Research

Audit a summary against its source

Find unsupported claims, changed meaning, and material omissions before a summary is shared

7 min setupTested with:ChatGPTReviewed: 2026-08-28

Add your context

Your text stays in this browser. AILesson Prompts does not send it to a model or server.

Complete sourceSummary to auditAudience and materiality

Your prompt

Unfilled fields remain visible as placeholders, so you can still copy and edit the prompt

Audit the summary against the complete source.

Source:
[source]

Summary:
[summary]

Audience, use, and materiality rules:
[purpose]

Break the summary into atomic claims. For each claim label it Supported, Partly supported, Unsupported, Contradicted, Overstated, or Unclear; cite the source locator and explain any difference in numbers, dates, scope, causality, obligation, certainty, or attribution. Then identify material source facts omitted from the summary, distinguishing required omissions from optional detail. Do not penalize harmless paraphrase. Return a severity-ranked issue table, a coverage assessment, unresolved source ambiguities, and a corrected summary that introduces no new claim.

Private by defaultPrompt assembly happens locally in your browser. Avoid placing confidential information into any AI service unless your organization allows it.

From input to outcome

A worked example

See how concrete context turns this recipe into a usable result

Actual input

Complete source: [Section: Results] In a voluntary survey of 84 of 230 employees, 62% of respondents preferred two office days per week; 21% preferred three days and 17% another arrangement. Response patterns may not represent non-respondents. [Section: Decision] Leadership will run a three-month trial of two anchor days starting 4 September. Teams may request exceptions. No permanent policy decision has been made. [Section: Review] Attendance, retention, and employee feedback will be reviewed in December.
Summary to audit: Most employees want two office days, so leadership has permanently adopted a two-day office policy from 4 September. The change will improve retention and has no exceptions.
Audience and materiality: Brief all 230 employees. Survey coverage, temporary status, exceptions, dates, and uncertainty are material.

Example output

High-severity issues: “Most employees” is overstated—the source says 62% of 84 voluntary respondents, not a majority of all 230 employees [Results]. “Permanently adopted” contradicts a three-month trial and the explicit statement that no permanent decision has been made [Decision]. “Will improve retention” is unsupported; retention is only a future review measure [Review]. “Has no exceptions” contradicts teams being allowed to request exceptions [Decision].

Supported elements: two office days and the 4 September start date are supported when framed as a trial [Decision]. Material omissions: voluntary response count, representativeness limitation, three-month duration, exception process, lack of permanent decision, and December review.

Corrected summary: Leadership will start a three-month trial of two anchor office days on 4 September, with teams able to request exceptions. In a voluntary survey, 62% of 84 respondents out of 230 employees preferred two days; responses may not represent non-respondents. No permanent policy decision has been made. Attendance, retention, and feedback will be reviewed in December.

Why this works

1
Atomic claim review reveals small qualifiers that whole-paragraph comparison misses.
2
A separate material-omission pass checks what the summary never mentioned.