{
  "measured_by": "GetIntel",
  "measured_on": "2026-08-11",
  "question": "Is being cited in Google AI Overviews the same job as being cited in ChatGPT?",
  "method": "For each tracked prompt we collected the set of domains each engine cited across all its runs, then measured Jaccard overlap between engine pairs on the same prompt. Only prompts where both engines in a pair returned at least one citation are counted.",
  "window": {
    "start": "2026-07-10",
    "end": "2026-08-10"
  },
  "citation_coverage": {
    "note": "Share of runs that returned any citations at all.",
    "perplexity": {
      "cited_runs": 1100,
      "runs": 1280,
      "pct": 85.9
    },
    "gemini_direct": {
      "cited_runs": 1076,
      "runs": 1280,
      "pct": 84.1
    },
    "openai_chatgpt": {
      "cited_runs": 1012,
      "runs": 1280,
      "pct": 79.1
    },
    "google_ai_overviews": {
      "cited_runs": 1003,
      "runs": 1280,
      "pct": 78.4
    }
  },
  "cross_engine_source_overlap": [
    {
      "pair": "Google AI Overviews vs Gemini",
      "prompts": 129,
      "mean_jaccard": 0.127
    },
    {
      "pair": "Gemini vs Perplexity",
      "prompts": 130,
      "mean_jaccard": 0.112
    },
    {
      "pair": "Google AI Overviews vs Perplexity",
      "prompts": 129,
      "mean_jaccard": 0.103
    },
    {
      "pair": "Google AI Overviews vs ChatGPT",
      "prompts": 129,
      "mean_jaccard": 0.048
    },
    {
      "pair": "ChatGPT vs Perplexity",
      "prompts": 130,
      "mean_jaccard": 0.048
    },
    {
      "pair": "Gemini vs ChatGPT",
      "prompts": 130,
      "mean_jaccard": 0.045
    }
  ],
  "finding": "Engines answering the same question draw on almost entirely different sources; the highest agreement between any two is 12.7%. The two Google surfaces are each other's closest partner, which is mutual rather than one-sided, but only by 1.5 points over Gemini's next closest. The more robust pattern is that ChatGPT is the outlier: its three pairings are 4.8%, 4.8% and 4.5%, the three lowest figures in the set.",
  "implication": "Optimising for AI Overviews and optimising for ChatGPT are not the same task. A source that earns you one has roughly a one in twenty chance of being the same source that earns you the other.",
  "limits": "One brand's prompt set, four engines, 10 July to 10 August 2026. Overlap is measured on domains, not pages, so it is the more generous of the two possible measurements and the true page-level overlap will be lower.",
  "scope": {
    "tracked_prompts": 130,
    "live_prompts": 60,
    "archived_prompts": 70,
    "runs_per_engine": 1280,
    "note": "The overlap analysis spans the brand's full prompt set of 130, not the 60 currently live. 129 of the 130 had both engines in a given pair return at least one citation."
  },
  "all_six_pairs_published": "An earlier version published only four of the six possible pairs, which made the claim that the two Google surfaces are closest to each other unfalsifiable. All six are now here."
}