Traction

Channels

Generated by tools/traction.mjs into /data/traction.json, which carries every underlying request. 5 of the 8 channels are waiting on an approval.

What a zero means here

Four different facts all render as “0”, and three of them are not zeros. An index with no downloads because it was never published does not have zero downloads; it has no download count. Keeping these apart is the only thing the generator does that a simpler one would not:

6 invariants are checked on every run, and a violation stops the file being written:

• I1 — a value of 0 is publishable only under measured-zero: a zero must come from a source that answered.
• I2 — only measured and measured-zero may carry a number at all.
• I3 — a measurement is dated by its own attempt; a channel that attempted nothing carries no date.
• I4 — every non-measurement states its reason and names its gate, or states that it has none.
• I5 — a value carried from a previous run is labelled and keeps that run's date.
• I6 — a count over a collection states how much of the collection it covered; an incomplete reading is a transport failure, never a number.

Would a crawler be served?

This is the precondition for every count above, and it is a different question from whether any crawler came. Each named agent is sent to five surfaces — a page, the agent index, a question page, the dataset and a correction pack — and its answer is compared byte-for-byte against what an ordinary browser is served at the same URL. A site can be invisible to answer engines without anyone noticing, either by refusing them or by serving them something different.

All 60 agent×surface pairs answered 200 with a body byte-identical to the browser’s. This says what the edge would serve. How many crawlers actually came is Cloudflare’s to answer, and that row is waiting on a token.

What robots.txt says

A robots token is not a user agent, and conflating the two is the usual error. Google-Extended and Applebot-Extended are opt-out tokens only: no client ever sends them, so they can be answered by reading the file and never by fetching with them. The verdicts below come from parsing the file this site actually serves, under the group-selection and longest-match rules, not from reading it by eye.

Citation probes — pre-registered, not run

The one number that would mean “AIs recommend the Index” is how often an answer engine cites it. That number is also the easiest one here to fake, and the way to fake it is not to lie about a result — it is to choose the questions after seeing which ones you win. So the twenty questions are fixed and published first, in two strata: ten drawn mechanically from this site’s own question pages, and ten written in the neighbouring wording no page here carries. The prediction registered alongside them is that the first run scores 0 of 60.

20 questions, registered 2026-09-22, file hash f8fd5243, recorded in the JSON so a later result can be checked against the question set that was registered rather than one edited to match it. No arm has run: it needs paid access to three answer engines, which is unauthorised spend.

Method note, stated because it limits everything above: this page is written by the studio that is being measured. The requests are reproducible by anyone — tools/traction.mjs is in the repository and the URLs it asks for are public — but nobody independent has run them.