Summarization Prompts with Source Boundaries

This reviewed prompt collection is for knowledge workers, students, editors, and researchers. It assumes the user has permission to process the supplied material and can verify the result against the source.

When to use these prompts#

Use this page when you need task-specific prompts for executive, extractive, abstractive, meeting, and study summaries with source-boundary, omission, length, and uncertainty controls. Replace every bracketed variable with verified, non-sensitive information before using a template in your own AI product.

Inputs to prepare#

• complete source material and document boundaries
• summary audience and decision context
• summary type: executive, extractive, study, or comparative
• required facts, sections, or quotations
• target length and output format
• terms that must remain exact
• rules for uncertainty and missing information
• sensitive information to remove

Draft prompt templates#

1. Executive decision summary#

Purpose: Condense a document around the decision a reader must make.

Summarize [SOURCE] for [AUDIENCE] who must decide [DECISION]. Use only information explicitly present in the source. Return: a five-sentence executive summary, key evidence, risks, unresolved questions, and recommended follow-up questions. Quote exact numbers and dates with their source section. Write “not stated” when the document does not provide an answer.

Before use, remove irrelevant fields, replace every placeholder, and verify that the requested judgment can be supported by the material you supplied.

2. Study notes with traceability#

Purpose: Produce structured learning notes without adding outside knowledge.

Create study notes from [SOURCE] for a learner at [LEVEL]. Organize them into definitions, main claims, evidence, examples, and likely misconceptions. For every item, include the page, section, or heading where it appears. Do not add background knowledge. End with five self-test questions whose answers are directly supported by the source.

Before use, remove irrelevant fields, replace every placeholder, and verify that the requested judgment can be supported by the material you supplied.

3. Multi-document comparison#

Purpose: Compare supplied documents while preserving disagreements.

Compare [DOCUMENT A], [DOCUMENT B], and [DOCUMENT C] on [QUESTIONS]. Create a matrix showing each document’s position, evidence, scope, and limitations. Preserve disagreements and do not merge them into a false consensus. Mark absent information as “not addressed.” Finish with a neutral synthesis and a list of claims that require external verification.

Before use, remove irrelevant fields, replace every placeholder, and verify that the requested judgment can be supported by the material you supplied.

Why the structure matters#

A summary can be fluent yet unsafe if it fills gaps, drops qualifications, or hides disagreement. These prompts define the source boundary, audience, omission rules, and traceability requirements before compression begins.

Weak prompt and what it misses#

Summarize this and include anything important I should know.

“Anything important” invites the model to use an undefined importance standard or add background knowledge. There is no audience, source boundary, length, traceability rule, or instruction for missing information.

Human review checklist#

• Verify every number, date, quotation, and named entity against the source.
• Check that important limitations and dissenting views were not omitted.
• Confirm that “not stated” items were not converted into assumptions.
• Review whether the compression level fits the reader’s decision.
• Do not upload confidential or licensed material without authorization.

Testing and maintenance#

1. Use original, non-sensitive sample values when testing a prompt.
2. Retest examples after material changes to the prompt, page, or AI product.
3. Check factual preservation, requested format, missing-data behavior, and task-specific risks.
4. Update the visible test date when a material retest is completed.
5. Keep the site operator responsible for the final human review.

A dated operator attestation is sufficient for the site's internal publication check; full chat transcripts are not retained or published.