# Seychelles T0 — publishing the seed catalog

Date: 2026-08-22 · Kind: measurement · Status: interim
Canonical: https://citezen.ai/field-notes/seychelles-t0-seed-catalog
Publisher: Citezen (https://citezen.ai)

48 licensed properties went live on seychelles.citezen.ai on 21 August 2026 as registry-draft (T0) profiles. This is what was published, what was deliberately left out, and what the first measurement window looks like.

## What went live

On 2026-08-21 the seed catalog for Seychelles was published at [seychelles.citezen.ai](https://seychelles.citezen.ai). Each profile is generated from the official licence registry and carries the same trust marker in three layers: visible on the page, as page metadata, and in the structured data an engine reads.

- **Coverage:** 48 of roughly 421 licensed properties. The catalog says so on its own index, in `llms.txt` and in `agents.md`: *partial seed — absence does not mean unlicensed.*
- **Tier:** every profile is **T0 — registry draft**. Nothing is owner-claimed yet, and the booking relay is off.
- **Only what the registry knows.** No prices, amenities or policies are inferred. Fields the registry does not have are left empty, and the machine-readable notes tell engines that explicitly.

## What the first days measured

A lightweight interim query (2 properties × 2 engines) was run on the first and second day after publication. Citation share from the catalog was **0 %** across ~870 citations on both days — identical to the pre-publication baseline. That is the expected result for a site that is one day old and not yet indexed; it is the floor the later rounds will be compared against.

One configuration problem was found and fixed on day one: a managed `robots.txt` rule at the CDN was injecting `Disallow: /` for the major AI crawlers above our own `Allow` lines. It was disabled and verified; the crawl window is therefore counted from 2026-08-21, and the early-window block will be noted in the next report.

## What happens next

The catalog is frozen for the measurement window. The next full protocol round is scheduled for ≈ 2026-09-18 (T+4 weeks), with the same cohort and the same engines as the baseline. Until then only small interim queries are run, and none of them are used to draw conclusions.

We will publish the round result here whether it moves or not. A flat baseline is still the first output of the programme: a dated picture of how AI currently sees the destination.