ChatGPT Prompt Library: 50 Ready-to-Use Templates for Work

This ChatGPT prompt library is for you if you want to use AI for real work, not just a few test questions for fun. Many marketing, sales and operations teams are losing time because their instructions are too generic, and they get back long answers that are hard to use. MONA wrote this article as a set of practical templates, with an input formula, context, examples and ways to check the output. We also add an internal rollout view, so you can see when ChatGPT is enough and when your business needs its own software to keep data connected.

Get to work in 10 minutes: a ChatGPT prompt library that turns vague requests into usable output

You do not need to relearn all of ChatGPT to start using it. What matters is picking the right task group, swapping in the right variables, and asking ChatGPT to return the right format.

Straight to the point. We often use this method inside MONA when we need content ideas, service descriptions, sales emails or document summaries.

Pick the right task group before writing the prompt

A good prompt starts with a clear goal. If the goal is vague, ChatGPT will answer in generalities that sound right but are hard to use in real work.

Pick a task group before you type the prompt. This step helps ChatGPT understand whether it is supporting writing, analysis, sales or operations.

• Content group: write SEO articles, social posts, emails, product descriptions, short video scripts.
• Sales group: write consultation messages, reply to customers, handle objections, summarize customer needs.
• Management group: summarize meetings, build to-do lists, write internal procedures.
• Analysis group: analyze competitors, sort customer feedback, pull key points from long documents.
• Website and SEO: write titles, descriptions, article outlines, keywords and service page content.

Pick the group first. Write the prompt second.

For example, with the same request, “write content for an SEO service”, the result will be very different if you choose the SEO group instead of the sales group. MONA always separates the goal first before using ChatGPT, to avoid heavy rewriting.

Replace 5 required variables so the prompt is not generic

A sample prompt is only worth using when you replace the right variables. If you keep the sample as is, the output looks more like reference material than usable content.

We recommend that you always replace these 5 variables before you hit send.

• Role: ask ChatGPT to play an SEO specialist, a customer support agent, a sales manager or an editor.
• Context: describe the industry, product, target customers, channel and the situation you are handling.
• Goal: state clearly whether you need to increase orders, explain a service, book an appointment or clarify an issue.
• Tone of voice: choose a professional, friendly, benefit-first or brand-fit tone.
• Output format: ask for a table, bullet list, email, script or outline.

The short formula looks like this.

Act as [role]. Write [content type] for [context]. The goal is [goal]. The tone is [tone]. Return the result as [format].

A real example for marketing.

Act as an SEO specialist. Write a blog outline for website design service businesses. The goal is to help business owners understand why a website needs to generate revenue. The tone is clear and practical. Return the result as H2, H3 and notes on the main points.

This prompt is short enough. But it has a structure.

Use one prompt template for 3 different output versions

A prompt should not produce just one result. You can use the same frame to get 3 versions, then pick the one that fits the channel.

This saves editing time. It also helps the content team see the direction before writing in detail.

• Short version: for messages, quick descriptions, titles and opening paragraphs.
• Standard version: for blog posts, emails, service page content or consulting documents.
• Conversion version: for ads, landing pages, sales scripts or calls to action.

You can add this line at the end of the prompt.

Create 3 versions: a short one under 80 words, a standard one of about 300 words, and a conversion-focused one with a clear call to action.

An example for a sales team.

Act as a website consultant. Write a reply to a customer asking about website design pricing. The context is that the customer does not yet understand the difference between a template website and a custom website. The tone is clear and not pushy. Create 3 versions: short, standard, and one aimed at booking a consultation.

The results will be easier to use. The team will also find it easier to stay consistent.

In website, SEO and marketing projects, MONA uses ChatGPT as a tool for drafting, sorting ideas and speeding up preparation. The decisions still rest on strategy, business data and how things are actually rolled out.

ChatGPT is handy for personal use and one-off tasks. When a business needs connected data, automated processes and security control, you need custom software built by MONA around how your operation really runs.

