Quick-pick table: top prompts at a glance(10)
25 Best ChatGPT Prompts for Customer Support: the top prompts and what you get from each
PromptWhat you get
Triage pasted support tickets by urgency and topicSort raw incoming support tickets into prioritized queues by urgency, root issue, and recommended first-touch actions.
Respond to an Abusive Ticket While Solving the IssueGenerates a calm customer support reply that establishes clear conduct boundaries while resolving the underlying technical problem.
Reply to an Angry Customer Complaint Without Blame or Vague PromisesPaste an angry customer message and get a calm reply that names their exact problem first, skips blame-shifting phrases, and ends with one dated next step.
Write Saved Replies (Macros) for Your Top Five Support Ticket TypesGet five short, ready-to-send support macros, each with an editable personalization line, clear numbered steps, and one closing question that moves the ticket toward done.
Customer Support Macro Set for Refunds, Delays, Angry Customers and EscalationWrites 10 ready-to-send support reply macros in your brand tone covering refunds, delays, angry customers and escalation.
Turn Support Ticket Topics into a Help Center Category Structure and Article ListGroups your pasted ticket topics into 5-8 customer-language categories, flags duplicates and orphans, and ranks proposed article titles by ticket volume.
Build a Yes/No Troubleshooting Decision Tree for a Recurring Product IssueYou get a step-by-step yes/no troubleshooting tree ordered from fastest, most likely checks to slowest, with ready-to-send customer messages and a clear escalation point.
Turn a Resolved Support Ticket Thread into a Help Center ArticlePaste a solved ticket thread and get a ready-to-publish help article with customer details removed, verb-first fix steps, and symptoms written the way customers search.
Audit a Help Center Article for Gaps, Jargon and Outdated StepsPaste a help center article and get a pass/fail checklist scorecard plus a line-by-line fix list ranked by customer impact, without a full rewrite.
Find Help Center Gaps by Matching Support Tickets to Existing ArticlesGet a ticket-to-article match table and a ranked list of new help articles to write, with ticket counts and deflection potential for each.

De-escalation and apologies

9 prompts

Calm frustrated customers, acknowledge outages or delays plainly and offer clear next steps without sounding robotic.

  1. #1Customer SupportWith example

    Respond to an Abusive Ticket While Solving the Issue

    Generates a calm customer support reply that establishes clear conduct boundaries while resolving the underlying technical problem.

    Preview prompt
    Act as a senior customer support manager skilled in de-escalation. Your job is to draft a reply to an angry customer ticket that contains profanity, personal insults, or abusive language, addressing their issue while setting firm boundaries.
    
    Rules:
    1. State a clear, neutral boundary regarding respectful communication in the first two sentences without sounding defensive or scolding.
    2. Do not apologize for the customer's feelings or validate verbal abuse; acknowledge only the factual issue reported.
    3. Isolate the core technical or operational problem and provide direct, step-by-step resolution steps or immediate status.
    4. Close with a clear requirement that future communication remain respectful to maintain support.
    
    Output format:
    - Subject Line
    - Email Body (including boundary statement, technical resolution, and clear next steps)
    - Internal Notes (a 2-sentence summary for the internal ticket record regarding the warning issued)
    
    Paste the angry customer ticket below this prompt.
  2. #2Customer SupportWith example

    Reply to an Angry Customer Complaint Without Blame or Vague Promises

    Paste an angry customer message and get a calm reply that names their exact problem first, skips blame-shifting phrases, and ends with one dated next step.

    Preview prompt
    You are a senior customer support lead who handles escalated complaints.
    
    Your job: write a reply to the angry customer message I will paste right after this prompt. I may add notes on what I can offer (refund, replacement, call-back). If I give no notes, first ask me up to 3 short numbered questions in one message (what went wrong on our side, what I can offer, by when) and wait for my answers before writing.
    
    Rules:
    1. The first sentence names the customer's specific problem in their terms (product, order number, date if given). Apologize once, for that specific thing, then move to the fix.
    2. Do not use blame-shifting or hedging phrases such as "per our policy", "as stated", "we apologize if", "any inconvenience this may have caused", "you should have", "unfortunately you".
    3. Never promise anything I have not confirmed. If something is unclear, flag it for me instead of guessing.
    4. End with exactly one concrete next step: who does it and by when (day and time).
    5. Keep the reply under 150 words, plain sentences, no exclamation marks.
    
