AI Chatbot Prompt Templates: 50 Reusable Prompts for Research, Writing, Support and Productivity

Overview

Good chatbot prompts do more than ask for an answer. They define the task, provide relevant context, set boundaries and explain what a useful result should look like. The most reliable structure is:

• Role: Tell the assistant what perspective or function to use.
• Context: Supply the information it needs, including audience, source material and business rules.
• Task: State the action clearly and use one main objective where possible.
• Constraints: Set limits for length, tone, evidence, exclusions or permissions.
• Output format: Specify headings, tables, bullets, JSON, a checklist or another useful structure.
• Quality check: Ask the assistant to identify uncertainty, missing information or assumptions rather than filling gaps silently.

Replace the bracketed fields before using a prompt. For sensitive work, remove unnecessary personal or confidential information and review the output before it is shared or used in a decision. If you are choosing an assistant for a team, compare its controls and workflow fit using this guide to choosing the right AI chatbot.

Checklist by scenario

Research and analysis prompts

1. Research plan: “Act as a research assistant. Help me investigate [topic] for [audience]. Break the work into questions, useful source types, search terms and a proposed order of work. Mark which points require verification.”
2. Source summary: “Summarise the material below for [audience]. Separate key claims, supporting details, limitations and open questions. Do not add information that is not present. Material: [paste text].”
3. Evidence table: “Extract the main claims from [text]. Return a table with claim, supporting passage, source location, confidence level and information still needed.”
4. Compare viewpoints: “Compare the positions in these documents about [issue]. Present areas of agreement, disagreement, assumptions and unresolved questions. Quote only short passages where necessary.”
5. Research brief: “Create a concise brief on [topic] using the supplied notes. Include background, current question, relevant evidence, risks, decisions required and next steps. Flag any unsupported statement.”
6. Question generator: “Generate 15 useful questions for evaluating [product, proposal or topic]. Group them by cost, usefulness, implementation, risk and long-term maintenance.”
7. Fact-check preparation: “Review the following draft and list every statement that should be checked. Explain what evidence would confirm or challenge each statement. Draft: [text].”
8. Research gap analysis: “Given this research summary, identify missing information that could change the conclusion. Rank each gap by likely importance and suggest a practical way to investigate it.”
9. Decision matrix: “Build a weighted decision matrix for choosing between [options]. Use these criteria and weights: [criteria]. Show assumptions and explain how changing a weight could alter the result.”
10. Reading notes: “Turn the following material into structured notes with headings, definitions, examples, questions and a five-point recall checklist. Material: [text].”

Writing and editing prompts

11. Outline: “Create an outline for [content type] about [subject]. The intended reader is [audience], the goal is [goal], and the desired length is [length]. Include the purpose of each section.”
12. First draft: “Draft [content type] for [audience] using this brief: [brief]. Use a [tone] tone, make the advice concrete and leave placeholders where information is missing.”
13. Plain-language edit: “Rewrite this text for a reader who understands the topic but is not a specialist. Keep the meaning, remove unnecessary jargon and preserve important qualifications: [text].”
14. Structure edit: “Review this draft for flow. Identify repetition, missing transitions, weak ordering and sections that should be combined or separated. Recommend changes before rewriting: [text].”
15. Headline options: “Suggest 12 accurate headlines for [article]. The reader wants [benefit]. Keep them specific, avoid exaggerated claims and vary the angle without changing the subject.”
16. Summary and excerpt: “Write a one-sentence summary and a short excerpt for this article. Reflect the actual content, avoid hype and keep the excerpt under [character limit]: [text].”
17. Style consistency: “Edit the following copy to match these style rules: [rules]. Preserve factual meaning, mark any substantive change and list unresolved inconsistencies: [text].”
18. Counterargument: “Read this proposal as a careful sceptic. List the strongest objections, the evidence needed to answer them and revisions that would make the proposal clearer: [proposal].”
19. Content repurposing: “Convert this article into [number] short posts, one email outline and a checklist. Keep each version faithful to the source and label any detail omitted: [article].”
20. Final proofing: “Proofread this text for spelling, grammar, punctuation, unclear references and formatting errors. Return a table with original wording, suggested correction and reason: [text].”

Customer support and business prompts

21. Support reply: “Draft a helpful reply to this customer. Acknowledge the issue, answer only from the supplied policy, explain the next step and avoid promising an outcome not stated in the policy. Message: [message]. Policy: [policy].”
22. Escalation check: “Review this support request and decide whether it needs escalation under these rules: [rules]. Give the reason, missing information and a suggested internal note: [request].”
23. FAQ extraction: “Turn these repeated customer questions into an FAQ. Group similar questions, write concise answers from the source material and flag answers that need owner approval: [tickets or notes].”
24. Sentiment and intent: “Classify this message by intent and sentiment using these labels: [labels]. Provide a short reason and identify urgency or risk indicators. Do not infer personal characteristics: [message].”
25. Conversation handoff: “Create a handoff note from this conversation. Include customer goal, facts provided, steps already taken, unresolved issue, promised follow-up and relevant priority: [conversation].”
26. Product feedback: “Extract product feedback from these comments. Group themes, include representative paraphrases, estimate frequency only from the supplied sample and separate requests from complaints: [comments].”
27. Meeting agenda: “Create a [length]-minute agenda for [meeting purpose]. Include decisions required, preparation, owners, time boxes and a final action review.”
28. Meeting actions: “Extract decisions, actions, owners, deadlines and unresolved points from these notes. Use ‘not assigned’ or ‘not stated’ instead of guessing: [notes].”
29. Process documentation: “Turn these informal instructions into a standard operating procedure. Include prerequisites, numbered steps, checks, exception handling and a definition of completion: [instructions].”
30. Risk review: “Review this proposed workflow for operational, privacy, security and quality risks. Rank each risk, explain the trigger and suggest a proportionate control: [workflow].”

