22 ChatGPT Prompts for Customer Service You Can Copy and Paste

ChatGPT prompts for customer service are ready-made instructions that tell an AI tool what reply, macro, or summary you need and how to format it. Good ones include the customer situation, your tone, and the output shape, so you get a usable draft instead of a stiff form letter. This page gives you 22 of them, grouped by the support job you are doing.

Most support prompts fail for one reason: they are too vague. "Write a reply to an angry customer" gives you the same generic apology every brand sends. The prompts below are different because each one carries context, a clear task, and a format.

You can copy any prompt as-is, then swap the bracketed placeholders for your product, policy, and situation. Use them with ChatGPT, Claude, or Gemini. They work the same way in each.

We built these around the Context/Task/Format framework, the same anatomy we teach in our pillar guide to prompt engineering for business. Below, we also show one before-and-after rewrite so you can see why adding tone and situation changes the output.

Key Takeaways

ChatGPT prompts for customer service work best when they name the customer situation, your tone, and the exact output format.

The points below summarize how to get the most from the prompt library on this page.

• Every prompt here uses the Context/Task/Format anatomy: who you are, what the customer needs, and how the reply should look
• Copy a prompt, then replace the bracketed placeholders with your real product, policy, and ticket details
• Grouped by job: replies, empathy and de-escalation, macros, help-center FAQs, ticket summaries, and follow-ups
• Never paste customer names, emails, or order details into a public AI tool, and always have a person review a reply before it reaches a customer
• The same prompts work in ChatGPT, Claude, and Gemini with no changes
• Use the AI prompt generator to build a custom prompt when none of these fit exactly

How to Use These Customer Service Prompts

Use these prompts by copying one, filling in the bracketed placeholders, and pasting it into ChatGPT, Claude, or Gemini. Each prompt already follows the Context/Task/Format framework, so you rarely need to add anything.

The Context/Task/Format framework has three parts. Context is who you are and the situation, for example "a support agent for a subscription meal-kit service handling a late delivery." Task is the specific thing you want done. Format is the shape of the answer, such as a short email, three macro options, or a bulleted summary. Our prompt engineering for business guide covers this anatomy in depth.

One rule comes first: describe the situation in general terms, not with real customer data. Do not paste a customer's name, email, order number, or account details into a public AI tool. The AI can write the wording; you add the specifics after.

Here is the information-gain point most listicles skip: the tone and situation context does more for quality than any clever wording. Watch what happens when we add it.

Before (weak): "Write a reply to a customer whose order is late." The output is a generic apology that could come from any company and names no next step.

After (strong): "You are a support agent for a small coffee-roasting business. A loyal customer's order is three days late because of a carrier delay. Write a short, warm email that apologizes without making excuses, explains the carrier delay in one line, tells them the new expected date is [date], and offers [gesture, for example free shipping on their next order]. Keep it under 90 words and sign off as [agent name]."

The second prompt names the situation, the tone, the facts to include, and the length. That is why it returns a reply you can almost send, instead of a draft you have to rewrite. When you need a prompt built to this standard for a job not covered here, our free AI prompt generator will assemble one for you.

Here is the failure mode most teams hit: auto-sending AI replies for account-specific issues. AI writes empathetic wording well, but it cannot see your systems, so it will state a refund amount, a policy, or an order status that sounds confident and is wrong. The win is a human-in-the-loop macro system, where AI drafts the tone and structure and a person adds the account facts, not full automation.

Reply and Response Template Prompts

These prompts turn a described situation into a ready-to-edit customer reply. They cover the everyday tickets that fill a support queue: delays, questions, and simple problems.

Describe the situation in general terms, fill in the brackets, then add the account-specific facts yourself before you send.

• You are a support agent for [describe business]. A customer wrote in about [describe issue, for example a product that arrived damaged]. Write a friendly reply that acknowledges the problem, states the next step we will take, and asks for [any detail you need, for example a photo]. Keep it under 100 words, use a warm and plain tone, and end with a clear sign-off.
• Act as a customer support writer for [describe business]. A customer asked [paste the general question, no personal data]. Write a clear, direct answer in plain language, break any steps into a short numbered list, and close by inviting them to reply if anything is still unclear. Keep it under 120 words.
• You are a support agent. A customer is confused about how [feature or policy] works. Explain it in simple terms a first-time user would understand, give one short example, and avoid jargon. Format the reply as two or three short paragraphs, not a wall of text.
• Act as a support agent for [describe business]. Write a reply for a customer reporting [common issue] where the fix is [describe fix]. Walk them through the fix in numbered steps, note what to do if it does not work, and keep the tone calm and reassuring. Leave a [bracketed placeholder] anywhere I need to add an account-specific detail.

Tone, Empathy, and De-escalation Prompts

These prompts help you respond to upset or frustrated customers without sounding scripted. They are for the moments when wording matters most.

AI is genuinely good at empathetic phrasing, which is exactly why a person still has to check the facts underneath it. A warm reply that promises the wrong thing does more harm than a plain one.

• You are a calm, experienced support agent. A customer is angry about [describe the general problem]. Write a reply that first acknowledges their frustration in one genuine sentence, avoids defensive language, states what we will do next, and gives a realistic timeframe. Keep it under 110 words and never blame the customer.
• Act as a de-escalation specialist. Rewrite this draft reply [paste your draft, no customer names] so it sounds more human and less scripted. Keep the same facts, soften any phrase that could read as dismissive, and lead with empathy before the solution. Return only the rewritten reply.
• You are a support agent replying to a customer who feels ignored after [describe situation, for example a slow first response]. Apologize for the wait without over-apologizing, take clear ownership of the next step, and reassure them a real person is now on it. Keep it warm, honest, and under 90 words.
• Act as a tone editor. Here is a reply I plan to send [paste draft]. Rate its tone for a frustrated customer on a scale of warm to cold, point out the two phrases most likely to escalate the situation, and rewrite each one. Do not change any factual claim in the message.

