SERP Research Pack for AI Agents

SERP Research Pack for AI Agents turns a natural-language search query into a structured research brief that another AI agent can consume directly. It returns deduped source records, snippets, inspected-page summaries, evidence sentences, relevance scores, entities, research risks, and suggested next searches.

Use it as a bounded discovery step before an agent writes a market brief, selects tools, plans a scrape, prepares an SEO outline, researches competitors, or gathers cited source leads for a human analyst.

Use This Actor When

• You need a compact SERP research brief instead of raw browser output.
• You want source URLs, snippets, evidence sentences, and relevance signals in one JSON record.
• You are building an AI-agent workflow that needs search context before deeper crawling or summarisation.
• You need quick market, product, competitor, topic, or tool-discovery research.
• You want next-search suggestions when the first result set is weak or incomplete.

Do Not Use This Actor When

• You need guaranteed Google ranking data, rank tracking, or paid keyword volume.
• You need exhaustive coverage across every search engine or country.
• You need facts from paywalled, login-only, or heavily blocked pages.
• You need legally or commercially verified facts without human review.

Input

{
  "query": "AI agents Apify actors MCP",
  "market": "UK",
  "maxResults": 8,
  "maxPagesToInspect": 5
}

Input Fields

• query: Required. Natural-language research query.
• market: Optional. Market, country, language, or location context appended to the query.
• maxResults: Optional. Number of search results to return, from 1 to 20.
• maxPagesToInspect: Optional. Number of result pages to fetch for summaries and evidence, from 0 to 10. Use 0 for SERP-only output.

Output

The Actor writes one JSON research-pack record to the default dataset.

Key fields:

• query: The final query used.
• summary: Counts, average relevance, source-type mix, and dataQuality.
• sourceRecords: Inspected sources with rank, title, url, domain, snippet, status, sourceType, pageSummary, evidence, relevanceScore, and fetch status.
• uninspectedResults: Extra deduped SERP results that were not fetched.
• topFindings: Evidence-backed finding seeds with source URLs and confidence labels.
• entities: Extracted terms useful for follow-up research.
• risks: Data-quality and interpretation warnings.
• nextSearches: Follow-up searches an agent can run.
• downstreamWorkflowHints: Suggested next actions for agent pipelines.
• limitations: Explicit boundaries of the result.

Example shape:

{
  "query": "AI agents Apify actors MCP UK",
  "summary": {
    "inspectedSources": 5,
    "returnedResults": 8,
    "averageRelevance": 0.58,
    "dataQuality": "medium",
    "sourceTypeCounts": {
      "technical_reference": 2,
      "guide_or_article": 2,
      "general_source": 1
    }
  },
  "sourceRecords": [
    {
      "rank": 1,
      "title": "Example source",
      "url": "https://example.com/research",
      "domain": "example.com",
      "snippet": "Search-result snippet...",
      "status": 200,
      "sourceType": "guide_or_article",
      "pageSummary": "Short page summary...",
      "evidence": ["Relevant sentence extracted from the page."],
      "relevanceScore": 0.67,
      "inspected": true,
      "error": ""
    }
  ],
  "topFindings": [
    {
      "finding": "Relevant sentence extracted from the page.",
      "sourceUrl": "https://example.com/research",
      "sourceTitle": "Example source",
      "confidence": "medium"
    }
  ],
  "entities": ["agents", "apify", "mcp"],
  "risks": ["Treat findings as research leads with source URLs, not verified final facts"],
  "nextSearches": ["AI agents Apify actors MCP UK official documentation"]
}

Common Downstream Workflows

1. Run this Actor with a market or competitor query.
2. Pass high-relevance sourceRecords to a crawler, summariser, SEO writer, or research agent.
3. Use topFindings as cited context seeds in a memo or brief.
4. Run nextSearches when dataQuality is low or the agent needs corroboration.

What This Actor Cannot Guarantee

Search engines can personalise, block, or reshape result pages. This Actor uses bounded public web search and lightweight page inspection, so it should be treated as a fast discovery layer rather than a final authority.

Pricing And Runtime

The Actor is designed for short agent loops. Keep maxResults and maxPagesToInspect low for fast runs, then increase them only when a workflow needs broader coverage.