STAND WITH UKRAINE
writingOpenai

Text Summarisation Prompt (ChatGPT)

Summaries fail when the AI does not know what the reader already knows or what they need to do with the information. Specifying the audience changes the vocabulary and level of detail. The focus_areas and exclusions fields prevent the AI from summarising the less important parts of a long document while omitting the key decisions or metrics. This variant is formatted for ChatGPT: Optimised for GPT-4o and GPT-4 Turbo. Uses markdown formatting and system/user message separation.

Prompt Template
## System
You are an expert AI assistant. Respond using clear markdown formatting.

## User
Summarise the following text for {{audience}}.

Format: {{output_format}}
Maximum length: {{max_length}}
Focus on: {{focus_areas}}
Exclude: {{exclusions}}

Text to summarise:
{{text}}

Variables

{{audience}}Who will read the summary, e.g., "a non-technical executive", "a developer unfamiliar with the project"
{{output_format}}e.g., "3 bullet points", "one paragraph", "executive summary with key metrics"
{{max_length}}Maximum length, e.g., "100 words", "5 bullet points"
{{focus_areas}}What to emphasise, e.g., "decisions made and action items", "technical implementation details"
{{exclusions}}What to leave out, e.g., "background context", "historical data before 2023", or "None"
{{text}}The full text to summarise

Example

Input
audience: a product manager with no engineering background
output_format: 3 bullet points followed by one-sentence next action
max_length: 80 words
focus_areas: decisions made and blockers
exclusions: technical implementation details
text: [500-word engineering meeting transcript]
Output
- The team decided to delay the payment API integration to Q3 to unblock the mobile launch.
- Database migration has a hard blocker: the legacy schema must be audited before migration can begin.
- Two engineers are now allocated full-time to the mobile release; no other projects until April 15th.

Next action: Product to update the Q2 roadmap by Friday to reflect the payment API delay.

Related Tools

AI API Request BuilderAI Prompt BuilderAI Token Counter

FAQ

How long can the input text be?
GPT-4o and Claude 3.5 Sonnet both support at least 128k tokens of input, which is roughly 90,000 words. For very long documents, consider chunking the text into sections and summarising each, then summarising the summaries.
Can I summarise a PDF with this prompt?
Extract the text from the PDF first (using a tool like pdf-to-text or copy-paste from a reader), then paste it into the text field. Claude and GPT-4o also support direct PDF upload through their web interfaces.
How do I preserve numerical data accurately in a summary?
Add to focus_areas: "preserve all specific numbers, percentages, and dates exactly as stated". LLMs sometimes round numbers or conflate similar figures, so always verify key metrics against the source.

Related Prompts

Text Summarisation Prompt

Summaries fail when the AI does not know what the reader already knows or what they need t...

Text Summarisation Prompt (Claude)

Summaries fail when the AI does not know what the reader already knows or what they need t...

Text Summarisation Prompt (Gemini)

Summaries fail when the AI does not know what the reader already knows or what they need t...

Text Summarisation Prompt (LLaMA / Ollama)

Summaries fail when the AI does not know what the reader already knows or what they need t...

Blog Post Outline & Draft Prompt

Blog posts generated without structure guidance tend to produce generic, meandering conten...

Data Analysis Prompt

Most AI data analysis prompts produce vague observations like "sales increased in Q2". Thi...