    Output format:
    - Subject line
    - Reply email
    - Notes for me: phrases avoided and why, plus anything to confirm before sending
    
    Finish with one line naming any assumption you made about the facts.
  3. #3Customer SupportWith example

    Rewrite a Support Reply to Be Clearer, Shorter and Friendlier

    Returns your support draft cut by at least 30 percent, with every fact and commitment kept and a table of risky promises plus safer wording.

    Preview prompt
    Act as a senior customer support editor who trains agents on tone and clarity.
    
    I will paste a draft support reply right after this message, plus the customer's original message if I have it. Rewrite the draft so it is clearer, shorter and friendlier.
    
    Rules:
    1. Cut the word count by at least 30 percent. Report the original and new word counts.
    2. Keep every fact, number, date, link, order or ticket reference and commitment from the original. Do not add new promises, discounts or policies.
    3. Open with the answer or next step in the first sentence. Use at most one short apology, and only if something went wrong.
    4. Use plain words, sentences under 20 words, and address the customer as "you". Put any steps the customer must take in a numbered list.
    5. Flag every promise the agent may not be able to keep, such as exact deadlines, guaranteed outcomes or "this won't happen again", and use safer wording in the rewrite.
    
    Output format:
    **Rewritten reply** (ready to send)
    **Word count:** original vs. new, percent cut
    **Facts and commitments kept:** checklist
    **Risky promises:** table with columns Original wording | Why it's risky | Safer wording
    **Tone note:** one line, only if the customer seems upset
    
    If anything in the draft is unclear or contradicts itself, point it out instead of guessing.
  4. #4Customer SupportGeminiWith example

    Reply to a Public Social Media Complaint and Move It to DM

    Get a 280-character public social media reply that de-escalates a complaint, plus a matching direct-message follow-up script to resolve it privately.

    Preview prompt
    Act as a senior customer support specialist skilled in social media de-escalation. I will paste a public customer complaint and any relevant context below.
    
    Your job is to write a calm public reply that moves the conversation to direct messages, followed by a private DM script to resolve the issue.
    
    Rules:
    1. Keep the public reply under 280 characters, including spaces.
    2. Acknowledge the customer's frustration without being defensive, making excuses, or blaming the customer.
    3. Do not ask for sensitive details (order numbers, emails, or account IDs) in the public reply.
    4. Make the DM script warm, direct, and focused on gathering only the exact information needed to fix the problem.
    
    Output format:
    - **Public Reply** (include the exact character count)
    - **Private DM Follow-Up** (the opening message to send once they reach out)
    - **Internal Handoff Note** (2-3 bullet points on what to check before replying further)
    
    End with one line noting any missing context or policy risks to verify before posting.
  5. #5Customer SupportGeminiWith example

    Score a Support Agent Reply Against a Quality Assurance Rubric

    Get a four-criteria QA scorecard table grading tone, accuracy, completeness, and next steps, plus a ready-to-send rewrite of your support agent's reply.

    Preview prompt
    Act as a senior customer support quality assurance lead. I will paste a customer's message and a support agent's draft reply below. Grade the draft reply against a four-criteria QA rubric and provide a rewritten version.
    
    Rules:
    - Score each criterion—Tone, Accuracy, Completeness, and Next Steps—on a 1 to 5 scale, where 5 means ready to send with no edits.
    - Cite one specific phrase from the draft reply in every scorecard row to justify the score.
    - Flag any policy or factual claims that cannot be verified from the customer's message alone.
    - Write the improved reply in plain, empathetic language under 180 words with a clear, single next step.
    
    Output format:
    1. Scorecard Table: Markdown table with columns for Criterion, Score (1-5), Evidence from Draft, and What to Fix.
    2. Overall QA Score: Total out of 20 and a one-sentence pass or revise verdict.
    3. Rewritten Reply: The full, ready-to-send message.
    
    End with one line noting any missing account context or unverified policy details.
  6. #6Customer SupportWith example

    Write Executive Escalation Email for At-Risk Customers

    Generate a professional, high-impact email from a VP to retain a churn-risk customer by offering ownership, a concrete fix, and direct communication.

    Preview prompt
    Act as a VP of Customer Success. You are writing to a high-value customer who has threatened to churn due to recurring service issues. Your goal is to de-escalate the situation and re-establish trust. First, ask the user these 3 questions: 1. What is the customer name and company? 2. What are the 2 primary reasons they are threatening to leave? 3. What is the specific 'make-good' offer or fix you are prepared to give them? Once the user answers, write the email using these rules: 1. Start with a direct apology that takes full ownership of the failure. 2. Use a professional, human tone without corporate jargon. 3. List an immediate 3-step action plan to resolve their issues. 4. Propose a specific time for a 15-minute call next week. 5. Keep the email under 200 words. Output format: Subject Line, then the Email Body. This is not professional legal advice; consult with your legal department before making binding financial commitments.
  7. #7Customer SupportWith example

    Draft a Bug Acknowledgment Email Without an ETA

    Generate a clear support email that validates a reported bug, provides a safe temporary workaround, and sets realistic follow-up expectations.

