AI for Marketing: ChatGPT Prompts With Real Output

Marketers use AI to draft and rework copy: launch posts, email subject lines, brand-voice rewrites, repurposed blog content, SEO outlines and plain-language summaries of campaign results. The key is to paste a short product fact sheet and a brand voice card into every prompt, so the AI writes from your facts instead of guessing. Every claim about the product still gets checked against the source before it is published.

• Give it a fact sheet. Without one, the AI invents features, numbers and customer results.

• Give it your voice. A short voice card and two real examples beat adjectives like "friendly".

• Check claims, not just grammar. Numbers, features, prices, trial lengths and "trusted by" lines.

• AI suggests, your audience decides. It can write variants for a test. It can't tell you which one will win.

• Results vary by tool. ChatGPT, Claude, Gemini and Copilot word things differently, but these prompts work in all of them.

A chat assistant writes the most likely text for the words in front of it. That makes it fast and fluent, and it makes it a poor source of facts. It doesn't know your product, your customers or this week's search results unless you paste them in or connect it to them.

For the general prompt method behind these examples, see the pillar guide AI prompts for work.

Each task has a prompt you can copy, the model's real answer (trimmed with "…") and what to check. The answers come from gpt-5-mini, a small OpenAI model. The product, Fieldnote, a scheduling app for plumbers and electricians, is made up, and so are its numbers.

1. Write launch posts from a fact sheet

Most people start like this:

Write a LinkedIn post announcing our new appointment reminder feature.

Version 2 — Product-Focused
Announcing Appointment Reminders — a simple, powerful way to cut no-shows and free your team from repetitive outreach.

Key benefits:
- Automated reminders via email and SMS
- Customize message copy and timing (24 hours, 2 hours, etc.)
- Two-way confirmations and rescheduling links
- Syncs with Google/Outlook calendars and existing booking flows

Results you can expect: fewer missed appointments, better staff utilization, and improved customer experience. We rolled this out to beta customers and saw a 40% drop in no-shows — now it’s available to all users. …

Email reminders, calendar sync and a 40% drop in no-shows: none of it is real. The model filled the gaps with a typical product. Give it the facts:

Task: Write 3 versions of a LinkedIn post announcing a new feature, plus 3 email subject lines for the same announcement.

Product facts:
- Fieldnote is a scheduling and invoicing app for home-service businesses (plumbers, electricians, HVAC) with 1 to 20 technicians.
- New feature, live September 9: automatic text reminders to customers 24 hours and 2 hours before an appointment. Customers can reply C to confirm or R to reschedule.
- Included in the Pro plan ($49 per month for up to 5 users).
- In our beta with 120 businesses (June to August), missed appointments fell from 8.1% to 5.2%.
- Available in the US and Canada.
- 14-day free trial, no credit card.

Audience: Owners and office managers of small home-service businesses.

Brand voice: Plain and practical, a little warm. Sound like an experienced dispatcher, not a tech company. No exclamation marks. Never use "revolutionary", "game-changer" or "seamless".

Format: Each post under 90 words, with a different angle: (1) the problem, (2) the beta result, (3) how it works. One call to action per post. No hashtags.

Rules: Use only the facts above. Don't add numbers, customer names, quotes or features.

Post 2 — The beta result
We tested these automatic text reminders with 120 businesses from June to August. Missed appointments dropped from 8.1% to 5.2%. That’s why we’re rolling the feature out on Sept 9. It’s part of the Pro plan ($49/month for up to 5 users) and available in the US and Canada. See if it tightens your schedule — 14-day free trial, no credit card.

Post 3 — How it works
Fieldnote will automatically text customers 24 hours and 2 hours before an appointment. Customers reply C to confirm or R to reschedule, and your office gets the update. …

Email subject line options
- New: automatic appointment texts (launching Sept 9)
- Cut no-shows: reminders that let customers reply C or R
…

What to check: The numbers, price and trial are right. One feature slipped in anyway: "your office gets the update" isn't in the facts. It may be true, so ask the product team. All three posts also repeat the price and the trial, which makes them read like ads; cut that to one post. Run "Cut no-shows" past whoever approves claims at your company.

2. Rewrite copy in your brand voice

Task: Rewrite the draft below in our brand voice.

Brand voice card:
- We sound like an experienced dispatcher: calm, practical, a little warm.
- Short sentences. Concrete details over adjectives.
- Talk about the customer's day (fewer callbacks, fewer empty driveways), not about our technology.
- No exclamation marks. Never use "revolutionary", "game-changer", "seamless", "leverage" or "empower".

