Uravation AI Checker logs 121,608 judgments; business docs up 2.2x

Uravation, a Tokyo-based company that provides generative-AI training and consulting, first published a survey of about 73,000 judgments on September 15. This follow-up covers 121,608 completed judgments from July 18 to October 16, 2026, excluding 8,997 that failed due to usage limits or errors, and tracks only metadata such as time, character count, language and document type — not the judgments themselves.

The weekly mix reflects the school calendar. Reports and assignments were 71.2% of judgments in the week of August 3, but in the weeks of August 31 and September 7, entry sheets edged ahead (36.2% vs 35.5%, then 36.1% vs 32.4%). From the week of September 28, reports rose back above half at 52.7%, continuing in the 50% range through October, around when university second-semester classes begin. September's daily average fell to 854, a dip Uravation attributes partly to temporary malfunctions on September 10, 15 and 16, before recovering to 1,657 in October's first half.

Uravation also updated the AI model behind the tool's judgment and classification on September 16, so it did not report the distribution of AI-rate results this time; it says it checked that document-type classification did not differ greatly between the old and new models on the switch date. In its guidance, the company advises against using judgment results alone for evaluation or pass/fail decisions, and recommends deciding in advance which parts of a document may use AI and which should be written in one's own words.