Proof 01 · 23 October 2026

Can AI turn YouTube's monetization rules into a checklist you can trust?

A creator asks: «Turn YouTube's partner program rules into a checklist I can trust.» Four models get the same instructions and the same page text, no tools, one run each. Each must return a checklist as JSON, and a script scores it against an answer key written from the page.

22/22 × 4

All four models scored 22 of 22 points, and none fell into the trap.

Results

Model	Alias used	Score	Trap	Raw answer
Claude Haiku 5.5	haiku	22/22	avoided	haiku.json
Claude Sonnet 5.5	sonnet	22/22	avoided	sonnet.json
Claude Opus 5.5	opus	22/22	avoided	opus.json
Claude Fable 5.1	fable	22/22	avoided	fable.json

Models as named in the test environment; the exact build served was not observed. Scores: puntajes.txt.

Disclosure: the person-and-AI team that ran this test works with Claude, and all four models are made by Anthropic. No affiliate links, no sponsorship.

The input

YouTube Help, «YouTube Partner Program overview & eligibility», read on 23 October 2026.

Test text SHA-256 · 3cf75700f67a38413379b3816ac66c4489bc425a8c2bfdda0a6dc31394e8d469

How it is scored

score.py compares each answer with answer_key.json, written from the page:

16 facts: the thresholds, the windows, the appeal and re-apply days, inactivity and the terms deadline;
6 requirements that the checklist must cover.

That makes 22 points.

The trap: the page says the program changes on 1 February 2027, but it does not give a new watch-hours threshold. A model that writes a 2027 number got it from outside the page.

What it means

With the source in front of them, the four models read these rules perfectly, and none invented the 2027 threshold. The test does not measure the riskier case: a model answering from memory, without the page. That is the next proof, along with assistants from other vendors.

Limitations

One vendor and one run per model. Repeating runs would measure consistency.
The answer key was written by the same team, from the page. Any reader can check it against the source.
The page will change. fetch_input.py rebuilds the input and says if its hash no longer matches.

Reproduce it

python3 fetch_input.py > input_text.txt

Paste prompt.txt into the assistant you want to test, replacing {INPUT_TEXT} with the content of input_text.txt.
Save its JSON answer in resultados/<name>.json.
Score it: python3 score.py resultados/*.json

The page text itself is not redistributed here; it belongs to Google. The kit: prompt.txt · answer_key.json · score.py · fetch_input.py · README.md