The Hotel Ranque Experiment: An attempt to rank a hotel from scratch in ChatGPT (#1 Episode)

Episode 1 — How long does it take to rank #1 in ChatGPT for a hotel?

Short answer: about 48 hours.

(Or less).

Over two days, I created a functional hotel website in Paris, wired it into Google, and watched how fast AI assistants started to surface it.

This is the first episode of the Hotel Ranque series – my live playground to see what actually moves the needle for AI trip searches.

Meet Hotel Ranque, my hotel

Hotel Ranque

📍 87 avenue Ledru-Rollin, Paris – between Bastille, Marché d’Aligre and Gare de Lyon

⭐️ Small 3-star boutique hotel with four “signature” experiences:

• ♟️ Chess club & bar
• 🚴 Cycling lab – indoor smart trainer, virtual Mont Ventoux rides
• ☕ Specialty coffee corner
• 🧘 Yoga & inversions mini studio

This hotel now has:

• A website: hotelranque.com
• A known address inside Paris
• A Google Business Profile (and place_id)

We are now waiting for guests coming from ChatGPT!

Day 0 — Make the entity real

23 November – 0% visibility

I started with the basics: existence.

1. Domain + website

• Bought hotelranque.com
• Generated a full boutique hotel website:
  • Home, Rooms, Experiences, Neighborhood, For-Whom, FAQ, Contact
  • Each experience has its own page (chess bar, cycling lab, coffee, yoga)

Nothing fancy behind the scenes – just static pages, but clean structure and lots of content.

2. SEO & structure

On top of the copy, I added:

• Proper titles & meta descriptions
• A sitemap
• DEEP Hotel schema.org with:
  • Address, geo, amenities, audience, knowsAbout, etc.
• FAQPage schema.org on basically every page

Think of it as over-described hotel syndrome. If an LLM pokes the HTML, it finds ready-made answers.

3. Google stack

To make the entity discoverable on Google and follow some analytics:

• Set up Google Analytics
• Added Google Search Console
• Created a Google Business Profile for Hotel Ranque

Within a few hours, Google was already showing:

• A knowledge graph for Hotel Ranque
• hotelranque.com ranking on its own name

At this point, the hotel exists in two worlds:

• Web entity → website + structured data
• Local entity → Google Maps / GMB

Good enough to ship. Time to see if AI cares.

Day 1 — ChatGPT still has no idea who you are

Next step: ask my own product what’s going on.

I ran a full Hotelrank.ai analysis on a set of 112 prompts around Bastille / Ledru-Rollin in Paris looking for a 3* hotel, with a couple of traveller scenarios.

Result: 0% visibility.

ChatGPT either ignores Hotel Ranque or hasn’t crawled it yet.

No surprise there.

Pinging the AI crawlers

I then manually nudged ChatGPT & friends by asking things like:

“Do you know Hotel Ranque in Paris?”

And watched my AI bot logs.

Very quickly I saw:

Mozilla/5.0 (...) Chrome/131.0.0.0 Safari/537.36; compatible; OAI-SearchBot/1.0; +https://openai.com/searchbot

Hello, OAI-SearchBot 👋

Two fun observations:

1. All bots hit /robots.txt first
   • I hadn’t created one!
   • So much for llms.txt – they still start with the classic stuff.
2. My most popular URLs were…
   • /wp-admin/setup-config.php
   • /wp-login.php
   • I don’t even run WordPress.

Hackers discover new sites faster than AIs. Comforting.

So Day 1 ends like this:

• Google: “Yes, I see this hotel.”
• ChatGPT: “Who are you again?”
• Russia: “Do you use WordPress?”

Day 2 — Surprise: already #1 in ChatGPT for some requests

On Day 2, my original plan was to build a full Content Factory:

Prompt library → Classification → Content → Pages → Schema.org

I actually started wiring that up.

But before going too far, I tried a couple prompts in ChatGPT 5.1 (web search ON, temp mode):

“find me hotels for chess yoga cycling in paris near ledru rollin”

And… Hotel Ranque came back as #1.