    Preview prompt
    Act as a senior customer support specialist handling software defects. Write a customer response acknowledging a confirmed bug where engineering cannot commit to a release date yet.
    
    First, ask me these 3 numbered questions and wait for my response:
    1. What is the bug and which customer workflow does it disrupt?
    2. What is the approved workaround, and what internal tracking ID is tied to it?
    3. What is our realistic follow-up cadence (e.g., weekly update, or only upon patch release)?
    
    Once I provide those details, draft the reply following these rules:
    - Validate the user's issue clearly without assigning blame or making vague promises.
    - Outline the temporary workaround in numbered steps that prevent data loss or duplicate work.
    - State transparently that an exact release date is not yet scheduled, and define precisely when you will contact them next.
    - Include the tracking ticket ID so the customer has a concrete reference.
    
    Output format:
    - Subject Line (under 60 characters)
    - Email Body (greeting, confirmation, step-by-step workaround, status explanation, committed next touchpoint, sign-off)
    - Internal Note (one sentence recording tracking ID and next check-in date)
    
    State one operational risk if the workaround is performed incorrectly.
  8. #8Customer SupportWith example

    Reply calmly to a negative public customer review

    Draft composed, professional public responses to critical reviews that acknowledge mistakes, de-escalate tension, and move resolution offline.

    Preview prompt
    Act as a small business owner drafting a calm, constructive public reply to a negative review.
    
    - Acknowledge the reviewer's specific complaint directly without making defensive excuses or arguing facts publicly.
    - Demonstrate accountability and state the standard your business normally upholds.
    - Provide a direct private contact channel (such as a specific owner email address) to resolve the matter one-on-one.
    - Write for two audiences: the dissatisfied customer and future shoppers evaluating how you handle problems.
    
    Output format:
    Public reply draft
    Internal checklist of 2 key facts to verify before posting
    
    Here is the negative review:
  9. #9Customer SupportWith example

    Write a Refund or Return Denial Email That Keeps the Customer

    Get a clear, polite denial email that explains the policy reason in plain words, offers an approved alternative, and includes a short chat version and escalation-risk note.

    Preview prompt
    You are a senior customer support lead who writes refund and return replies that hold the line on policy without losing the customer.
    
    Your job: write a reply denying a refund or return request. Before writing, ask me these questions in one numbered message and wait for my answers:
    1. What does the relevant policy say? Paste the exact wording if you have it.
    2. What are the order details: product, order date, request date, price, and the customer's stated reason?
    3. What can the team actually offer instead (store credit, exchange, repair, partial refund, discount)? List only what is approved.
    4. Paste the customer's message or describe their tone.
    
    Rules:
    - State the decision within the first two sentences.
    - Explain the reason in plain words, naming the specific policy point and the date or condition that applies.
    - Offer at least one alternative from my approved list only; never promise anything I did not approve.
    - No blame, use "unfortunately" at most once, and avoid "per our policy".
    - Keep the email under 170 words and end with one clear next step for the customer.
    
    Output format:
    - Subject line
    - Email reply
    - Internal note: escalation risk (low, medium or high) and why, in one line
    - Two-sentence version for chat or SMS
    
    If any policy detail or approved alternative is missing, say so instead of guessing.

Refunds and policy replies

3 prompts

Explain billing limits, return windows and policy exceptions clearly while preserving customer trust.

  1. #10AI Business WorkflowsWith example

    Customer Support Macro Set for Refunds, Delays, Angry Customers and Escalation

    Writes 10 ready-to-send support reply macros in your brand tone covering refunds, delays, angry customers and escalation.

    Preview prompt
    Act as a senior customer support lead who writes replies agents can send with minimal edits. Job: write 10 reply macros for my support team. First ask me up to two short questions: what type of business we are, including our refund window and typical delivery times, and what our brand tone is. Wait for my answers.
    
    Rules:
    - Write exactly these 10: 1 refund approved, 2 refund denied, 3 partial refund, 4 delay with a new date, 5 delay with no date yet, 6 angry customer first reply, 7 angry customer follow-up, 8 escalation to a manager, 9 escalation to a specialist team, 10 check-in after resolution.
    - Use bracketed fields such as [customer name] and [order number] for agent details. Never invent policies. Write [CONFIRM POLICY] where you do not know one.
    - Keep each body under 120 words, with no blame, one clear next step and the agreed tone.
    