Data extraction and transformation prompts

31. Structured extraction: “Extract [fields] from the text below. Return valid JSON using exactly these keys. Use null when a value is absent and do not infer missing data: [text].”
32. Table conversion: “Convert these notes into a table with columns [columns]. Keep wording concise, preserve units and mark ambiguous entries for review: [notes].”
33. Duplicate detection: “Compare these records and identify likely duplicates. Explain the matching signals and return a review list rather than deleting or merging anything: [records].”
34. Keyword extraction: “Extract the most useful keywords and phrases from this text for [purpose]. Group them by topic, exclude generic filler and show the passage that supports each group: [text].”
35. Entity list: “Identify organisations, products, locations, dates and named concepts in this text. Keep the original wording and include the sentence where each item appears: [text].”
36. Classification: “Assign one or more labels from this controlled list: [labels]. Give a short evidence-based reason and flag cases where none of the labels fit: [content].”
37. Data quality audit: “Check this dataset for missing values, inconsistent formats, duplicates, impossible values and unclear labels. Return issue, example, likely cause and recommended correction: [data].”
38. Format conversion: “Convert the following content from [source format] to [target format]. Preserve all factual fields, explain any incompatible fields and do not create values: [content].”
39. Language detection: “Identify the primary language of each passage and note mixed-language or uncertain cases. Return passage ID, language, confidence description and reason: [passages].”
40. Redaction review: “Identify information in this text that may need removal before wider sharing, such as personal, confidential or security-sensitive details. Explain the category and suggest a neutral replacement: [text].”

Brainstorming and productivity prompts

41. Idea generation: “Generate 20 ideas for [goal] under these constraints: [constraints]. Group them by effort and potential usefulness, then recommend three to test first.”
42. Prioritisation: “Prioritise these tasks using urgency, impact, effort and dependency. State the assumptions and return a suggested order with the first action for each task: [tasks].”
43. Project plan: “Create a practical plan for [project]. Include milestones, dependencies, owners, risks, decision points and a definition of done. Assume [constraints].”
44. Weekly review: “Turn these notes into a weekly review with completed work, unfinished work, lessons, risks, decisions and three priorities for next week: [notes].”
45. Learning plan: “Create a four-week learning plan for [skill] at [level]. Include outcomes, short practice tasks, review points and a way to demonstrate progress.”
46. Alternative approaches: “Suggest three ways to solve [problem]: a fast approach, a low-risk approach and a scalable approach. Compare trade-offs, prerequisites and likely failure points.”
47. Workflow improvement: “Map this workflow from trigger to completion. Identify avoidable steps, handoffs, bottlenecks and automation opportunities, while noting where human review is still needed: [workflow].”
48. Decision preparation: “Prepare me for a decision about [subject]. List the choices, criteria, unknowns, questions for stakeholders and a recommendation format without making the decision for me.”
49. Reflection prompt: “Ask me 10 focused questions to review [project or experience]. Ask one question at a time, then summarise patterns only after I answer all questions.”
50. Action checklist: “Convert this goal into a checklist of actions that can each be completed in one session. Include order, dependencies, a completion test and the next smallest step: [goal].”

What to double-check

Before adopting any prompt, run a small test with realistic input. Check whether the assistant follows the requested format, distinguishes source material from assumptions and behaves sensibly when information is missing. Test at least one ordinary example, one incomplete example and one edge case.

For research or summarisation, verify important claims against the original material. For extraction, inspect whether fields are shifted, duplicated or silently inferred. For customer support, confirm that the response follows the approved policy and does not invent refunds, deadlines or capabilities. For business workflows, review permissions, data handling and the point at which a person must approve the result. A chatbot may produce a plausible answer even when the prompt or input is incomplete, so a review step is part of the design.

Keep a simple prompt record containing the name, purpose, owner, model or tool used, input assumptions, expected output, test examples and revision date. This makes prompts easier to maintain and compare. For implementation work, see the AI chatbot integration guide; for document-heavy tasks, the PDF and research summarisation guide provides a useful adjacent checklist.

Common mistakes

• Vague objectives: “Help with this” does not define a useful result. Name the task and audience.
• Too many tasks at once: Split research, drafting and checking into separate stages when accuracy matters.
• Missing source boundaries: Tell the assistant whether it may use general knowledge or only the supplied material.
• Unclear formats: Specify the fields, headings, length or structure required downstream.
• No uncertainty instruction: Ask for missing information and assumptions to be labelled rather than guessed.
• Confusing confidence with correctness: A fluent response still needs appropriate human review.
• Reusing stale context: Update product rules, team names, deadlines and workflow constraints when they change.
• Changing several variables at once: Revise one part of a prompt at a time so you can see what improved or worsened the result.

When to revisit

Review this prompt library before seasonal planning cycles, new projects and recurring reporting periods. Re-test prompts whenever you change the chatbot, connected tool, source format, business rule or intended audience. A prompt that worked well for a short email may need different constraints for a support queue, website chatbot or internal knowledge workflow.

Use this maintenance checklist:

• Confirm the prompt still solves the intended task.
• Replace outdated context, examples, policies and output fields.
• Run ordinary, incomplete and edge-case inputs.
• Compare results with the previous version using the same test examples.
• Record known failure modes and the required human review.
• Archive versions rather than overwriting a prompt that is still in use.
• Retire prompts that duplicate a better workflow or no longer produce a reliable result.

Start with three templates that match a real task this week. Save the original prompt, the test input and the edited output, then make one deliberate improvement. Over time, this small version-tracking habit will turn a general prompt library into a dependable set of chatbot prompts for your own work.