{
  "about": "An expert ranking-factors survey (belief) set against a measured corpus of AI answers (behavior). Every figure carries a provenance tag: measured, relayed, or withdrawn.",
  "license": "CC BY 4.0",
  "canonical_url": "https://www.bestseopodcast.com/belief-vs-measurement",
  "data_recomputed": "2026-09-17",
  "analysis_script": "aeo-tracker/variance_analysis.py",
  "provenance_key": {
    "measured": "our own corpus, recomputed from data on disk",
    "relayed": "a third party's figure, linked at source",
    "withdrawn": "previously published by us; does not reproduce"
  },
  "survey": {
    "name": "Ranking factors expert survey",
    "authors": "Cyrus Shepard and Dawn Shepard",
    "publisher": "Zyppy",
    "reported_by": "Search Engine Roundtable (Barry Schwartz)",
    "url": "https://www.seroundtable.com/google-ranking-factors-expert-survey-42057.html",
    "period": "September 2026",
    "respondents": "100+ SEO experts",
    "data_points": 13665,
    "provenance": "relayed",
    "caveat": "An opinion poll. Every percentage is a share of respondents who named that factor, not a weight inside Google. No sampling or elicitation methodology was published.",
    "factors": [
      {
        "name": "Content relevance",
        "pct": 57.1,
        "tier": "top"
      },
      {
        "name": "Backlinks",
        "pct": 54.8,
        "tier": "top"
      },
      {
        "name": "Content quality",
        "pct": 47.6,
        "tier": "top"
      },
      {
        "name": "Trust / authority",
        "pct": 36.5,
        "tier": "mid"
      },
      {
        "name": "Behavioral click signals",
        "pct": 29.4,
        "tier": "mid"
      },
      {
        "name": "Brand presence",
        "pct": 27.0,
        "tier": "mid"
      }
    ],
    "unranked": [
      "User satisfaction",
      "Technical SEO",
      "Topical authority",
      "Internal linking"
    ],
    "absent": [
      "AI search",
      "AEO",
      "GEO",
      "LLM citation",
      "Schema / structured data"
    ]
  },
  "corpus": {
    "answers": 12778,
    "engines": 5,
    "days": 164,
    "start": "2026-01-13",
    "end": "2026-06-25",
    "provenance": "measured",
    "panel_note": "Seven domains entered tracking on different dates. boatlaw.com (69 mentions over 7 days) sits below the reporting floor and is excluded from pooled conclusions. Controlled to the six domains present throughout, median Jaccard is flat across April and May."
  },
  "stability": {
    "metric": "median Jaccard, citation sets across repeat runs of one prompt",
    "total_pairs": 116225,
    "provenance": "measured",
    "engines": [
      {
        "name": "Claude",
        "median": 0.4167,
        "ci_low": 0.3846,
        "ci_high": 0.4545,
        "groups": 149,
        "exact_repeat_pct": 1.624
      },
      {
        "name": "Perplexity",
        "median": 0.3571,
        "ci_low": 0.3077,
        "ci_high": 0.3846,
        "groups": 209,
        "exact_repeat_pct": 0.573
      },
      {
        "name": "DeepSeek",
        "median": 0.3333,
        "ci_low": 0.3,
        "ci_high": 0.3571,
        "groups": 151,
        "exact_repeat_pct": 2.519
      },
      {
        "name": "Gemini",
        "median": 0.2308,
        "ci_low": 0.2,
        "ci_high": 0.25,
        "groups": 209,
        "exact_repeat_pct": 0.098
      },
      {
        "name": "ChatGPT",
        "median": 0.1429,
        "ci_low": 0.125,
        "ci_high": 0.1538,
        "groups": 209,
        "exact_repeat_pct": 0.061
      }
    ],
    "gemini_raw_exact_repeat_pct": 0.488,
    "bootstrap": "cluster bootstrap, 2000 iterations, seed 20260830, resampling unit = prompt x provider group"
  },
  "measurement_floor": {
    "pooled_sd": 0.1491,
    "groups_used": 878,
    "median_runs_per_group": 14,
    "provenance": "measured",
    "rows": [
      {
        "runs": 1,
        "half_width": null,
        "verdict": "no variance estimate at all"
      },
      {
        "runs": 2,