    Output: for each macro, give name, when to use, when not to use, subject line, body, fields to fill, and escalate-if note. After the 10, add a tone checklist, risks and policy items to confirm, and three next actions to roll the set out. The set is good if an agent could send any macro in under a minute.
  2. #11Customer SupportWith example

    Turn Around Negative CSAT Scores with a Calm Follow-Up Email

    Generates a calm, non-defensive CSAT follow-up email that asks two diagnostic questions to uncover support failure points and win back upset customers.

    Preview prompt
    Act as a customer support lead handling churn risks. Write an email to follow up on a negative CSAT rating or dissatisfied survey response.
    
    First, ask me 3 brief questions and wait for my answers:
    1. What was the customer's original issue, and what went wrong during the interaction?
    2. What is the customer's name, their company or account type, and their assigned representative's name?
    3. What immediate fix, credit, or next step can we offer right now?
    
    Once I reply, generate the email adhering to these rules:
    - Keep the tone calm, accountable, and entirely non-defensive; do not blame tooling or company policies.
    - Include exactly two focused diagnostic questions asking what went wrong, phrased to invite open feedback rather than a yes/no response.
    - Keep the total length under 180 words.
    - Never make up fake policy exceptions or guarantee timeline promises that depend on third parties.
    
    Output format:
    Subject line options (provide 2)
    Email body
    Recommended internal tag or tracking category for this case
    
    State one assumption about the customer's contract status or remaining goodwill.
  3. #12Customer SupportGeminiWith example

    Analyze Customer Cancellation Reasons to Surface Preventable Churn Fixes

    Get a structured churn analysis table separating controllable from uncontrollable cancellations, with exact counts, root causes, and one retention fix per category.

    Preview prompt
    Act as a customer retention analyst. I will paste raw customer cancellation reasons, exit survey notes, or support tickets right after this prompt. Analyze the data to separate preventable churn from uncontrollable churn and surface concrete fixes.
    
    Rules:
    - Count every cancellation entry and group them into clear, mutually exclusive categories.
    - Classify each category as Controllable (product gaps, onboarding, support, pricing) or Uncontrollable (business closed, seasonal pause, acquired).
    - Identify the specific root cause behind each category using only the pasted data; do not invent reasons or statistics.
    - Propose exactly one actionable retention fix per controllable category that a support or product team can test within 30 days.
    
    Output format:
    1. Summary Table: Columns for Category, Type (Controllable or Uncontrollable), Exact Count, Root Cause, and 30-Day Retention Fix.
    2. Top 3 Priorities: A short numbered list of the highest-impact controllable categories to tackle first, with one sentence explaining why.
    
    End with one line noting any ambiguous entries, missing context, or assumptions made while grouping the data.

Help center articles and macros

11 prompts

Convert solved tickets into scannable self-service guides, FAQ entries and reusable agent macros.

  1. #13Customer SupportWith example

    Write Saved Replies (Macros) for Your Top Five Support Ticket Types

    Get five short, ready-to-send support macros, each with an editable personalization line, clear numbered steps, and one closing question that moves the ticket toward done.

    Preview prompt
    You are a support team lead who writes saved replies (macros) that agents can send in under a minute.
    
    Before writing anything, ask me these questions in one message and wait for my answers:
    1. What is your product, and who are your customers?
    2. What are your top five ticket reasons, in order of volume?
    3. What tone does your team use (formal, friendly, casual), and how do you sign off?
    4. What must agents never promise (refunds, timelines, features)?
    
    Then write one macro per ticket reason. Rules:
    - Each macro is 60-110 words in plain language with no internal jargon.
    - Start with one line labeled "Personalize:" that shows the agent what to change per ticket, such as the customer's name or the exact error they quoted.
    - Give the fix or next step as numbered steps, no more than 4.
    - End with exactly one closing question that either confirms the issue is solved or asks for the single missing detail.
    - Never promise anything from my answer to question 4.
    
    Output format:
    For each ticket reason: a heading with the macro name, a "When to use" line, the macro text, and a "Don't use if" line.
    
    Finish with a short note on which macros need a policy or accuracy check before going live, and any detail you had to assume.
  2. #14Customer SupportWith example

    Turn Support Ticket Topics into a Help Center Category Structure and Article List

    Groups your pasted ticket topics into 5-8 customer-language categories, flags duplicates and orphans, and ranks proposed article titles by ticket volume.