A 4-layer formula that makes ChatGPT understand the job, the role and the tone

Weak prompts make ChatGPT ramble, use the wrong tone and leave out the usable parts. The 4-layer formula helps you make the AI clearly understand the role, context, task and output standard.

We use this formula inside MONA when writing content, analyzing SEO, drafting sales scripts and standardizing customer support replies. The method is not complicated. What matters is writing every layer.

Layer 1: give a clear professional role

ChatGPT needs to know who it is playing before it handles the job. The clearer the role, the less generic the answer.

Do not write: “Write an ad”. That is too broad. The AI will guess.

Write it in this pattern:

• Act as an SEO expert with experience building service websites in Vietnam.
• Act as a marketing manager in charge of generating leads for an online course.
• Act as a customer support agent handling a complaint about late delivery.
• Act as an operations manager who needs to write a procedure for new hires.

The role layer helps ChatGPT choose the right point of view. This matters a lot. An answer for a CEO is different from an answer for a frontline employee.

Layer 2: give enough context so the AI does not guess

Context is the part many people skip. Without context, ChatGPT often answers the words correctly but misses the job.

Give the background details that affect the result. It does not need to be long. It just needs to be enough.

• Industry: spa, education, interior design, software, F&B or e-commerce.
• Audience: business owners, young mothers, students, branch managers or sales staff.
• Channel: website, Facebook, Zalo, email, landing page or phone call script.
• Goal: increase sign-ups, reduce repeated questions, keep existing customers or book consultations.
• Constraints: professional tone, no exaggeration, no overly technical words.

A better example prompt:

“Act as an SEO expert. I am writing for a website design service in Ho Chi Minh City. The reader is a mid-sized business owner who needs a website that brings in leads. Write a short, clear opening paragraph, without an over-the-top advertising tone.”

This has enough background. ChatGPT guesses less. The result is usually closer to what you need from the first try.

Layer 3: lock the task with a specific verb

A vague task produces a vague result. Use a specific verb so ChatGPT knows what to do.

Instead of “handle this content”, state the action clearly. Each prompt should have only one main task.

• Rewrite this paragraph in an easier-to-understand tone.
• Summarize the content into 5 key points for the sales team.
• Turn the product description into a consultation script for Zalo.
• Write 3 SEO titles that include the main keyword.
• Sort customer feedback into urgent items and items that can be followed up later.

The verb is the steering wheel. The clearer it is, the easier it is to control. MONA often splits one big job into several small prompts to reduce errors.

For marketing, for example, you can follow this chain:

• Step 1: summarize the customer profile.
• Step 2: write 5 content angles.
• Step 3: pick 1 angle and write a draft.
• Step 4: review the brand tone.
• Step 5: turn it into a Facebook description and an SEO article.

This is a little slower. But it is easier to control. The marketing team also finds it easier to learn.

Layer 4: set the output standard before you get the result

The output standard is a quality fence. If you do not set it first, ChatGPT tends to give long, even answers that are hard to use right away.

State the length, structure, tone and what to avoid. This is the layer that cuts down on manual edits.

• Length: 120 words, 3 paragraphs, each with no more than 2 sentences.
• Structure: open the problem, explain, give an example, close with an action.
• Tone: clear, close to Vietnamese business owners, no clichés.
• Prohibited: no exaggeration, no rhetorical questions, no overly technical words.
• Format: return a table, bullet list or a complete paragraph.

A prompt with all 4 layers can be written like this:

“Act as a marketing expert for service businesses. The context is a website design company that wants to write content for business owners who need more customers from Google. The task is to write 5 blog titles about website SEO. The output standard: titles under 65 characters, a clear benefit, no sensationalism, no clichés.”

You can change the context to use this for sales, customer support or internal operations. The frame stays the same. Only the job changes.

• Marketing: write titles, outlines, service descriptions and social content.
• Sales: write consultation scripts, handle objections, summarize customer needs.
• Customer support: write polite replies, group errors, write service guides.
• Operations: build procedures, checklists and training content for new staff.

