and scroll properly.

Oops thx, fixed!

thx for fixing. how do you select which hotels get on the list?

Tested it. It zoomed straight out to the full europe map like problem solved 🙃 Needs some tuning on the zoom and results focus. Good effort though. Once it clicks this will be way more useful than old school filters! Real estate properties need this approach too.

- would be good if it skipped booking programmatically with a booking site and used whisper to call the hotel directly to make the booking. But I see why you may not have done this already - I don’t think lots of people would use it enough to justify a subscription BUT paired with a flight booking system it would - especially if it tied into your flight safety report

Bug - Input is hidden by safari bottom navigation

On MacOS or iOS?

Fixed I think

This is great, going to check it out. I tried Odessia for booking a holiday to Poland and was disappointed. My go to stack is still Google Flights and Maps (for Hotels) and it is super slow filled with ads that objectively make it harder.

looks great, any plan to allow user to update info if wrong. I noticed some types on my hometown Bareilly hotel.

IMO the assistant finding you a hotel is the easy half. The price is the hard part, the same room is a different number depending what country your IP sits in. The agent that just books the cheapest version is the real unlock

I just wish there was better data to help me travel with my family. Just finding a place as a Single traveler really isn't that much of an issue...

Bug report: it shows some hotels that are closed, for example `Ikador Luxury Boutique Hotel & Spa` Feature request: ability to filter on family rooms (4 or 5 pax), I tried asking the chat but it seems to hallucinate which hotels have these.

Can you add a filter for hotels for families? Meaning big rooms (2 bedroom suite, etc). When you have many young kids its not ideal to be split up in multiple hotel rooms. Takes me hours of scouring the internet and calling to find one.

Been thinking about this for my site (similar but for local events, not hotels). Are you giving an agent tools to query a dB or just embedding all the info in the context?

Here's how it works: The one clever idea: the AI never searches the database or makes up hotels. It just decides what you want and then drives the real website UI for you — exactly as if you'd clicked the filters yourself. So every search runs through the same code as a normal human search. No separate "AI path" that can drift or hallucinate listings. The flow, step by step 1. You type something like "cheap hotels with a gym in Bangkok" or "which of these is best for a couple?" 2. The browser sends your message to a small backend endpoint (/ai_chat) along with some context: the chat history, what city the map is showing, and the IDs of the hotels currently on screen. 3. That endpoint is a thin proxy to Grok (xAI). Grok doesn't get database access — it gets a list of "tools" it's allowed to call and a description of what each does. 4. Grok reads your message and either: - calls a tool with structured arguments (e.g. search_hotels(city: "Bangkok", chain: "Marriott", amenities: ["gym"], sort_by: "price_asc")), or - just answers in text (for questions or travel advice). 5. The browser receives that decision and operates the actual filter controls — it picks the city, sets the chain dropdown, ticks the amenity checkboxes, sets the sort order, moves the map — and then runs the site's normal search. The results you see are 100% the real search engine, not the AI's imagination. The tools Grok can use - search_hotels — city / country / region, chain, min rating, max price, "newness" (built in last N years), amenities, sort order, boutique/collection. The browser maps each of these onto a real dropdown or slider. - find_hotel_by_name — global search for a specific hotel. - open_hotel — opens a specific hotel's detail card ("open the Four Seasons one"). Asking questions about the results When you ask "which of these is cheapest?" or "any with a sauna?", the browser sends the on-screen hotel IDs up; the backend pulls those hotels' real details (price, amenities, wifi, pros/cons) straight from the database and hands them to Grok as context. Grok then answers from real data only — it's told not to invent anything. Nice touches - If a place isn't one of our ~1,400 known cities (say "Bali" — it's an island), it geocodes the name and flies the map there anyway. - If a search returns zero hotels, it honestly says "couldn't find any — try loosening the filters" instead of pretending. - Your conversation and whether the chat is open are remembered (local storage). - It's rate-limited (per-person and global daily caps) so it can't run up a huge bill. Why this design is good Because the AI's only job is translation (English → filter settings) and reasoning over real data — never retrieval or generation of facts — it basically can't show you a hotel that doesn't exist or a price that isn't real. The trustworthy parts (search, ratings, data) stay deterministic; the AI just makes them easier to reach.

Awesome, thanks for this… pretty much validated the approach I was thinking about 👌

Wow, the response times and the UI updates are really smooth and fast on here. Well done.

Thank you man!

This feels less like a hotel search engine and more like a travel agent that never sleeps. what query made you realize the AI assistant was genuinely useful, not just a gimmick?

how do you deal with costs though? it inevitably becomes an issue once you get enough users.

Not really, LLM costs are super super cheap

does it also list like really cheap hotels? (for us who haven‘t made it yet)

Areyou using webMCP ?

What is your architecture for map based web apps? Is everything server side rendered? What frameworks do you use for the tiles?

Sorry, something went wrong — try again?

I think in future people will be bored of AI, and simple apps which integrate AI seamlessly will Win

People would pay for an hotelist MCP to connect their AI agents with trustful and curated info (not the SEO/marketing bullshit in other platforms)

I have been testing built in AI Chats in several products and that it is not working either. Two main takeways: 1. trust - you have to go lengths to prove that this is not just another dumb chat 2. commitment - engaging with chat is commitment. Almost like downloading an app

Love this but it's not working for me

This is the right way to do it. Most people slap an AI button on top — you rebuilt the whole experience around intelligence from day one. That’s why it feels like magic instead of a gimmick.

Imagine making an AI system to optimize hotels while being unable to distinguish “energy policy” from “woke agenda.” The algorithm is smarter than the user.

This is pretty cool. I just built a site that’s niche without AI that will have vacation type of listings. Mostly because I didn’t build it I used boxed software because it’s just a fun project (now that I’m using Claude deeply at work I realize my mistake) But I will rebuild it with my own code once it’s launched. Two concerns I have with using ai 1. The sheer cost of using it for listings, but I assume I can mitigate most of this with super aggressive caching 2. Copyright issues with the images and such that I’m able to find and use with AI #2 is the one I know least about but I’ll figure it out