    Preview prompt
    Act as a help center content strategist who organizes self-service sites for support teams.
    
    I will paste a list of ticket topics right after this prompt, ideally with a ticket count next to each. Turn them into a help center category structure and a prioritized article list.
    
    Rules:
    1. Create 5-8 categories named in the words customers use in tickets, not internal team or feature names (e.g. "Billing and payments", not "Revenue Ops").
    2. Put every topic in exactly one category. If a topic fits two, pick one and note the cross-link.
    3. Flag duplicates: topics that describe the same problem in different words become one article, with the merged topics listed.
    4. Flag orphans: topics that fit no category or appear only once. Say whether to drop, merge or keep each one.
    5. Write article titles as the task or question a customer would type into search, under 70 characters.
    
    Output format:
    1. Categories: name, one-line scope, total ticket count.
    2. Article table with columns: Rank | Article title | Category | Ticket topics covered | Ticket volume. Sort by volume, highest first.
    3. Duplicates merged (bullets).
    4. Orphans and recommendation (bullets).
    
    If counts are missing, rank by how often a topic appears and say so. End with one line on gaps or assumptions in the data.
  3. #15Customer SupportWith example

    Build a Yes/No Troubleshooting Decision Tree for a Recurring Product Issue

    You get a step-by-step yes/no troubleshooting tree ordered from fastest, most likely checks to slowest, with ready-to-send customer messages and a clear escalation point.

    Preview prompt
    Act as a senior support operations lead who writes troubleshooting guides for frontline agents.
    
    Your job: build a yes/no troubleshooting decision tree for one recurring product issue my team handles.
    
    Before writing anything, ask me these questions in one message, then wait for my answers:
    1. What is the issue, in the customer's words, and which product or plan does it affect?
    2. What are the known causes, and roughly how often is each one the culprit?
    3. What can frontline agents check or fix themselves, and what needs another team?
    4. What must an escalation ticket include?
    
    Rules:
    - Order checks by speed and likelihood: quick checks for common causes first, slow or rare checks last.
    - Every step is one question answerable only Yes or No, and each answer points to a fix ID or the next step number.
    - Give each step a time estimate in minutes.
    - Each fix includes 1-2 sentences an agent can send the customer word for word.
    - Mark the single escalation point with the word ESCALATE. No step may require access agents don't have.
    
    Output format:
    1. Issue summary (2 lines)
    2. Decision tree: numbered steps, each with Check, Time, If Yes, If No
    3. Escalation block: trigger, receiving team, required info
    4. Fix table with columns: Fix ID, Cause, Agent action, Customer message
    
    End with one line listing your assumptions and any likely causes I didn't mention.
  4. #16Customer SupportWith example

    Turn a Resolved Support Ticket Thread into a Help Center Article

    Paste a solved ticket thread and get a ready-to-publish help article with customer details removed, verb-first fix steps, and symptoms written the way customers search.

    Preview prompt
    You are a senior support content writer who turns solved tickets into help center articles.
    
    I will paste a resolved support ticket thread right after this prompt. Turn it into one public help center article that the next customer with the same problem can follow without contacting us.
    
    Rules:
    1. Remove every customer detail: names, emails, account IDs, order numbers, company names, private links and anything only that customer would know. Use generic wording like "your account".
    2. Write the fix as numbered steps. Each step starts with a verb, covers one action, and names the exact button, menu or field used in the thread.
    3. List symptoms in the customer's own words, lightly cleaned from the thread, so the article matches what people type into search.
    4. Leave out things the agent tried that did not work, unless a step needs to warn against them.
    5. If the thread does not show how the fix was confirmed, say so instead of guessing.
    
    Output format:
    - Title: phrased as the problem, under 70 characters
    - Who this affects (1-2 sentences)
    - Symptoms (bullets, customer wording)
    - Cause (2-3 plain sentences)
    - How to fix it (numbered steps)
    - If this didn't work (what to send support)
    - Search keywords (5-8 phrases)
    
    End with Notes for the editor: details removed, assumptions made, and gaps to check with the product team.
  5. #17Customer SupportWith example

    Audit a Help Center Article for Gaps, Jargon and Outdated Steps

    Paste a help center article and get a pass/fail checklist scorecard plus a line-by-line fix list ranked by customer impact, without a full rewrite.

    Preview prompt
    You are a help center content lead who audits support articles before they are republished.
    
    I will paste an existing help center article right after this message. Audit it for gaps, jargon and outdated steps. Do not rewrite the whole article. Give targeted fixes only.
    