ChatGPT handles one-off tasks very fast. But it has a ceiling. When data is scattered and each person uses it differently, the results are still hard to control.

For businesses that need connected data, security and real automation, MONA usually recommends building custom software. We use AI internally, and we also train customers to apply it properly across websites, SEO, CRM, ERP and operating processes.

50 prompt templates in 7 task groups, ready to edit and use right away

This section is a ready-to-use prompt library for websites, SEO, marketing, sales and operations. You only need to replace the text in square brackets with your real data.

We wrote it the way MONA does in internal training and customer guidance. The goal is clear. Get work done the same day.

Prompts for website content and SEO

Websites rarely lack words. They usually lack the right search intent. This group of prompts helps you get a structured draft, then refine it with your professional experience.

• 1. Write an SEO outline for the topic [topic], the readers are [customer group], and the business goal is [business goal].
• 2. Write an intro under 80 words for the article [article name], in a clear tone, without exaggeration, mentioning [main keyword].
• 3. Create 10 SEO titles for the keyword [keyword], each with one specific benefit.
• 4. Write a description of the service [service] for the website, focusing on the problem [pain point] and the result [desired result].
• 5. Write 5 H2s for the article [topic], each handling a different search intent.
• 6. Write an FAQ section for the article [article name], with 6 short, easy-to-understand answers, without a sales tone.
• 7. Improve the following paragraph so it reads more easily, keeping the main idea and adding the keyword [keyword] naturally: [paragraph].
• 8. Suggest a structure for the service page [service], including hero, benefits, process, proof and call to action.

Prompts for ads, emails and social posts

Ad copy easily sounds all the same if staff only change a few words. This group pushes the AI tool to change the angle, the promise and the way it opens.

• 9. Write 5 ad variations for the product [product], each aimed at a different pain point of [customer].
• 10. Write 3 follow-up emails to customers who left their details about [service], in a consulting tone, not pushy.
• 11. Write 10 social post openers for the topic [topic], each under 25 words and with a specific angle.
• 12. Turn the following content into a 150-word Facebook post, in an expert but easy-to-understand tone, without clichés: [content].
• 13. Write a 7-day content calendar for [industry], with a topic, an angle and a call to action for each day.
• 14. Write 5 short description versions for the video [topic], each suited to a different customer group.
• 15. Shorten the following email to 120 words, keeping the polite tone and the main offer: [email].

Prompts for sales and customer care

Sales and customer care need wording that fits the context. One wrong sentence and customers read it as dodging or pushing.

This group is for messages, call scripts and handling feedback. After using it, a person still needs to approve.

• 16. Write a consultation script for a customer interested in [service], including greeting, discovering needs, advising and setting the next step.
• 17. Write 5 replies for when a customer says the price is high, in a calm tone, explaining by the value they receive.
• 18. Write a reminder message for a customer about [type of appointment], short and polite, with the time and place.
• 19. Sort the following list of leads into hot, warm and cold, and suggest how to handle each group: [list].
• 20. Write a reply template for when a customer complains about [issue], accepting the right level of responsibility and giving a clear plan.
• 21. Create a question list to discover needs for a customer buying [product], with 10 questions in an easy-to-answer order.
• 22. Write a message to re-engage a past customer after [time period], recalling the earlier value and suggesting a fitting next step.

Prompts for document analysis and meeting summaries

Long documents make teams spend time rereading. Long meetings make it easy to lose important items.

This is well worth using. You only need to give clean data, then ask for a structured output.

• 23. Summarize the following document into 5 key points, 5 action items and 5 risks to note: [document].
• 24. Turn the following meeting minutes into a table with action items, owner, deadline and notes: [minutes].
• 25. Read the following content and point out contradictions, missing data or items that need further verification: [content].
• 26. Summarize the following report for a busy CEO, 200 words at most, focusing on the decisions to be made: [report].
• 27. Create action notes after the meeting, split by department, priority and deadline: [notes].
• 28. Turn the following technical document into an explanation for marketing staff, keeping examples easy to understand: [document].
• 29. Create a list of questions to ask a partner after reading this proposal: [proposal].

