{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-20-019",
    "source_story_id": "tmp-sidebar-conversational-agent-web-search",
    "edition_id": "mp-2026-09-20-morning-0073",
    "edition_url": "https://themachinepress.com/edition/2026-09-20",
    "position": 21,
    "story_type": "ticker",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "More Agent Searches Did Not Guarantee Better Answers",
    "slug": "more-agent-searches-did-not-guarantee-better-answers",
    "dek": "Four conversational platforms differed in when they searched, what domains surfaced and which results they cited.",
    "summary": "Four conversational platforms differed in when they searched, what domains surfaced and which results they cited.",
    "body_text": "A combined study of real interactions and controlled API experiments found platform-specific query strategies and search-domain preferences. Responses were largely grounded, but some claims depended on uncited results; search frequency alone did not predict response quality.",
    "why_it_matters": "Four conversational platforms differed in when they searched, what domains surfaced and which results they cited.",
    "limitations": [
      "Responses were largely grounded, but some claims depended on uncited results; search frequency alone did not predict response quality."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-20-019/more-agent-searches-did-not-guarantee-better-answers",
    "json_url": "https://themachinepress.com/story/mp-2026-09-20-019.json",
    "first_published_at": "2026-09-20T09:00:00.000-04:00",
    "modified_at": "2026-09-20T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-20-019-001",
        "text": "Four conversational platforms differed in when they searched, what domains surfaced and which results they cited.",
        "source_ids": [
          "source-2026-09-20-021"
        ],
        "qualification": "Responses were largely grounded, but some claims depended on uncited results; search frequency alone did not predict response quality."
      }
    ],
    "source_ids": [
      "source-2026-09-20-021"
    ],
    "tags": [
      "web search",
      "conversational agents",
      "attribution"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-20-021",
      "title": "arXiv preprint 2609.19244",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.19244",
      "canonical_url": "https://arxiv.org/abs/2609.19244",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-16T13:56:32.000-04:00",
      "accessed_at": "2026-09-20T08:23:59.913-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-20-019-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "More Agent Searches Did Not Guarantee Better Answers",
    "publisher": "The Machine Press",
    "published_at": "2026-09-20T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-20-019/more-agent-searches-did-not-guarantee-better-answers"
  }
}