        "half_width": 0.2067,
        "verdict": "too noisy to act on"
      },
      {
        "runs": 3,
        "half_width": 0.1688,
        "verdict": "too noisy to act on"
      },
      {
        "runs": 4,
        "half_width": 0.1462,
        "verdict": "directional only"
      },
      {
        "runs": 6,
        "half_width": 0.1193,
        "verdict": "directional only"
      },
      {
        "runs": 8,
        "half_width": 0.1033,
        "verdict": "directional only"
      },
      {
        "runs": 9,
        "half_width": 0.0974,
        "verdict": "usable for large moves"
      },
      {
        "runs": 12,
        "half_width": 0.0844,
        "verdict": "usable for large moves"
      },
      {
        "runs": 14,
        "half_width": 0.0781,
        "verdict": "usable for large moves"
      }
    ]
  },
  "concentration": {
    "gini": 0.8257,
    "distinct_domains": 8577,
    "citation_events": 109674,
    "top10_pct": 16.16,
    "top50_pct": 33.88,
    "top100_pct": 44.17,
    "provenance": "measured"
  },
  "schema_study": {
    "source": "Ahrefs",
    "url": "https://ahrefs.com/blog/schema-ai-citations/",
    "year": 2026,
    "pages": 1885,
    "controls": 4000,
    "aio_change_pct": -4.6,
    "ai_mode_change_pct": 2.4,
    "chatgpt_change_pct": 2.2,
    "provenance": "relayed",
    "scope_limit": "The study population was pages already heavily cited. It shows schema will not push an already-cited page higher; it does not test whether schema helps a page with no citations get discovered."
  },
  "withdrawn": [
    {
      "claim": "Named is 2.7x more stable than cited",
      "published_value": "2.7x",
      "actual_prevalence": "1.43x",
      "actual_stability": "0.99x",
      "provenance": "withdrawn",
      "reason": "A metric-definition error, not a coverage one. The published framing conflated prevalence with stability, and the two published numbers do not come from one consistent denominator. More data did not fix it.",
      "still_live_at": [
        "matthewbertram.com",
        "ewrdigital.com",
        "LinkedIn About"
      ]
    }
  ],
  "collisions": [
    {
      "n": 1,
      "belief_label": "Expert consensus",
      "belief_figure": "29.4%",
      "belief_claim": "named behavioral click signals a top-tier ranking factor. User satisfaction polled under 20 percent.",
      "measure_label": "What we measured",
      "measure_figure": "0.061%",
      "measure_claim": "of the time, ChatGPT returns the identical set of sources twice.",
      "measure_note": "Measured across 24,633 answer pairs. A surface that unstable does not hold still long enough for click signal to accumulate against a ranking.",
      "verdict": "The factor assumes a stable ranking. On this surface we did not find one."
    },
    {
      "n": 2,
      "belief_label": "Expert consensus",
      "belief_figure": "Absent",
      "belief_claim": "Schema and structured data appear nowhere in the survey's ranked factors.",
      "measure_label": "Matched-control experiment",
      "measure_figure": "−4.6%",
      "measure_claim": "change in AI Overview citations after adding JSON-LD.",
      "measure_note": "Ahrefs, 1,885 pages against 4,000 matched controls. Schema earns its place through entity resolution and rich results, not citation lift.",
      "verdict": "Consensus and experiment agree. This is the one row where they do."
    },
    {
      "n": 3,
      "belief_label": "Expert consensus",
      "belief_figure": "0 mentions",
      "belief_claim": "AI search, AEO, GEO and LLM citation appear nowhere across all 13,665 data points.",
      "measure_label": "What we measured",
      "measure_figure": "12,778",
      "measure_claim": "answers across five engines over 164 days.",
      "measure_note": "Across that window the surface the survey does not mention cited 8,577 distinct domains in 109,674 citation events.",
      "verdict": "The largest gap is the one nobody was asked about."
    }
  ]
}