Prompts for hiring, training and internal procedures

Hiring and training take effort because everyone writes in their own way. Good prompts help standardize the framework before managers approve.

• 30. Write a job description for the position [position], including duties, requirements, performance metrics and benefits.
• 31. Create a set of interview questions for [position], split by expertise, attitude, situations and teamwork.
• 32. Write a 14-day training plan for a new hire in the position [position], with a goal and a practice task each day.
• 33. Turn the following procedure into a step-by-step guide for new staff, using short sentences and clear tasks: [procedure].
• 34. Create a probation evaluation sheet for [position], with criteria, descriptions of pass and fail, and manager notes.
• 35. Write an internal reminder about [topic], in a clear, respectful tone, without excessive pressure.
• 36. Standardize the following rule into an internal document, split into sections, with correct and incorrect examples: [rule].

Prompts for product ideas and rollout plans

Many ideas do not produce revenue without a validation step. This group of prompts helps you pull ideas down into concrete work.

MONA often uses this when sketching websites, SEO and custom software modules. Clarify first. Code later.

• 37. Analyze the idea [idea] by target customer, the problem to solve, benefits and rollout risks.
• 38. Create a 30-day rollout plan for [project], with main tasks, owners, checkpoints and expected results.
• 39. Suggest 10 features for the product [product], split into must-have, should-have and later.
• 40. Write a requirements description for the feature [feature], including the goal, user flow, input data and output data.
• 41. Create a comparison table of 3 rollout options for [project], by cost, time, risk and fit.
• 42. Write a launch plan for the service [service], including website content, SEO, ads, email and customer care.
• 43. Point out the risky assumptions in the following plan and suggest simple ways to test them: [plan].

Prompts for error checking, shortening and rewriting

This is the group of prompts we use most when we need to fix a draft fast. It does not replace an editor, but it noticeably cuts down on mechanical edits.

Do not give a generic command. Give specific criteria.

• 44. Check the following paragraph for spelling errors, long sentences and repeated ideas, and suggest a tighter version: [paragraph].
• 45. Rewrite the following content in a professional, easy-to-understand tone, without adding new information: [content].
• 46. Shorten the following paragraph by 50%, keeping the main argument and removing repeated sentences: [paragraph].
• 47. Turn the following content into slide bullets, each point under 12 words: [content].
• 48. Check the following article for exaggeration, unsupported claims or anything easy to misread: [article].
• 49. Rewrite the following paragraph to fit the brand tone [tone description], keeping the important information: [paragraph].
• 50. Grade your previous answer on 5 criteria: meets the request, clear, enough data, easy to read and ready to use.

These 50 prompts help you work faster at the individual and small-team level. But scattered tools still have a ceiling, because data is spread out and processes are still done by hand.

When a business needs connected data, clear permissions and automation tied to CRM, ERP or the website, custom software is a better fit. MONA builds software, designs websites and does SEO on top of a real operating foundation, not just a few one-off commands.

Command with sample data: a trick that makes answers less generic after a single paste

Content written by ChatGPT is often generic when the input is only a few lines of request. The less data you give, the more likely the tool is to return mass-produced wording.

We find the quick fix is sample data. Instead of just saying “write a sales post”, add old samples, a product table, a good example and a bad example.

MONA usually teaches internal teams and customers to input data in 4 layers. The tone layer, the information layer, the output standard layer, and the rules layer drawn from the samples.

Paste old content samples to keep a familiar tone

A business’s tone usually lives in the posts that have already worked well. Paste 2 to 3 old paragraphs before asking for new writing.

The samples do not need to be long. They only need to be enough for ChatGPT to pick up the way you address readers, sentence length, how you name products and how you close a point.

• Paste 1 service introduction already used on the website.
• Paste 1 Facebook post with solid engagement.
• Paste 1 sales email your sales team still uses.
• State clearly: “keep a tone close to the samples, do not copy the wording”.

