{
  "measured_by": "GetIntel",
  "measured_on": "2026-08-11",
  "question": "Do the posts that earn citations use the same words as the buyer questions?",
  "method": "For each post title we computed Jaccard overlap of its content words against each of the 60 tracked prompts and kept the best match. Stopwords removed. Titles only, not full body text.",
  "scope": {
    "posts_compared": 68,
    "cited": 9,
    "uncited": 59,
    "note": "68 of the 83 in-window posts had a reliably extractable title; 9 of the 12 cited posts parsed. The 15 that did not are legacy entries whose titles could not be pulled cleanly."
  },
  "results": {
    "cited": {
      "n": 9,
      "mean": 0.203,
      "median": 0.182
    },
    "uncited": {
      "n": 59,
      "mean": 0.109,
      "median": 0.083
    },
    "cited_above_uncited_median": 8
  },
  "finding": "Cited posts' titles overlap the nearest tracked prompt about twice as much as uncited ones, median 0.182 against 0.083, with 8 of 9 above the uncited median.",
  "the_confound": "Both the posts and the prompts derive from the same product and ICP framing, so some vocabulary overlap is structural rather than causal. This shows an association between phrasing a page like a tracked question and being cited on it; it does not show that rewording a title would earn a citation.",
  "what_we_could_not_measure": "We attempted to compare word counts of cited against uncited posts and abandoned it: legacy posts stored in blogData.ts could not be parsed reliably and only one of the twelve cited posts extracted. No length comparison is published because we do not have one.",
  "sample_size_caveat": "The cited group is n=9. A single outlier title would move the median materially, so treat the gap as directional rather than precise."
}