Two examples of our voice:
"Your techs show up. Your customers are home. That's the whole idea."
"Invoices go out from the truck, so you're not doing paperwork at 9 p.m."

Draft to rewrite:
"""
Introducing Smart Reminders! Our revolutionary AI-powered reminder engine leverages cutting-edge automation to empower your business and deliver a seamless customer experience. Say goodbye to no-shows forever!
"""

Format: The rewrite (under 50 words), then a list of what you changed and why, one line each.

Rules: Keep the meaning, but don't promise results. The feature sends text reminders 24 hours and 2 hours before an appointment. Don't add other facts.

Smart Reminders sends text messages 24 hours and 2 hours before appointments. Reminders run automatically. Fewer callbacks. Fewer empty driveways. Your team spends less time chasing customers and more time on the job.

Changes and why:
- Replaced marketing hype with a clear feature sentence that states when reminders are sent.
- Removed banned buzzwords and exclamation to match voice guidelines.
…
- Added a practical benefit line about your team's time to show impact without promising outcomes.

What to check: The voice is much closer. But "Fewer callbacks. Fewer empty driveways." and "less time chasing customers" are promises of results, which the prompt ruled out. The change list even says it didn't promise outcomes. So check the copy itself, not the AI's summary of what it changed. Two real examples of your voice do more than a list of adjectives. The guide to few-shot prompting explains why.

3. Repurpose a blog post

Task: Turn the blog post below into a newsletter blurb and two social posts.

Format:
1. Newsletter blurb: under 60 words, ends with "Read the full post".
2. LinkedIn post: under 80 words.
3. Facebook post: under 50 words, friendlier.

Rules: Keep every number exactly as it is in the post. Don't add tips or facts that aren't in the post. No hashtags, no exclamation marks.

Blog post:
"""
Four ways to cut no-shows without nagging your customers

A missed appointment costs more than the job. Your tech drives out, waits, calls, and drives back. For most shops that is two hours gone.

1. Confirm the day before, and again on the day. One reminder helps. Two, at the right times, help more. We suggest 24 hours and 2 hours before.
2. Make rescheduling easy. Customers who can't make it often just don't answer. If they can reply with one letter to move the visit, many will.
3. Give an arrival window, not a time. "Between 1 and 3" is honest, and customers plan around it.
4. Call back new customers. First-time customers miss more appointments than regulars. A short call after booking builds trust.

In our beta with 120 businesses, missed appointments fell from 8.1% to 5.2% after shops turned on two reminders.
"""

Newsletter blurb:
Cut no-shows without nagging: confirm 24 hours and 2 hours before, let customers reschedule with one reply, give an arrival window, and call new customers. In a beta with 120 businesses, missed appointments fell from 8.1% to 5.2% after turning on two reminders. Read the full post

LinkedIn post:
A missed appointment costs more than the job — often two hours lost. …

Facebook post:
Missed appointments can cost shops two hours. Try confirming 24 hours and 2 hours before, allow one-letter reschedules, give an arrival window, and call new customers. In a 120-business beta, missed appointments dropped from 8.1% to 5.2%.

What to check: Every number survived. Now look at the source of the claim. The post says "our beta", while the blurb says "a beta". Dropping one word makes your own test sound like independent research, and small wording shifts like that are where claims go wrong. The Facebook post also isn't friendlier than the others. Ask for a second try rather than fixing it by hand.

4. Find customer language in reviews

Task: Find what our customers value most, in their own words, so we can use their language in our marketing.

Format:
1. A table: Theme | Number of reviews | Exact phrases customers used
2. The 5 most useful exact phrases for headlines.
3. One theme that is a complaint, and how many reviews mention it.

Rules: A review can count toward more than one theme. Copy phrases exactly. Use only these reviews.

Reviews (anonymous, from our app store page):
"""
1. "Invoices from the truck changed my life. No more Sunday paperwork."
2. "My guys finally know where they're going each morning."
3. "The reminders mean fewer empty driveways. Worth it for that alone."
4. "Setup took an afternoon. Support answered every question the same day."
5. "Customers like getting a text before we show up."
6. "Wish it worked offline. Basements kill my signal."
7. "Getting paid faster since we send invoices on site."
8. "Simple enough that my 60-year-old office manager loves it."
9. "Offline mode please. Half my jobs have no service."
10. "Fewer no-shows since we turned on reminders."
11. "Dispatch board is clear. I can see the whole day at a glance."
12. "Support is real people who know the trades."
"""

1) Table: Theme | Number of reviews | Exact phrases customers used

- Invoicing / getting paid faster | 2 | "Invoices from the truck changed my life. No more Sunday paperwork." ; "Getting paid faster since we send invoices on site."
- Dispatch / route visibility | 2 | …
- Reminders / fewer no-shows | 2 | "The reminders mean fewer empty driveways. Worth it for that alone." ; "Fewer no-shows since we turned on reminders."
- Setup & Support responsiveness | 2 | …
- Customer notifications / texts | 1 | "Customers like getting a text before we show up."
…
- Offline / connectivity (complaint) | 2 | "Wish it worked offline. Basements kill my signal." ; "Offline mode please. Half my jobs have no service."

