How I made an app that LLMs can use: GEO for FOMO Sun

The thesis

When someone asks ChatGPT or Gemini "where is it sunny near Zurich right now?", I want FOMO Sun to be the source it cites. This is not SEO. This is GEO: Generative Engine Optimization.

What GEO means in practice

Traditional SEO optimizes for Google's link-ranking algorithm. GEO optimizes for how LLMs discover, understand, and cite your content. The difference matters because when an AI agent answers a user's question, it does not show ten blue links. It shows one synthesized answer, maybe with a source citation. If your data is not structured for machine consumption, you are invisible.

What I built into FOMO Sun from day one

llms.txt at the root. A plain text file that describes the service, its API, and its data sources in language that LLMs parse well. Think of it as robots.txt for AI agents.

OpenAPI 3.0 spec published at /api/docs. Any agent that supports function calling can discover and use the FOMO Sun API programmatically. The schema is versioned and stable.

Schema.org markup on every page: Place, TravelAction, WeatherForecast. This means when an LLM crawls our destination pages, it can extract structured facts without parsing messy HTML.

SSR destination pages at routes like /escape/chasseral. These are real server-rendered HTML pages, not JavaScript blobs. LLMs and their crawlers can read them.

Self-describing JSON in every API response. The _meta field includes data freshness timestamps, weather source attribution (MeteoSwiss CC BY 4.0, Open-Meteo), confidence explanations, and API version. An agent can assess data quality without external documentation.

Content negotiation. Send Accept: application/json and you get structured data. Send Accept: text/html and you get a human-readable page. Same URL, two audiences.

Why this matters for distribution

The next wave of travel planning happens through AI assistants. People will ask their phone "escape the fog near me" and get an answer synthesized from whatever sources the LLM trusts. If FOMO Sun's data is well-structured, attributed, and fresh, it becomes a cited source. That is free, high-intent distribution.

Early signals

We are tracking API calls by user-agent to detect agent traffic. The API also supports CORS for browser-based agents. The architecture is ready. Now we need a real fog season to see if the thesis holds.

This is part of the FOMO Sun build-in-public series. The app is live at fomosun.com.