A short command can be written like this: “Read the 3 samples below, work out the writing tone, then rewrite the new content in the same style”.

Give a product information table before asking it to write

ChatGPT cannot know by itself how your product differs from competitors. Without an information table, the answer easily looks like every ad on the market.

We usually recommend that you enter data as a short table. The clearer the columns, the closer the answer gets to the work of selling, SEO and customer care.

• Product or service name.
• Suitable customers.
• The customer’s main pain point.
• Real benefits after using it.
• Differences that can be backed with proof.
• What must not be overstated.

For websites, this table keeps the content on target for the service page. For SEO, it keeps the article from drifting into general knowledge.

Attach a good output example and a bad output example

A good example helps ChatGPT understand the standard to meet. A bad example helps the tool avoid the kinds of sentences you do not want to see.

MONA often uses this pair of examples when writing service descriptions, SEO articles, consultation scripts and ad content. It cuts many rounds of gut-feeling edits.

• Good example: short sentences, with facts, no flattery.
• Bad example: long sentences, lots of slogans, no specific information.
• Good example: states clearly what the customer gets after doing it.
• Bad example: only says “elevate your brand” with no context.

You should add the command: “If your answer resembles the bad example, fix it yourself before sending the final result”.

Ask ChatGPT to draw rules out of the sample data

One trick worth using is to make ChatGPT analyze the samples before it writes. When the tool states the rules itself, the results afterwards tend to fit the context better.

You can ask the tool to list 5 rules. Only then ask it to write the article, email, product description or SEO content.

• Step 1: paste old content samples.
• Step 2: ask it to draw out rules on tone, structure and words to avoid.
• Step 3: check that the rules fit your business.
• Step 4: ask it to write new content following that set of rules.
• Step 5: ask it to review and fix its own work before returning the result.

MONA applies this in many internal jobs, from summarizing ideas to sketching website content. Since 2016, MONA has delivered over 14,000 projects, so we put real data ahead of polished wording.

ChatGPT helps you finish personal tasks faster and spend less time on rough drafts. But when data is scattered and staff use it unevenly, the results hit a ceiling.

A business that wants seamless automation needs custom software for CRM, ERP or an internal automation chain. MONA can use AI as one part of the system, but the core is still software designed around the business’s real data.

Make the AI grade its own work: a 3-round revision loop that sharpens results without rewriting from scratch

ChatGPT’s first draft usually sounds fine, but it easily loses focus, gets the tone wrong, lacks evidence or repeats itself. This is a very common problem when you use AI for SEO articles, emails, sales scripts and internal documents.

Do not rewrite from scratch. Make ChatGPT grade itself, spot its own errors, then fix them against clear criteria. This lets you keep the main idea while raising the quality of the final version faster.

MONA uses this 3-round process when training internal teams and customers who use AI tools for marketing. The goal is not for the AI to write everything, but to cut manual revision rounds and control quality better.

Round 1: ask ChatGPT to grade against a scale of criteria

Before fixing anything, make ChatGPT play the grader. The AI has to look at the draft against specific criteria and is not allowed to give vague praise.

Use this command template:

• Grade the content below on a 10-point scale.
• The criteria are: clear idea, right goal, right tone, easy to read, persuasive.
• For each criterion, give a score and a short explanation.
• Do not rewrite the content in this step.

The key is in the last line. You do not let the AI fix it right away, because fixing too early often loses the original idea.

Grade first.

Round 2: make the AI point out weak sentences and why they are weak

Once you have the scoring table, ask ChatGPT to examine each weak sentence. This step exposes the parts of the content that make readers lose focus.

Keep using this command template:

• List the weak sentences in the draft.
• For each sentence, state why it is weak.
• Group the errors by type: vague, wordy, missing benefit, wrong tone, missing evidence.
• Do not rewrite in this step.

At this point, you will see where the draft is weak. A sentence can be grammatically correct and still have no selling power or reading pull.