What to check: The phrases are exact, which is the hard part. The grouping is a judgment call: review 5 is about the reminders too, so that theme really has three reviews. Read the reviews yourself before you trust the counts. Phrases like "fewer empty driveways" are gold for headlines. Before you quote a customer in an ad, follow your company's rules for using reviews.

5. Plan an SEO article

Task: Plan a blog post for the search "how to reduce no-shows for home service businesses".

What we know (from our own experience):
- A no-show costs a tech about two hours of driving and waiting.
- Two reminders (24 hours and 2 hours before) work better than one.
- Easy rescheduling by text reduces silent no-shows.
- Arrival windows ("between 1 and 3") set honest expectations.
- First-time customers miss more appointments than regulars.
- Our beta: missed appointments fell from 8.1% to 5.2% across 120 businesses.

Format:
1. A title tag under 60 characters.
2. A meta description between 140 and 155 characters.
3. An outline: H2s and one line under each on what it covers.
4. Three questions the post should answer in an FAQ.

Rules: Use only what we know. Don't invent statistics or studies. Where a section needs a fact we haven't given, write "[need source]".

Title tag (under 60 chars)
Reduce No-Shows for Home Service Businesses

Meta description (140–155 characters)
Cut no-shows with 24h + 2h reminders, easy text rescheduling, clear arrival windows, and first-time customer follow-up — beta: missed appts fell 8.1% → 5.2%.

Outline (H2s with one-line descriptions)

H2: Why no-shows matter for home service businesses
Covers the real cost: a typical no-show costs a tech roughly two hours of driving and waiting, plus lost revenue and scheduling disruption.
…
H2: Metrics to track and measure success
Lists the key metrics to monitor … [need source] if external benchmark comparisons are required.
…

What to check: The meta description is 157 characters, two over the limit the prompt set. Models can't count characters reliably, so count them yourself (=LEN() in a spreadsheet does it). "Plus lost revenue" wasn't in your notes. The "[need source]" markers worked: they show where the article needs your own data. The AI can't see search volumes or what ranks today, so look at the real search results before you write.

6. Summarize A/B test results

Task: Summarize the results of our September email test for the marketing team.

Context: We sent two versions of the launch email for appointment reminders. The only difference was the subject line.

Results:
"""
Version,Subject line,Sent,Opened,Clicked,Trial signups
A,"New: text reminders that cut no-shows",4200,1218,214,31
B,"Fewer empty driveways, starting today",4200,1386,189,24
"""

Format: First the answer in one sentence: which version did better, and on which measure. Then a table: Version | Open rate | Click rate (clicks / sent) | Signup rate (signups / sent). Then at most 3 bullets on what to test next.

Rules: Show your math for each rate. Don't say a difference is significant unless you show how you know. Use only this data.

Version B got a higher open rate, while Version A produced higher click and trial-signup rates.

Version | Open rate | Click rate (clicks / sent) | Signup rate (signups / sent)
---|---:|---:|---:
A | 1218 / 4200 = 0.2900 = 29.00% | 214 / 4200 = 0.05095 = 5.10% | 31 / 4200 = 0.007381 = 0.74%
B | 1386 / 4200 = 0.3300 = 33.00% | 189 / 4200 = 0.04500 = 4.50% | 24 / 4200 = 0.005714 = 0.57%
…

What to check: The math is right; we checked every rate. The first sentence is the real insight: the subject line that won opens lost signups. "Show your math" lets you check the numbers in a minute, and "don't say significant" stopped the model from calling a winner on 31 versus 24 signups. Decide winners with your team's usual test rules, not the AI's opinion.

7. Check a draft's claims against your product facts

Task: Check the landing page draft below against our product facts. List every claim a reader could check.

Product facts:
- Fieldnote is a scheduling and invoicing app for home-service businesses with 1 to 20 technicians.
- Text reminders go out 24 hours and 2 hours before an appointment. Customers can reply C to confirm or R to reschedule.
- Pro plan: $49 per month for up to 5 users.
- In our beta with 120 businesses (June to August), missed appointments fell from 8.1% to 5.2%.
- Available in the US and Canada.
- 14-day free trial, no credit card.

