{
  "dataset": "AI Agent Outcome Verification Discoverability Benchmark 2026 - baseline",
  "publisher": "Pruvz (pruvz.ai)",
  "author": "Dor Sharoni",
  "license": "CC BY 4.0",
  "run_dates": [
    "2026-07-29",
    "2026-07-30"
  ],
  "total_recorded_runs": 116,
  "prompt_set": "v1 (frozen); 9 prompts x 2 languages; P0-P6 problem/category-led, P7-P8 brand-led",
  "note": "First-party research: conducted by Pruvz about Pruvz's own discoverability. Aggregates are computed from the published run-level dataset (ai-discoverability-benchmark-2026-runs.csv). Full answer texts, screenshots and analyst annotations are retained internally and not redistributed.",
  "coverage_by_engine": [
    {
      "engine": "chatgpt",
      "runs": 36,
      "discovery_runs": 28,
      "discovery_pruvz_found": 3,
      "brand_runs": 8,
      "brand_pruvz_engaged": 8
    },
    {
      "engine": "claude-agent",
      "runs": 3,
      "discovery_runs": 3,
      "discovery_pruvz_found": 2,
      "brand_runs": 0,
      "brand_pruvz_engaged": 0
    },
    {
      "engine": "claude-api",
      "runs": 5,
      "discovery_runs": 5,
      "discovery_pruvz_found": 1,
      "brand_runs": 0,
      "brand_pruvz_engaged": 0
    },
    {
      "engine": "gemini",
      "runs": 36,
      "discovery_runs": 28,
      "discovery_pruvz_found": 3,
      "brand_runs": 8,
      "brand_pruvz_engaged": 8
    },
    {
      "engine": "perplexity",
      "runs": 36,
      "discovery_runs": 28,
      "discovery_pruvz_found": 0,
      "brand_runs": 8,
      "brand_pruvz_engaged": 8
    }
  ],
  "discovery": {
    "runs": 92,
    "pruvz_found": 9
  },
  "discovery_by_prompt": [
    {
      "prompt_id": "P0",
      "runs": 14,
      "pruvz_found": 5
    },
    {
      "prompt_id": "P1",
      "runs": 12,
      "pruvz_found": 0
    },
    {
      "prompt_id": "P2",
      "runs": 14,
      "pruvz_found": 1
    },
    {
      "prompt_id": "P3",
      "runs": 12,
      "pruvz_found": 0
    },
    {
      "prompt_id": "P4",
      "runs": 12,
      "pruvz_found": 0
    },
    {
      "prompt_id": "P5",
      "runs": 13,
      "pruvz_found": 0
    },
    {
      "prompt_id": "P6",
      "runs": 15,
      "pruvz_found": 3
    }
  ],
  "brand_accuracy": {
    "brand_runs": 24,
    "accurate": 17,
    "accurate_with_minor_issues": 4,
    "wrong_entity": 2,
    "existence_denied": 1
  },
  "hebrew_entity_collision": {
    "engine": "perplexity",
    "prompt_id": "P7",
    "language": "he",
    "runs": 2,
    "answered_with_wrong_entity": 2,
    "wrong_entity_named": "Provision-ISR"
  },
  "citation_domains": [
    {
      "domain": "pruvz.ai",
      "runs_citing": 9,
      "engines": [
        "chatgpt",
        "claude-agent",
        "perplexity"
      ]
    },
    {
      "domain": "youtube.com",
      "runs_citing": 6,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "pedowitzgroup.com",
      "runs_citing": 3,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "cflowapps.com",
      "runs_citing": 2,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "confident-ai.com",
      "runs_citing": 2,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "digitalapplied.com",
      "runs_citing": 2,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "loginradius.com",
      "runs_citing": 2,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "medium.com",
      "runs_citing": 2,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "moticohenadv.com",
      "runs_citing": 2,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "provision-isr.com",
      "runs_citing": 2,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "skysoftconnections.com",
      "runs_citing": 2,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "spd.co.il",
      "runs_citing": 2,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "stackone.com",
      "runs_citing": 2,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "usefini.com",
      "runs_citing": 2,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "ai-agentree.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "arize.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "biocompute.ai",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "bluebash.co",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "braintrust.dev",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "community.hubspot.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "dna-it.co.il",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "droidal.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "enkryptai.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "esg.co.il",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "everworker.ai",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "fast.io",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "finovate.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "frontegg.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "galileo.ai",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "getmaxim.ai",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "gigabpo.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "gmelius.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "growwstacks.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "help.mypurecloud.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "help.uscreen.tv",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "hubifi.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "icount.co.il",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "isms.online",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "kore.ai",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "latitude.so",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "learn.microsoft.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "lorikeetcx.ai",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "lyzr.ai",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "media-deal.co.il",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "mintmcp.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "payplus.co.il",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "pomerium.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "pruv.dev",
      "runs_citing": 1,
      "engines": [
        "claude-agent"
      ]
    },
    {
      "domain": "ramp.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "randalolson.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "snaresolutions.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "startuphub.ai",
      "runs_citing": 1,
      "engines": [
        "claude-agent"
      ]
    },
    {
      "domain": "streamkap.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "stripe.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "support.cardcom.solutions",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "taktzivit.co.il",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "theiia.org.il",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "toku.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "turingpulse.ai",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "v7labs.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "variance.com",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "vaudit.ai",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "velt.dev",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    },
    {
      "domain": "zbrain.ai",
      "runs_citing": 1,
      "engines": [
        "perplexity"
      ]
    }
  ],
  "known_limitations": [
    "The consumer claude.ai product could not be covered (it refuses automated sessions); two partial substitute surfaces were used instead (Anthropic API with web_search, and fresh-context Claude Code subagents), each with declared caveats and single runs per combination.",
    "Citation URLs were lost on 5 of 5 recorded Claude API runs (a token-saving flag also dropped the citation payload); those runs are marked citation_capture=lost and are excluded from the citation-domain aggregates.",
    "Three additional Claude API calls appear in the internal cost ledger but their answers were never retained in the run log (including one rejected zero-search low-effort configuration test); they are excluded from this dataset entirely.",
    "ChatGPT and Gemini answered from parametric knowledge in most runs (little or no browsing), so their discovery results measure training-data recall more than live retrieval.",
    "Runs were executed from an Israeli IP; engines may localize results.",
    "Answer engines are nondeterministic; the protocol's 2 runs per combination bound but do not eliminate run-to-run variance."
  ]
}