    Rules:
    1. Score the article against this checklist, marking each item Pass, Partial or Fail with a one-line reason: one task per article, title names the task, numbered steps with one action each, UI labels written exactly and in bold, screenshots noted where a step is ambiguous, prerequisites listed first, last-verified date present, related articles linked.
    2. Quote the exact line for every issue before the fix, so I can find it with Ctrl+F.
    3. Flag any term a new customer would not know and give a plain replacement.
    4. Mark steps that look outdated (old menu names, removed features, version numbers) as "Verify in product" instead of guessing the current UI.
    5. Rank each fix: High (blocks the task), Medium (causes confusion), Low (polish).
    
    Output format:
    1. Scorecard: table with columns Check, Result, Reason, then a total out of 8.
    2. Fix list: table with columns #, Quoted line, Problem, Fix, Priority.
    3. Split recommendation: if the article covers more than one task, list proposed separate article titles.
    4. Three quick wins to do today.
    
    End by listing what you could not check without access to the live product.
  6. #18Customer SupportWith example

    Find Help Center Gaps by Matching Support Tickets to Existing Articles

    Get a ticket-to-article match table and a ranked list of new help articles to write, with ticket counts and deflection potential for each.

    Preview prompt
    You are a support operations analyst who audits help centers for self-service gaps.
    
    Right after this prompt I will paste two things: (1) a batch of support tickets (subjects, summaries or full text) and (2) the list of article titles in our help center. Find which ticket reasons have no good article and tell me what to write first.
    
    Rules:
    1. Group tickets into contact reasons written in the customer's words, and count tickets per reason. Each ticket goes in exactly one reason.
    2. Mark a reason Covered only if an article title clearly answers it, Partial if an article exists but likely misses what customers ask, and Gap if nothing fits.
    3. Rate deflection potential High, Medium or Low. High only if a customer could fix it alone with clear steps; Low if an agent must act on the account (refunds, bugs, account changes).
    4. Rank articles to write by ticket count times deflection score (High 3, Medium 2, Low 1), not by count alone.
    5. Use only the tickets and titles I paste. Do not invent volumes or articles.
    
    Output format:
    - Table 1: Contact reason | Tickets | Matched article | Status
    - Table 2: Priority | Proposed article title | Reasons covered | Tickets | Deflection | Score | Why
    - 3 bullets: quick fixes for Partial articles
    
    End with one line on what this ticket sample cannot show.
  7. #19Customer SupportWith example

    Turn Feature Notes into a Step-by-Step Help Center How-To Guide

    Paste rough notes about a product feature and get a customer-ready how-to guide with one action per step, bold UI labels, expected results and a troubleshooting section.

    Preview prompt
    You are a technical writer who builds help center articles for a software support team.
    
    I will paste my notes about a product feature right after this prompt. They may be internal notes, a Slack thread, release notes or a rough draft. Turn them into a step-by-step how-to guide that a customer can follow without contacting support.
    
    Rules:
    1. One action per numbered step. If a step joins two actions with "and" or "then", split it into two steps.
    2. Write every on-screen label (buttons, menus, tabs, fields) in bold, spelled exactly as it appears in my notes. Do not rename or guess labels.
    3. After any step that saves, submits or changes the screen, add a line starting with "Result:" that says what the user should see.
    4. Use short sentences, second person ("you") and plain words. Remove internal jargon, code names and ticket numbers.
    5. Put prerequisites (plan, permissions, settings) before step 1, not inside the steps.
    
    Output format:
    - Title starting with "How to"
    - One sentence on what the guide helps the user do
    - "Before you start" bullet list
    - Numbered steps
    - "If this didn't work": 3-5 likely problems, each with cause and fix, based only on my notes
    - "Still stuck?": one line on contacting support
    
    End with a short list of any labels, steps or permissions my notes left unclear, so I can confirm them before publishing.
  8. #20Customer SupportWith example

    Write Help Article Titles and Meta Descriptions in Your Customers' Words

    Get help article titles that start with the customer's task verb and meta descriptions within search length limits, each with a character count and the customer phrase it matches.

    Preview prompt
    Act as a help center content strategist who writes article titles for search.
    
    Your job: write help article titles and meta descriptions that match the words customers type into your help center search and Google.
    
    Before writing, ask me these questions in one message, numbered, and wait for my answers:
    1. What is the product, and who uses it?
    2. Which help articles need titles? Give the topic of each.
    3. Paste real customer phrasing: ticket subject lines, search queries or chat messages.
    4. Which internal feature names differ from the words customers use?
    
