Case study · Consumer · AI concierge
Mes Prestiges: a dining platform that runs its own growth.
Mes Prestiges is a curated fine-dining guide covering 30 cities in 10 countries, built around a verified venue database and an AI concierge. We designed it, built it, and operate it ourselves, which makes it the most honest kind of case study: every number on this page comes from our own Search Console and our own deploy logs.
It is also a working demonstration of a thesis we bring to client engagements: content, localization, and distribution can be engineered like software, so that one small team ships what normally takes an editorial staff, a translation agency, and a social media manager.
- Stack:
- Astro · Cloudflare edge · Claude API
- Status:
- Live in production
- Team:
- Built and run by one person
The platform in numbers · data as of 30 Jun 2026
3,023
curated venues, each verified and geocoded
30
cities across 10 countries
9
languages, every one a first-class experience
8,999
URLs in the live sitemap, all edge-rendered
445K
Google impressions in June 2026 alone
0
manual Instagram publishes, software posts daily
The traffic curve
Google first indexed the site on 21 December 2025. For three months almost nothing happened: 76 clicks in January, 130 in February, 108 in March. That flat stretch is the part most projects do not survive, and the reason we automate distribution instead of waiting on it.
Then the compounding started. April closed at 875 clicks, May at 3,018, June at 5,965 clicks and 445,484 impressions, ending the month near 390 clicks per day. All of it is organic search, measured in Search Console: people looking for specific venues, menus, and neighborhoods and landing on pages the platform generated for exactly those queries.
Organic clicks per month · Google Search Console
mesprestiges.com · Dec 2025 to Jun 2026
| Month | Clicks | Impressions |
|---|---|---|
| December 2025 (from 21 Dec) | 20 | 422 |
| January 2026 | 76 | 4,286 |
| February 2026 | 130 | 7,044 |
| March 2026 | 108 | 12,291 |
| April 2026 | 875 | 30,841 |
| May 2026 | 3,018 | 286,478 |
| June 2026 | 5,965 | 445,484 |
Where 8,999 pages come from
Nobody wrote 8,999 pages. We built an engine that produces them from a curated database, in every language each city supports, and keeps them consistent as the data grows. One new venue becomes a family of pages across nine languages, without a writer, a translator, or a CMS touching any of it.
The interesting part is not that the pages exist, it is that their quality is enforced by machinery rather than discipline. How that machinery works is the part we build for clients, so we will keep the details for a conversation.
Built to be read by AI assistants, not just Google
A growing share of "where should we eat tonight" happens inside ChatGPT and Claude rather than in a search box. Mes Prestiges treats that as a distribution channel in its own right: the platform is built so AI assistants can draw on its verified venue database directly, instead of guessing from whatever they crawled.
On the site itself, an AI concierge does the same job for humans: describe the evening you want, get venues from the verified database rather than model guesswork. The database is the moat; the AI is how you talk to it.
Distribution as software
The Instagram channel is operated end to end by software we built. It decides what to publish, produces the post, ships it, and learns from how each one performs. In its first month it published 21 posts with zero human publishes, and it has been running hands-off since.
The same philosophy runs through the rest of the operation: search engines and AI assistants are kept fed and monitored by systems, not by someone's to-do list. The founder's weekly job is curation and quality, everything repetitive belongs to the machines.
What this means for your engagement
- Growth can be engineered, not just advised.
- The curve on this page was produced by systems: a data model that generates pages, pipelines that localize them, and automation that distributes them. We build the same machinery inside client businesses.
- Small team, industrial output.
- One person operates 8,999 pages in 9 languages and a daily social channel, because the operating work is done by software. That ratio is the point of AI adoption done properly.
- Real numbers, from our own dashboards.
- Everything above is measured, not projected: Search Console for the traffic, the repository and deploy logs for the rest. We hold client work to the same standard of evidence.
Want a growth system like this inside your business?
Start with the free AI Opportunity Score, a 3-minute assessment that maps where AI can reach into your current workflows. Then let's talk about what to build.