Source Quote Verifier

What does Source Quote Verifier do?

Source Quote Verifier checks whether a public web source actually contains claimed text and returns the matching passage, surrounding context, character offsets, retrieval metadata, and a content hash. It is a small deterministic API primitive for citation checks, evidence pipelines, research agents, and fact-checking workflows. It does not use an LLM and does not claim that a passage semantically supports an argument.

The Actor retrieves one public HTTPS document with bounded standard HTTP, extracts readable text from HTML or textual formats, and evaluates the supplied quote using an explicit algorithm. Responses are capped at 1 MiB after decompression. You can invoke it from the Apify API, scheduled runs, integrations, or Apify MCP and monitor every run through the Apify Console.

Why use Source Quote Verifier?

• Verify that a quotation is present before an agent cites a source.
• Replace ad-hoc page retrieval, HTML cleanup, Unicode normalization, and passage-location code with one stable JSON contract.
• Receive SHA-256 content provenance, redirect history, retrieval time, match type, score, and algorithm name.
• Keep matching deterministic and reproducible without runtime AI or a paid data supplier.
• Block localhost, private networks, link-local addresses, metadata services, unsafe redirects, oversized responses, and slow requests by default.

Example agent queries include “Does this article contain this quote?”, “Find this claimed text in the source,” “Verify this quotation and return its context,” and “Check whether this URL contains the cited sentence.”

How to verify a quote on a web page

1. Open the Actor's Input tab.
2. Enter the public HTTPS source URL.
3. Paste the claimed text.
4. Choose normalized for most citations, exact for byte-like literal checking, or fuzzy only when small textual differences are expected.
5. Start the Actor and inspect the single dataset result.

Always ensure your retrieval and use of the source complies with its terms, copyright rules, and applicable law.

Input

See the Input tab for the full schema. The contract deliberately has only three fields:

{
  "url": "https://example.com/article",
  "text": "The product launched in June 2026",
  "matching": "normalized"
}

normalized is the default. It applies Unicode NFKC normalization, case folding, whitespace collapsing, and equivalent quote/dash punctuation. exact is case- and whitespace-sensitive. fuzzy first attempts normalized exact matching, then uses a fixed token-overlap and character-trigram algorithm with a documented 0.82 threshold. There is no semantic inference.

Output

The Actor writes exactly one versioned JSON envelope. You can download the dataset in various formats such as JSON, HTML, CSV, or Excel.

{
  "contractVersion": "1.0",
  "ok": true,
  "source": {
    "finalUrl": "https://example.com/article",
    "httpStatus": 200,
    "contentHash": "sha256:...",
    "retrievedAt": "2026-08-09T12:00:00.000Z"
  },
  "verification": {
    "match": true,
    "matchType": "normalized_exact",
    "algorithm": "unicode-nfkc-casefold-whitespace-punctuation-v1",
    "score": 1,
    "threshold": 1,
    "occurrenceCount": 1,
    "passage": "The product launched in June 2026",
    "startOffset": 418,
    "endOffset": 458
  },
  "warnings": []
}

Field	Type	Meaning
ok	boolean	Whether retrieval and a trustworthy textual determination completed
source	object	Final URL, status, content type, byte count, hash, timestamp, and redirects
extraction	object	Detected format, extraction method, title, and readable character count
verification	object	Match decision, type, algorithm, score, passage, context, and offsets
error	object	Stable machine-readable free failure when ok is false

How much does Source Quote Verifier cost?

The initial pricing proposal is one quote-verification pay-per-event charge of approximately $0.03 for a completed determination. Network, timeout, unsupported-content, resource-budget, and internal failures are not charged. Final Store pricing remains subject to a private Apify benchmark and margin gate before publication.

Accuracy and advanced options

Use normalized unless the application explicitly needs literal or fuzzy behavior. A fuzzy result means only that a deterministic textual similarity threshold was met; inspect score, algorithm, and passage. Client-rendered text, images, audio, video, PDFs, authenticated pages, and content loaded only after JavaScript execution are intentionally unsupported in v1. Query values are used for retrieval but redacted from output and logs.

FAQ, responsible use, and support

Does a match prove the source supports my claim?

No. It proves that extracted source text matched the supplied text under the named algorithm. Interpretation and semantic entailment are outside this Actor's contract.

Why can visible browser text be missing?

The Actor uses bounded HTTP and does not run page JavaScript. Use a browser-assisted retrieval tool when text is rendered only on the client.

The Actor reads public source text and does not intentionally collect private user data. Public pages can still contain personal or copyrighted information. Use results only with a lawful purpose and consult qualified counsel when unsure. Use the Actor's Issues tab for bug reports and the API tab for programmatic invocation.