    Rules:
    - Every title starts with the task verb the customer would use (Cancel, Export, Change), not a feature name or "How to".
    - Prefer the customer's wording over internal terms. If an internal term matters, put it in the meta description.
    - Titles are 30-60 characters. Meta descriptions are 120-155 characters. Count characters including spaces and report the exact count.
    - Each meta description states the outcome plus one detail that helps readers pick the right article, such as plan, device or limits.
    
    Output format:
    A table with columns: Topic | Title | Chars | Meta description | Chars | Customer phrase matched
    Then a list titled "Internal terms replaced," showing each old term and the new wording.
    
    End with one line naming any article where the customer phrasing was thin and the title is a best guess.
  9. #21Customer SupportWith example

    Turn Pasted Customer Questions into a Clear FAQ Page with Short Answers

    You get a ready-to-publish FAQ page with duplicate questions merged, answers under 60 words drawn only from your facts, and a list of gaps.

    Preview prompt
    You are a customer support lead who writes help-center FAQ pages.
    
    After this message I will paste two things: raw customer questions (emails, chat logs, tickets, reviews) and the facts you may use (policies, product notes, pricing, shipping details). If I paste only questions, ask me for the facts and wait before writing.
    
    Your job: turn the questions into a clean FAQ page customers can scan in under two minutes.
    
    Rules:
    1. Merge questions that ask the same thing in different words into one entry, and count how many times it came up.
    2. Write each question the way a customer would type it, in plain words, under 15 words.
    3. Answer each question in under 60 words, using only the facts I provide. Never guess prices, dates, timelines or policies.
    4. Start every answer with the direct answer (yes, no, a number or the first step), then add one line of detail or the next step.
    5. Group entries into topic sections and order entries by how often they were asked.
    
    Output format:
    - Page title
    - Topic sections as H2 headings, each with Q&A pairs: question in bold, answer below, then (asked X times)
    - Table titled Questions I Could Not Answer: Question | Times asked | Fact needed
    - List titled Duplicates Merged: the original wordings combined into each entry
    
    Finally, flag any facts that conflict or look out of date, and name the answers that depend on them.
  10. #22Customer SupportWith example

    Write Chatbot Intent Answers with Trigger Phrases and Human Handoff Rules

    Get a set of chatbot intents for your top support questions, each with customer-style trigger phrases, a short bot answer, and an exact rule for handing off to a human.

    Preview prompt
    You are a support operations lead who designs chatbot flows for help centers.
    
    Your job: turn my top support questions into chatbot intents, each with trigger phrases, a short bot answer, and a clear handoff rule. First ask me these in one message and wait for my answers:
    1. What does your company sell, and who contacts support?
    2. List your top 5-10 support questions, or paste ticket subjects or FAQ text.
    3. What can the bot actually do (look up orders, reset passwords, or only answer)?
    4. When are agents available, and how does a handoff work?
    
    Rules:
    - Each bot answer is 60 words or fewer and ends with one next step or one question.
    - Give 5-8 trigger phrases per intent, written the way customers type, including typos and short fragments.
    - Each handoff rule names an exact, checkable condition, such as the customer asking twice, mentioning a charge dispute, legal or safety issue, angry language, or an account the bot cannot verify.
    - The bot never promises refunds, dates or policy exceptions it cannot confirm.
    - Include one fallback intent: one clarifying question at most, then handoff.
    
    Output format:
    For each intent: Intent name, Trigger phrases, Bot answer, Handoff rule, Handoff message to customer.
    Then a short Global handoff rules list.
    
    End with one line listing gaps: questions my info did not cover and policies I need to confirm.
  11. #23Customer SupportGeminiWith example

    Write a Customer Troubleshooting Guide for Specific Error Codes

    Get a customer-facing troubleshooting guide that maps each error code to a plain-English cause, three numbered fix steps, and a clear escalation trigger.

    Preview prompt
    Act as a senior technical writer for a customer support team. Turn my raw error-code notes into a customer-facing troubleshooting guide.
    
    I will paste my error codes, internal causes, and product context right after this prompt.
    
    Rules:
    1. Write in plain, calm English that non-technical customers can follow.
    2. Give each error code one plain-English cause, three numbered self-serve fix steps, and one exact moment to contact support.
    3. Start every fix step with a clear action verb and name the exact menu, button, or setting to click.
    4. Do not invent error codes, UI labels, or support hours that I did not provide.
    
    Output format:
    - Overview: Two sentences explaining how to use the guide.
    - Error Code Sections: For each code, include a heading with the code and short title, followed by **What this means**, **Try these 3 steps** (numbered 1-3), and **When to contact support**.
    - Quick-Reference Table: Columns for Error Code | Likely Cause | First Step | Escalate If.
    
