I've read the full EY report. Now I'll do two things in parallel: load the canvas skill (since the article is a substantial document deliverable) and run a stylometric analysis of the report's prose to ground my findings in measurable evidence.

The stylometric analysis gives me concrete evidence. Now I'll write the article as a Canvas document.

I've analyzed the report and written up the findings in a article canvas: "The Machine in the Room: How Much of EY's AI Survey Was Written by AI?"

My verdict: plausibly 70–90% of the narrative prose (~2,200 words) was drafted or heavily rewritten by an LLM, with humans supplying the survey data, structure, methodology, and sign-off.

Key evidence from the stylometric analysis:

• Triad density: 21 "X, Y and Z" constructions in 99 sentences (~1 in 5) — the signature LLM cadence
• Refrain repetition: "confidence gap" ×5, "keep pace" ×5, "as intended" ×3 — a report generated from a thesis prompt rather than written toward one
• Near-verbatim recycling: the same 20+ word sentences about the 72% regulation concern and the 36% incident rate appear on both pages 10 and 12; "data loss, financial damage, operational disruption and brand damage" appears word-for-word twice
• Templated structure: three sections end with an identical "ORGANIZATIONS SHOULD ASK:" device
• Tells of absence: no contractions, no digressions, no em-dashes — frictionless competence that human writers under deadline rarely achieve
• Counter-evidence: the methodology math checks out (±7pp MOE for n=202 is arithmetically correct), and one sentence ("Just as assurance professionals feel they've caught up...") carries a genuinely human, insider voice

The juiciest finding is the irony on page 18: EY explicitly certifies that its research was "conducted by human researchers... screened for bots and AI agents" — yet says nothing about the prose. A report about the gap between governance-on-paper and governance-in-practice turns out to be its own case study.

The article includes full evidence, counter-evidence, the estimate with its limits, industry implications, and a self-aware closing note about an AI judging AI authorship.

Here's the cartoon to accompany the article — a giant robot typing the entire report while the humans concentrate on stamping the last page "100% HUMAN-MADE":

It captures the article's core irony: EY certifies the humans behind its data (the "100% HUMAN-MADE" stamp on the final page) while the machine visibly writes everything else — which is exactly the disclosure asymmetry on page 18 of their methodology.

Want me to tweak it (different style, other visual gag — e.g., a robot getting a performance review from the report it wrote), or would you like it embedded at the top of the article canvas as a header illustration?