Canned Response and Macro Prompts

These prompts build reusable macros for the questions you answer over and over. Draft once, save to your help desk, and edit per ticket.

This is where a human-in-the-loop macro system pays off: the AI writes a clean, on-brand template, and your team fills the account-specific blanks each time instead of trusting the AI to know them.

• You are a support operations writer for [describe business]. Create a reusable macro for the common question [describe question, for example how to reset a password]. Write it so it works for any customer, use [bracketed placeholders] for details like name or order number, and keep the tone friendly and clear. Include a one-line opener and a short set of steps.
• Act as a help-desk template designer. Write three macro variations for [common request, for example a refund request] that share the same facts but differ in warmth: one brief, one standard, one extra-empathetic. Mark where the agent must insert account-specific details, and keep each under 100 words.
• You are a CX writer. Turn this long past reply [paste a general example, no customer data] into a tight, reusable macro. Remove anything one-off, replace specifics with [placeholders], and keep the parts that made it clear and kind. Return the macro and a short note on when to use it.
• Act as a support-content editor. Review this set of existing macros [paste macros] for consistency. Flag any that sound robotic or off-brand, suggest a shared sign-off, and rewrite the weakest one. Keep every factual placeholder exactly as it is.

FAQ and Help-Center Drafting Prompts

These prompts draft the self-serve content that deflects tickets before they reach your queue: FAQ answers, help-center articles, and step-by-step guides.

Have someone who knows the product check every draft. AI can describe a feature or policy that does not match how your system actually works.

• You are a help-center writer for [describe business]. Write a clear FAQ answer for the question [paste question]. Lead with the direct answer in the first sentence, then add any needed steps as a short numbered list. Keep it under 120 words and write at a plain, seventh-grade reading level.
• Act as a technical writer. Turn these rough notes on [feature or process] [paste notes] into a help-center article with a short intro, numbered steps, and a brief 'What to do if this does not work' section. Use simple language and avoid internal jargon a customer would not know.
• You are a content strategist for our support site. Based on this list of common tickets [paste general topics, no customer data], suggest ten FAQ questions we should publish, grouped by theme. For each, write a one-line summary of the answer so I can see the gaps in our current help center.

Ticket Summarization and Internal Note Prompts

These prompts compress long threads into summaries and handoff notes your team can act on fast. They save the most time on tickets that bounce between agents.

Strip personal details before you paste a thread, and treat the summary as a draft a teammate still reads, not a system of record.

• You are a support team lead. Summarize this ticket thread [paste the conversation with names and emails removed] in five lines or fewer: the core problem, what has been tried, the current status, and the single next step. Write it for a teammate picking this up cold.
• Act as a handoff assistant. Turn this thread [paste anonymized conversation] into an internal note for the next agent. Include the customer's goal, any promises we have already made, what is still open, and one thing to be careful about. Use short bullets and a neutral tone.
• You are a QA analyst. Read this resolved ticket [paste anonymized thread] and write a two-line internal summary plus one tag for the issue type, for example billing, shipping, or bug. Keep it factual and skip any praise or judgment of the agent.

Follow-up, CSAT, and Escalation Prompts

These prompts handle everything after the first reply: check-in messages, satisfaction surveys, and clean escalation notes. They close the loop so tickets do not go quiet.

Anything involving a refund, a complaint, or an account-specific promise should route to a person before it goes out. AI drafts the wording; a human confirms the commitment.

• You are a support agent for [describe business]. Write a short follow-up message to check whether the fix for [describe issue] worked. Keep it friendly and low-pressure, invite them to reply if anything is still broken, and keep it under 60 words.
• Act as a CX writer. Draft a one-question CSAT survey message that goes out after a resolved ticket for [describe business]. Ask for a rating in a warm, brief way, explain it takes ten seconds, and add one open line for extra feedback. Keep the whole message under 50 words.
• You are a support agent writing an internal escalation note. A ticket about [describe general issue] needs a manager or a specialist team. Summarize the problem, what the customer wants, what we have offered, and why it needs to go up, in four short bullets. Flag if a refund or account change is involved.
• Act as a retention-minded support agent. A customer hinted they may cancel over [describe general reason]. Write a reply that takes their concern seriously, addresses the specific point, and offers a next step without hard-selling. Leave a [placeholder] for any account-specific option a human should confirm first.

Putting These ChatGPT Prompts for Customer Service to Work

These ChatGPT prompts for customer service are starting points, and the context you add is what makes them yours. The moment you describe the real situation and your tone, the output stops sounding like a form letter.

In the AI workflow audits we run at Layer3 Labs, the support teams that get real value use AI to draft the tone and structure of a reply, then a person adds the account-specific facts before it goes out. The teams that try to auto-send AI replies for account-specific issues are the ones who end up cleaning up confident, wrong promises.

Start with one job this week, whether that is a set of macros, an FAQ refresh, or your ticket summaries. Run the prompt, edit the result, and note what you changed. That edit is a clue for how to sharpen the next prompt.

For more copy-paste libraries, see our ChatGPT prompts for business and ChatGPT prompts for small business. To go deeper on writing your own, build a custom prompt with the AI prompt generator. If you want help turning these prompts into a repeatable support workflow instead of copy-pasting them by hand, our team can map it to your systems.