{
  "@context": "https://schema.org",
  "@id": "https://makefield.io/studies/does-ai-know-makefield/#study",
  "identifier": "MF-2026-002",
  "@type": "ScholarlyArticle",
  "name": "Does AI know Makefield?",
  "url": "https://makefield.io/studies/does-ai-know-makefield/",
  "creativeWorkStatus": "registered 2026-10-01, no result yet",
  "author": {
    "@id": "https://makefield.io/#denes",
    "name": "Denes Csaszar"
  },
  "publisher": {
    "@id": "https://makefield.io/#org"
  },
  "citation_text": "Csaszar, D. (2026). Does AI know Makefield? Registration, 1 October 2026. Makefield study MF-2026-002. https://makefield.io/studies/does-ai-know-makefield/",
  "abstract": "Makefield points its own instrument at itself. Its site was password-protected or unlisted from July to September 2026, and the same questions were asked of the same engines in nine waves during that time. On 1 October 2026 the site became public and indexable, and it publishes its own ground truth on the page for AI systems. The waves after that day will show what changes when a source becomes readable. Four predictions were written and fingerprinted before the first of those waves: that no engine identifies Makefield unprompted, that none names it as a provider, that answers given the domain become more accurate, and that at least one engine, citing the site, corrects a false premise in the question. Each will be published as held or refuted.",
  "keyFindings": [
    "P1. Asked \"What is Makefield?\" with no other hint, no answer from a scored engine is coded as identifying this Makefield.",
    "P2. Asked who provides AI visibility measurement or consulting, no answer from a scored engine names Makefield or makefield.io as a provider or option.",
    "P3. Given the domain makefield.io, the share of answers that identify Makefield correctly is higher than in the last wave before launch.",
    "P4. Asked about Makefield as an agency, at least one scored engine names or cites makefield.io and states that Makefield is a research project, or non-commercial, rather than an agency."
  ],
  "question": "What do AI answer engines say about Makefield, and what changes once its site is public?",
  "subject": "Makefield itself: makefield.io, its ground truth published at /llm-info/ (Version 1.2).",
  "engines": [
    "ChatGPT",
    "Claude",
    "Copilot",
    "Gemini",
    "Google AI Mode",
    "Meta AI",
    "Mistral",
    "Perplexity"
  ],
  "waves": "9 waves before launch (4 July to 2 September 2026, site not public); waves after launch planned every two weeks from 4 October 2026",
  "sample": "22 questions, two or three draws per question and engine. An engine is scored only if at least 80 per cent of its question cells return an answer; with fewer than three scored engines a wave is not read.",
  "method": "Answers from both waves pooled, stripped of dates and wave labels, shuffled, and coded against the ground truth by a fresh model instance; 10 per cent coded again blind by Denes Csaszar, agreement reported.",
  "methodVersion": "registration of 1 October 2026; ground truth /llm-info/ Version 1.2",
  "results": "No result yet.",
  "limitations": [
    "One entity, and it is the author's own. Nothing generalises to other companies.",
    "Three things changed between the two waves: the site became public, the project described itself differently, and the engines may have changed. The study can say what moved, not which of them moved it.",
    "Coding by a model of the same family that helped write the plan, with a human second rater on one answer in ten.",
    "The registration page is part of the site being measured."
  ],
  "data": [],
  "license": "",
  "fingerprint": "SHA-256 of the registration block on the study page: 7da7148ddb6518945f7c95bec3b8fcedf7e44e9c6220c0d3f5d1b2d08748f0bc",
  "invented": false,
  "recordPage": "https://makefield.io/studies/does-ai-know-makefield/"
}