This is the part worth paying for.

For SEO articles, the error is usually an opening line that is too generic. For emails, it is usually the call to action. For internal documents, it is usually instructions with missing steps.

Round 3: ask it to rewrite against the criteria it just graded

Once ChatGPT has pointed out the errors itself, then let it rewrite. The revision now follows the criteria more closely, because the AI already has an evaluation frame from before.

Use this command template:

• Rewrite the draft based on the errors just identified.
• Keep the main goal and the important information.
• Prefer short sentences, clear ideas and few filler words.
• The tone is for Vietnamese business owners and marketing staff.
• After the rewrite, list 5 points that were improved.

Do not drop the line “keep the important information”. Without it, ChatGPT may make the content smoother but lose the facts you need to keep.

Keep the facts first.

You can apply the same loop to many types of content:

• SEO articles need a clear idea, the right search intent and no rambling.
• Sales scripts need to hit the pain point, have a lead-in and a closing line.
• Emails need to be tight, on goal and not sound like a form letter.
• Internal documents need to be easy to follow, with few gaps left to interpretation.

Lock the final version with a checklist before using it for real

The 3-round revision loop sharpens a draft, but you still need a final check. AI helps a lot, but it cannot replace the responsibility of approving content.

Use this checklist before publishing or sending:

• Does the content match the original goal.
• No invented information or unsourced figures.
• The tone fits the brand.
• Important sentences have not been softened too much by the AI.
• The call to action is clear and natural.
• No sensitive internal data is exposed.

MONA usually guides customers to use ChatGPT with a frame, criteria and an approval step. This suits marketing, sales and operations teams that need to move fast but still stay in control.

Still, standalone tools always have a ceiling. When data is scattered, staff use it unevenly and processes still depend on manual steps, a business needs its own software for real automation.

For CRM, ERP, marketing automation chains, web design and SEO tied to revenue, MONA usually combines internal AI with software we build ourselves. At that point, AI does not just answer in a chat box, it enters operating processes with data and clear permissions.

Turn ChatGPT prompt library into an internal handbook instead of a file saved for fun

Instead of saving ChatGPT prompts as scattered documents, you need to organize them into an internal handbook shared by the marketing, sales and operations teams. This creates an asset that runs across the company and helps everyone understand and use prompts more effectively.

Name prompts by the job to be done, not by inspiration

Naming prompts clearly by purpose makes them easier for you and your staff to find and use. Instead of inspirational names, choose names that reflect the specific job.

Record the user, the input data and the output standard

Each prompt should come with information about the user, the input data and the expected output standard. This helps everyone understand the context and use it for the right goal when applying the prompt.

Create prompt versions for new staff and for experienced staff

Split prompts into two versions: one for new staff with detailed guidance, and one for experienced staff to save time. This makes the most of each group’s abilities.

Schedule reviews so prompts do not go stale

Set a regular schedule to review and update prompts. This keeps the library in line with real practice and keeps it from going out of date. Do not forget to update when leadership sets a new strategy or direction.

The ceiling of one-off prompts: work moves faster, but your data is still not connected

Staff who use prompts well still have to copy data by hand

Staff who run prompts in ChatGPT usually only solve their own individual tasks. They still have to copy data by hand between systems. This makes work easy to interrupt and wastes unnecessary time.

Scattered data makes it hard for managers to control quality

When data is not connected, managers find it hard to control and assess the quality of work. Scattered data makes analysis and reporting complicated and slow. As a result, information is not updated in time, which affects business decisions.

Custom software keeps processes, permissions and data moving together

MONA develops custom software designed around each business’s needs. These systems keep processes seamless, permissions clear and data integrated. This keeps information secure and continuously updated, which improves performance and work quality.

MONA puts AI into the system when a business needs real operations

Whenever a business needs stable operations, MONA uses AI inside the software system. Running fast is one thing, running right is what pays off. With over 14,000 projects delivered and 85% of customers coming back, MONA stays with you for sustainable growth.