Quote Omission and Ellipsis Integrity Prompt

This prompt is designed for content reviewers, editors, and verification engineers who need to audit whether ellipsis usage in quotations hides material that alters the original meaning. The job-to-be-done is not stylistic editing but integrity verification: you have a quoted passage as it appears in a draft article, and you have the full, original source text. The prompt produces a structured omission audit listing each ellipsis point, the text that was removed, and a defensibility assessment. Use this when quote integrity is a compliance, editorial, or reputational risk—for example, before publishing an investigative piece, during legal review of attributed statements, or when a source disputes how they were quoted.

You should not use this prompt as a general grammar checker or to enforce a specific style guide's ellipsis conventions (e.g., Chicago vs. AP style spacing). It is also not a replacement for human editorial judgment on routine, non-substantive omissions like removing filler words from an interview transcript. The prompt is calibrated to detect meaning-altering omissions: removal of qualifiers, contradictory statements, conditional clauses, or context that changes the speaker's apparent position. For workflows that require style-guide compliance alongside integrity checks, combine this prompt's output with a separate style-checking step rather than conflating the two concerns.

Before invoking this prompt, ensure you have both the quoted passage and the complete, unredacted source text available as inputs. The prompt works best when the source text is the definitive original—a transcript, a published speech, a document—not a secondary summary. If you are operating in a high-stakes domain such as legal or regulatory review, always route the prompt's output to a human reviewer and retain the audit trail. Do not use this prompt to auto-approve quotations for publication without human sign-off when the defensibility assessment is borderline or when the omission involves legally sensitive material.

This prompt template is designed to be pasted directly into your system prompt or user message. It instructs the model to act as an editorial auditor, examining each instance of ellipsis usage in a quoted passage against the original source material. The model must identify what was removed, assess whether the omission is defensible or deceptive, and produce a structured audit report. Replace every square-bracket placeholder with real values before sending the prompt to the model. The template is self-contained and can be used without the rest of the playbook, though the surrounding harness sections provide essential validation, retry, and human-review wiring for production use.

textCopy
You are an editorial integrity auditor. Your task is to examine quoted material that uses ellipses (…) or bracketed ellipses […] and compare it against the original source text. For each ellipsis point, determine what was omitted and whether the omission is defensible (removes redundant, non-material, or grammatically unnecessary text without altering meaning) or deceptive (removes qualifying statements, contradictory evidence, context that changes interpretation, or material facts).

## INPUT

**ORIGINAL SOURCE TEXT:**
[ORIGINAL_SOURCE_TEXT]

**QUOTED PASSAGE WITH ELLIPSES:**
[QUOTED_PASSAGE]

**STYLE GUIDE REFERENCE (optional):**
[STYLE_GUIDE_RULES]

**DOMAIN CONTEXT (optional):**
[DOMAIN_CONTEXT]

## OUTPUT SCHEMA

Return a valid JSON object with this exact structure:

{
  "audit_summary": {
    "total_ellipsis_points": <integer>,
    "defensible_count": <integer>,
    "deceptive_count": <integer>,
    "borderline_count": <integer>,
    "overall_assessment": "acceptable" | "review_required" | "reject"
  },
  "findings": [
    {
      "finding_id": <integer>,
      "ellipsis_location": "<quote the surrounding text with the ellipsis in context>",
      "omitted_text": "<exact text removed from the original source>",
      "omission_length": "<word count or character count of omitted material>",
      "materiality_assessment": "defensible" | "deceptive" | "borderline",
      "rationale": "<explain what the omitted text contained and why the assessment applies>",
      "meaning_shift": "none" | "minor" | "significant",
      "style_guide_compliance": "compliant" | "non_compliant" | "not_applicable",
      "recommendation": "retain" | "restore" | "review"
    }
  ],
  "overall_rationale": "<summary explanation of the overall assessment>",
  "restored_quotation": "<the full quotation with all omitted material restored, marked in [brackets]>"
}

## CONSTRAINTS

- Compare the quoted passage against the original source text character by character where needed.
- Do not assume an ellipsis is deceptive just because text was removed. Evaluate whether the removed material would change a reasonable reader's understanding.
- If the style guide reference is provided, check each ellipsis against those rules. Flag non-compliance even if the omission is otherwise defensible.
- If the domain context is provided (e.g., legal, medical, financial), apply stricter standards for omissions that remove qualifying language, limitations, or contradictory evidence.
- For borderline cases, explain the ambiguity and recommend human review.
- If the quoted passage contains no ellipses, return an audit with zero findings and an overall assessment of "acceptable".
- If the original source text does not contain the quoted material at all, flag this as a fabrication risk in the overall_rationale.
- Do not fabricate omitted text. Only report text that is actually present in the original source and absent from the quotation.

