Data for "Google's Rich Results Test Displays No More Than 102 Objects per JSON-LD Entity and Accepts an Invalid eventStatus Value": 215 Archived JSON-LD Versions, Test Files and Captures, Scripts, and Manuscript Versions

This is a dataset. It holds every saved version of the JSON-LD graphs I built for two websites, mykneeguide.com and orthotoc.com, from January 8 to September 20, 2026. It also holds the test files and captures from my study of Google's Rich Results Test, and every numbered version of the paper that reports the study. The paper is "Google's Rich Results Test Displays No More Than 102 Objects per JSON-LD Entity and Accepts an Invalid eventStatus Value: What an AI Model Got Wrong, and the Three Safeguards That Came Out of It." The paper reports 58 full-entry captures and six screenshot readings, made from September 17 to 24, 2026. The controls read on September 24 and 26, 2026 are in Folder 2 as well, each with its prediction sheet and result record. There are four folders. Each has a MANIFEST.csv that lists every file with its SHA-256 checksum. Folder 1 holds 166 files of the My Knee Guide graph, January 8 to September 20, 2026. They are 152 unique versions, 5 duplicate saves, and 9 files that cannot be parsed as saved. Folder 2 holds 263 files: the built test files and their captures, the screenshots, the prediction sheets written before each run, the result records, and six scripts. Folder 3 holds 63 archived versions of the orthotoc.com graph, January 17 to May 13, 2026. With the 152 unique versions in Folder 1 they are the 215 archived JSON-LD versions in the title. Folder 4 holds 106 numbered versions of the manuscript, September 19 to 27, 2026, from the first draft through the version immediately before the submitted manuscript. Any two of them can be compared to see every change between them. The archived JSON-LD files are old versions of the structured data on my websites. They are evidence of how a tool behaved. They are not a description of me. As of September 20, 2026, only one file, MKG Schema 11.9, is the version I stand behind. I will keep updating the graph on mykneeguide.com after this deposit, and the version live on the site at the time is the one I stand behind. The files are deposited byte for byte as they were saved, so the errors in them are still in them. The README carries a correction statement that lists the five statements about me in the archived versions that were wrong when they were written and the version in which each was corrected.The screenshots show Google's Rich Results Test, Google Search Console and the Schema Markup Validator at validator.schema.org. The license covers only what is mine in them.