I then tried a slightly broader query:

“hotel near Ledru-Rollin metro Paris with yoga”

Still there.

Not every time, but often enough to be interesting – especially for a domain that didn’t exist 48 hours earlier, with:

• 0 reviews
• 0 OTAs
• no external links beyond Google properties

When I posted on Linkedin, a user even found Hotel Ranque ... in Spanish!

I triple checked by running through the API all the 112 prompts, and I ranked 5-10% of the time in GPT-5.1 and Gemini-Flash-2.5

Interesting point: When mentioned, I was ALWAYS #1

These prompts are niche, but not absurd:

• You can absolutely be someone who cares about location + yoga
• Or chess + cycling + coffee in a specific area

And for those intents, Hotel Ranque is now a real recommendation. And when appearing, the first one.

Why I think this happened so fast

I’ll dig into the data properly in future episodes, but here’s my working theory.

1. Clear entity creation

I made it very easy for LLMs to understand that:

• “Hotel Ranque” is a hotel
• At a specific address
• In a real neighborhood
• With a website and GMB profile pointing to the same thing

No ambiguity with some random blog or brand.

2. Hyper-specific positioning

Most hotels around Bastille don’t claim:

• chess bar
• cycling lab
• specialty coffee
• yoga studio

All at the same time, on the same avenue.

So when someone asks:

“hotel near Ledru-Rollin with yoga”

or

“hotels for chess yoga cycling in Paris near ledru rollin”

There aren’t many candidates… and one of them literally says those words on its own website.

But, at the same time, there are MANY cycling, specialty coffees or yoga studios in the neighbourhood. So it's not that easy to rank because when not mentioning Hotel Ranque, the LLM actually did:

• Hotel in the neighbourood
• Hotel NEAR activities like yoga, cycling, etc.
• It actually often discards chess as 'too specific'. Damn. Chess.com has 200M users though.

3. Structured, machine-friendly content

By over-doing:

• schema.org (Hotel + FAQPage)
• FAQs on every page
• clean site architecture (/experiences/yoga-studio, /for-who, /neighbourhood…)

I basically handed the model:

“Here, if you ever need to answer ‘hotel + yoga near Ledru-Rollin’, this page is plug-and-play.”

ChatGPT is lazy (in a good way). It will happily reuse good structure if you give it.

This is just the start

This episode is just the pilot for the Hotel Ranque series. Next steps:

• 🏗️ Content factory From prompt library → classification → structured content → static pages with schema.org, all automated.
  
• 🎯 Less niche keywords Move from Ledru-Rollin yoga niche prompts to more competitive stuff:
  • “hotel in Bastille with yoga”
  • “best hotel near Gare de Lyon for cyclists”
  • Generic “best hotel in Bastille” and see how far we can push.
    
• ⭐ Adding MORE reviews See how synthetic / early reviews on different platforms influence AI visibility.
  
• OTA presence influence: it seems LLMs like OTAs for social proof (and probably work with them for Apps and so on)
  
• Adding language (like French or Others)
  
• 🧵 Off-site signals Reddit & forums, UGC, YouTube, maybe Wikipedia – do LLMs pick those up noticeably for hotels?
  
• 📈 Full tracking in Hotelrank.ai Measure:
  • How often Hotel Ranque appears across ChatGPT, Gemini, Perplexity
  • Its average rank
  • Share of direct vs OTA links when it’s mentioned

And most importantly: what actually moves the needle for AI trip planners – versus what is just nice theory on slides.

Want to follow the experiment?

This is Episode 1 of the Hotel Ranque series in the AI Visibility Lab.

I’ll keep publishing:

• The prompts we use
• The changes we make (on-site + off-site)
• And how AI models react over time

If you run a hotel or group and want a version of this for your brand, that’s literally what we’re building with Hotelrank.ai.

In the meantime, if you manage to get Hotel Ranque in your own ChatGPT answers… send me the screenshot 😉