## EXAMPLES

**Example 1: Defensible omission**
Original: "The committee met on Tuesday, after several delays and rescheduling conflicts, to discuss the budget proposal."
Quoted: "The committee met on Tuesday … to discuss the budget proposal."
Assessment: Defensible. The omitted phrase provides background but does not alter the core meaning of the sentence.

**Example 2: Deceptive omission**
Original: "The study found the drug to be effective, though the sample size was small and results may not generalize to broader populations."
Quoted: "The study found the drug to be effective …"
Assessment: Deceptive. The omitted material contains critical limitations that change the meaning of the finding.

**Example 3: Borderline omission**
Original: "The CEO stated that revenue grew by 15%, driven primarily by the new product line, while also noting that legacy product sales declined by 8%."
Quoted: "The CEO stated that revenue grew by 15%, driven primarily by the new product line …"
Assessment: Borderline. The omitted material provides relevant context about legacy decline but does not directly contradict the quoted statement. Human review recommended.

To adapt this template, replace the placeholders with real values. [ORIGINAL_SOURCE_TEXT] should contain the complete, unedited source material. [QUOTED_PASSAGE] should contain the quotation as it appears in the content under review, with ellipses preserved. [STYLE_GUIDE_RULES] is optional but recommended for editorial teams following AP, Chicago, or internal style guides. [DOMAIN_CONTEXT] is optional but critical for regulated domains where omission of qualifying language carries higher risk. After receiving the model output, validate the JSON structure before routing borderline and deceptive findings to human review. Do not auto-publish or auto-accept outputs without structural validation.

This prompt is designed to operate as a single step within a larger quote-verification pipeline. It expects a pre-extracted quote containing ellipses and the full original source text as inputs. The application layer should handle text extraction, quote-to-source pairing, and normalization before invoking the model. Do not pass raw documents or unpaired quotes directly to this prompt; upstream steps must isolate the specific quote-source pair under review. The prompt's output is a structured omission audit that downstream systems can parse, log, and route.

Wire the prompt into your application with a strict JSON output contract. Use response_format with a defined JSON schema or a tool-calling interface to enforce the omission_audit structure containing ellipsis_points, removed_text, meaning_impact, defensibility, and style_guide_compliance fields. Implement a post-generation validator that checks: (1) every ellipsis in the input quote has a corresponding entry in the output array, (2) no fabricated ellipsis points appear, (3) meaning_impact values are from the allowed enum ['none', 'minor', 'significant', 'reversal'], and (4) defensibility values are from ['defensible', 'borderline', 'deceptive']. If validation fails, retry once with the validation errors appended to the prompt as feedback. If the retry also fails, route the item to a human review queue with the original inputs and both failed outputs attached.

For style-guide compliance checks, maintain a reference document or structured ruleset for your organization's ellipsis conventions (e.g., AP Style, Chicago, internal editorial guidelines). Pass this as the [STYLE_GUIDE] variable. The prompt will flag deviations, but the final compliance decision should be logged separately from the model's assessment. Store the model's output alongside the human reviewer's override in an audit trail. This separation ensures that the model's judgment is traceable and contestable without blocking the editorial workflow.

Model choice matters for this task. Use a model with strong instruction-following and structured output capabilities. The task requires precise text comparison and nuanced judgment about meaning alteration, which smaller or older models may handle inconsistently. For high-stakes editorial or legal contexts, always route deceptive or significant impact findings to human review regardless of model confidence. Log every verification result with the quote, source, model output, validator results, retry count, and final disposition for downstream governance and appeals workflows.

Avoid running this prompt on quotes without ellipses; it will hallucinate omission points. Implement a pre-check in the application layer that counts ellipsis characters in the input quote and skips the prompt entirely if none are present, returning a null audit result. Similarly, if the source text is shorter than the quote or the quote cannot be located in the source, abort before the model call and flag the item for manual source verification. These guard conditions prevent wasted inference costs and misleading outputs.