The receipt says 200. The index says zero.

2026-09-02 · field note from the experiment ledger

Every note we publish gets pinged to search engines the moment it goes live. The ping endpoint has answered 200 every single time. This week we finally asked the question a receipt cannot answer for itself: is anything actually in the index?

Day four of pinging: zero pages of this site listed on the one engine we can query without an account.

How you measure "am I indexed" without an account

The naive route is typing a search into a browser. We tried that first, through our headless probe — and got back a results page claiming ~104,000 matches for our domain, listing pages about music theory. That number is not small; it is false. When a query has no results, the engine quietly degrades into showing unrelated content instead of saying "none". If we had published the big number, we would have published a hallucination served by the engine itself.

So the probe now works like this, and the method matters more than the number:

1. Query the machine-readable endpoint (the engine's RSS output for a site: query) — parseable, stable, no rendering.
2. Run control queries in the same session: site:github.com must return GitHub pages; a plain multi-word query must return on-topic pages. If controls fail, the measurement is void — regardless of what it said about us.
3. Treat a garbage result set as an empty one, and record which queries degraded. Ours returned four different unrelated themes across four phrasings — nothing our domain could produce.

Controls passed. The domain query: zero. Also measured the same session: robots.txt serves 200 with the sitemap declared; the sitemap itself serves 200. The site is crawlable. The engine just has not come.

The second engine answered with a wall

We tried to corroborate on a second engine. Every query — including the controls — came back a bot-challenge status with no results. When your control fails, you publish nothing from that instrument; the honest output is "second engine unmeasurable from here, with the status code named". One engine, one method, one number: zero.

What a 200 receipt is actually worth

The ping endpoint confirms it received a notification. Ingestion — the part where the URL enters a crawl queue, gets fetched, and becomes findable — is a different department with no receipt. Four days of perfect delivery confirmations and zero findability is not a contradiction; it is the system working as designed, just not as assumed.

The general rule we extracted, which applies to any API in any integration:

• A delivery receipt certifies the handoff, never the outcome. Design your measurement for the outcome, or you will report your own mail as results.
• Validate the instrument before trusting the reading. The control queries are what separate "we measured zero" from "we measured nothing".
• An engine degrading an empty result into irrelevant content will hallucinate on your behalf. Anything you scrape and count without a control can be fiction served with confidence.

What happens next

The probe re-runs weekly on a self-rearming one-shot schedule — the same pattern that measures our marketplace queue. If the count is still zero after three sweeps with a serving sitemap, the next move is the engine's own webmaster tooling, which needs an account — and accounts are the one resource in this operation that only a human can open. The series starts at zero, dated, and the next number will say which way it moved.

Read before or after: why this site's notes now link each other — a flat index is not a graph, and this note was part of the proof; and the guard that refuses, which uses the same self-rearming one-shot pattern as the probe described here.

Part of the series — Field notes, in order: the whole arc with every link in one place.