    End with one line noting any missing UI labels, unverified fix steps, or assumptions.

Ticket triage and escalations

2 prompts

Summarise long threads cleanly for engineering, classify ticket themes and tighten handoff notes.

  1. #24Customer SupportWith example

    Triage pasted support tickets by urgency and topic

    Sort raw incoming support tickets into prioritized queues by urgency, root issue, and recommended first-touch actions.

    Preview prompt
    Act as a support triage lead. Sort the customer tickets pasted below. Use only what the tickets say and do not guess missing details.
    
    - Urgency: P1 = customer is locked out, cannot use the product, or has a billing error. P2 = something is broken but a workaround exists. P3 = how-to question, feedback, or feature request.
    - Topic: Billing, Bug, Feature Request, How-To, or Account Access.
    - Escalation Risk: Yes if the customer threatens to cancel, asks for a manager, or mentions a deadline. Otherwise No.
    - Next Action: one sentence on what the agent should do first.
    - If a ticket has no number or name, write Not given.
    
    Output format:
    1. Table with columns: Ticket #, Customer, Topic, Urgency, Escalation Risk, Next Action
    2. P1 Action Items: one bullet per P1 ticket with a short reason, or None.
    
    Tickets:
  2. #25Customer SupportWith example

    Turn vague feature requests into actionable workflow feedback

    Extract underlying user problems, manual workarounds, and workflow impact from brief customer feature requests.

    Preview prompt
    Act as a customer support lead and product operations specialist. Analyze customer feature requests to uncover the root problem behind what they asked for.
    
    - Identify the core job-to-be-done and the probable reason they requested this specific solution.
    - Determine the manual workaround or friction the user is likely enduring right now.
    - Draft a warm, two-question reply to the user asking clarifying questions about their actual workflow without committing to building the request.
    - Summarize the insight into an internal ticket format for product teams.
    
    Output format:
    1. Core Friction Analysis (underlying problem vs. surface request)
    2. Customer Follow-up Email (concise, polite, non-committal)
    3. Product Ops Ticket (Summary, User Goal, Current Blocker, Impact Level)
    
    Customer request is pasted below:

Playbook

How to get better results

Which AI model to pick

ChatGPT
Fast macro drafting, help centre article formatting and summarising weekly ticket clusters.
Claude
Nuanced de-escalation replies and reading long, messy customer threads without losing tone.
Gemini
Updating shared support playbooks and training guides directly in Google Docs.

Follow-up prompts to refine any output

  1. Cut any hollow apology phrasing and put the concrete resolution or next step in the very first paragraph.
  2. Create a shorter macro version with clear bracketed placeholders an agent can fill in seconds.
  3. Summarise this thread for an engineer in three bullets: expected behaviour, actual behaviour and steps already tested.

FAQ

About this list

Frustrated customers spot robotic apologies immediately. These prompts ask for the real ticket context, your policy boundaries and the concrete fix you can offer, then draft a direct, empathetic reply that owns the next step without overpromising or hiding behind stiff corporate language. Each prompt shows a complete example reply so you can check the tone before using it with customers.

Beyond one-off replies, the list includes operational prompts for support leads: turning solved tickets into searchable help centre articles, trimming macros so agents spend less time editing, packaging bug reports for engineering and spotting repeat friction themes across a week of conversations. Strip customer account details before pasting, and run any prompt in ChatGPT, Claude or Gemini.

How do I keep AI support replies warm without sounding scripted?

Paste a couple of your team's best real replies as a tone reference and ban hollow openers such as thanking the customer for their patience when something broke. State what happened plainly, explain what you have already done and give a realistic timeframe for the next update.

What customer details should I remove before pasting a ticket?

Always strip email addresses, phone numbers, billing addresses, order numbers, payment details and account tokens before pasting a thread into an external assistant. Replace them with simple placeholders such as Customer or Order ID. The prompt only needs the issue, timeline and policy context to draft a helpful response.

Can these prompts turn resolved tickets into help centre documentation?

Yes. Paste a resolved conversation after removing personal details and the documentation prompt rewrites it into a scannable help article with prerequisites, numbered troubleshooting steps and a short FAQ. Publishing guides from real tickets deflects repeat questions while keeping wording close to what customers actually search.

How do I handle a refund denial without escalating the conversation?

Use the policy-explanation prompt with the exact rule and any goodwill alternative you are allowed to offer, such as store credit, a partial adjustment or a plan pause. Leading with what you can do, explaining the boundary calmly and avoiding defensive wording keeps most difficult tickets constructive.