Citation Remediation Payload Generator

Turn supplied AI citation evidence and HTML snapshots into deterministic JSON instructions for a downstream HTML consumer.

This Actor helps prepare conservative changes when supplied evidence cites a competitor page and omits your target page. It uses facts already present in your target snapshot. It does not predict or guarantee that an AI system will cite your page.

What it does

• Processes one target page, one cited competitor page and one query per run.
• Checks the citation gap in the URLs supplied by the caller.
• Proposes a query heading and grounded excerpts when applicable.
• Proposes minimal WebPage JSON-LD when missing.
• Returns snapshot hashes, change preconditions and typed acceptance tests.

Input example

{
  "targetUrl": "https://example.com/product",
  "competitorUrl": "https://competitor.example/product",
  "query": "Does Acme export CSV?",
  "evidence": {
    "answer": "The competitor offers CSV exports.",
    "citedUrls": ["https://competitor.example/product"]
  },
  "targetHtml": "<!doctype html><html><body><h1>Acme</h1><p>Acme exports reports in CSV format.</p></body></html>",
  "competitorHtml": "<html><body><h1>CSV exports</h1></body></html>",
  "facts": [{
    "text": "Acme exports reports in CSV format.",
    "sourceUrl": "https://example.com/product"
  }]
}

Supply relevant, truthful facts and current snapshots. Unknown input fields are rejected. Citation evidence is caller-supplied and is not independently authenticated. Competitor HTML is used only for heading comparison.

Output

The Actor writes a structured JSON result to the run's default dataset. It includes schemaVersion, inputHash, status, provenance and snapshot hashes, diagnosis, changes, acceptanceTests and warnings. Heading comparison may also be present.

The example above produces a ready payload with an HTML section, minimal WebPage JSON-LD and acceptance tests. Other inputs may produce fewer changes, no_gap or no_action.

Supported changes are append_html and append_json_ld. Tests are typed instructions: selector_count, text_contains and json_ld_webpage_url. They are not arbitrary executable code.

Applying a payload

The Actor returns instructions; it does not edit your website. Before applying any change, your consumer must verify all snapshot hashes and selector preconditions against the original snapshot. Track applied change IDs, support rollback and refuse already-applied changes. Use an appropriate CMS adapter for production writes.

Limitations

The Actor does not fetch live pages, render JavaScript, collect citation evidence, perform semantic gap analysis, predict rankings, discover internal links, generate FAQ schema, write to a CMS or process bulk jobs. Excerpt matching verifies occurrence rather than factual truth or relevance. Minimal WebPage JSON-LD does not establish Google rich-results eligibility. A repeated query heading can suppress a content change even if existing content is inadequate.

Pricing

• Remediation payload: USD 0.02 for one result with status ready, regardless of the number of proposed changes.
• Actor start: USD 0.00005 per run with the supported memory allocation of 256–512 MB.
• Platform usage: included in the event prices.

A run producing one ready payload costs USD 0.02005. Results with status no_gap or no_action do not incur the remediation-payload charge; the startup charge may still apply. Invalid or failed runs may also incur a startup charge. There is no additional automatic dataset-item fee.

Set a maximum cost per run to control spending. If the remaining budget cannot cover a ready payload, it will not be delivered.