Audit a summary against its source

Author: AILesson7 min setupTested with:ChatGPTReviewed: 2026-08-28

Quick answer

Find unsupported claims, changed meaning, and material omissions before a summary is shared. Provide: Complete source, Summary to audit, Audience and materiality. Expected result: A claim-level fidelity audit, omission review, and corrected summary.

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.

Try in Playground

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.

Check the result

Was every summary claim assigned evidence and a fidelity label?
Were causality, certainty, scope, and numerical qualifiers preserved?
Does the corrected summary contain only source-supported claims?

Use it with confidence

Frequently asked questions

Practical answers about when to use this recipe, what to provide, and where human review still matters

What should I prepare before using “Audit a summary against its source”?

For “Audit a summary against its source,” prepare Complete source, Summary to audit, and Audience and materiality. Replace placeholders only with information you can verify. If a detail is unknown, preserve that uncertainty explicitly instead of asking the model to infer it.

When is the “Audit a summary against its source” result not ready to use?

The result is not ready if it does not yet deliver the stated outcome—A claim-level fidelity audit, omission review, and corrected summary—from the supplied evidence, or if it relies on unresolved assumptions, missing approvals, or invented details. Use the checks as release gates: revise the source inputs or assign a named, authorized reviewer instead of polishing an unsupported output.

Which AI tools have recorded tests for “Audit a summary against its source”?

The published test record for “Audit a summary against its source” lists ChatGPT as of 2026-08-28. This confirms recorded runs, not guaranteed compatibility or identical results in later product versions. For another tool or version, keep every constraint visible and repeat the result checks before use.

More ways to explore

Where this recipe fits

Text prompts Document reading Document summarization Document analysis Feedback & review For writers

Keep the work moving

Text prompt

Read clearly

Document reading 9 min setup

Organize multiple sources into an evidence table

Extract comparable claims, observations, methods, and limits from several documents while preserving source lineage

Open recipe

Text prompt

Read clearly

Document reading 6 min setup

Turn a long document into an actionable summary

Condense a document around decisions and next steps while preserving evidence and uncertainty

Open recipe

Text prompt

Read clearly

Document reading 6 min setup

Generate a source-grounded FAQ

Turn source material into useful questions and traceable answers without filling evidence gaps

Open recipe

AILesson · Recommended courses

Your next step: put AI to work

Go from understanding AI to getting things done. Practice prompting, reviewing, and improving results through interactive lessons for work and everyday life.

Explore all courses