Landing page draft:
"""
Cut no-shows in half with Fieldnote reminders.
Trusted by more than 10,000 home-service pros.
Reminders go out automatically, 24 hours and 2 hours before every visit.
Customers confirm or reschedule with one text.
Works anywhere in the world.
Start your free 30-day trial today.
"""

Format: A table: Claim | Supported by the facts? (Yes / No / Partly) | Why | Suggested fix. Then one line: how many claims need a fix.

Rules: Use only the product facts. If a claim needs a source we haven't given, say so.

Claim | Supported by the facts? (Yes / No / Partly) | Why | Suggested fix
--- | --- | --- | ---
Cut no-shows in half with Fieldnote reminders. | Partly | Our beta (120 businesses, June–Aug) showed missed appointments fell from 8.1% to 5.2% — a ~36% reduction, not 50%. | …
Trusted by more than 10,000 home-service pros. | No | We only have a fact about a beta of 120 businesses; no fact supports 10,000 users. | Remove or replace with a supported statement …
…
Works anywhere in the world. | No | Product fact: Fieldnote is available in the US and Canada only; no claim of worldwide availability. | Change to "Available in the US and Canada" or remove global claim.
Start your free 30-day trial today. | No | Product fact: trial length is 14 days and requires no credit card; 30-day trial is unsupported. | Change to "Start your free 14-day trial — no credit card required."

How many claims need a fix: 4

What to check: It caught all four problems, and the math is right: 8.1% to 5.2% is a drop of about 36%, not half. This is the most useful marketing prompt in this guide. Run it on every draft, including your own. Two limits: it only checks against the facts you paste, so an outdated fact sheet gives an outdated check. And a check by AI is not an approval. Your usual sign-off still applies.

Keep this in a shared document and paste it into every marketing prompt. Update it when the product changes.

PRODUCT FACTS (approved by [owner], last updated [date])
- What it is, in one sentence:
- Who it's for:
- Features (only what ships today):
- Prices and plans:
- Where it's available:
- Trial or offer terms:
- Numbers we may quote, with their source:
- Claims we must NOT make:

BRAND VOICE
- We sound like: [a person or role, e.g. "an experienced dispatcher"]
- We do: [short sentences, concrete details, ...]
- We never: [banned words, exclamation marks, ...]
- Two real examples of our voice:
  "..."
  "..."

RULE FOR EVERY PROMPT
Use only the facts above. Don't add numbers, customer names, quotes or features. If something is missing, write "[need source]".

To keep tone consistent across a whole team, the guide on style controls shows how to turn a voice card into a reusable style prompt.

• Follow your company's data rules. Use the AI tool your company approved, with your work account.

• Don't paste: customer lists, email addresses, CRM exports, unreleased launch plans or prices, agency contracts, or anything a partner shared in confidence, unless your approved tool and your company's rules allow it.

• Use summaries, not raw data. "Open rate 29%, 4,200 sent" is enough for analysis. The AI doesn't need the subscriber list.

• Check every fact and claim against the fact sheet or the original source, and keep your normal approval process for claims. Some uses of AI are regulated in some places. Ask your legal and compliance team what applies to you before you roll this out.

For what each vendor says it keeps, and which settings to check, see is it safe to use ChatGPT at work?. For why models invent facts, see hallucination basics.

When the same prompts run every week, teams usually build a shared prompt library first, then switch on AI features in the tools they already use. Some go further: software that pulls the week's campaign numbers and drafts the summary, or that runs the claims check on every draft before it reaches a reviewer. Software that takes steps in other systems is an AI agent, and it needs limits, approvals and logs that a chat window doesn't. Start with what is an AI agent? and the overview of AI agents for business leaders.

How do marketers get consistent, on-brand output from AI?

Give the AI the same three things every time: a product fact sheet, a short brand voice card, and two or three real examples of your voice. Add a rule such as "use only the facts above". Save these in a shared document or as custom instructions so the whole team starts from the same place.

What are the best ChatGPT prompts for marketing?

The best marketing prompts include the product facts, the audience, the channel and format, the brand voice and a rule against adding facts. Useful starting tasks are launch posts, brand-voice rewrites, repurposing, review analysis, SEO outlines, test summaries and claims checks. The seven prompts in this guide are ready to copy.

What is prompt engineering for marketing?

Prompt engineering for marketing means writing instructions that get usable copy from an AI tool on the first or second try. It comes down to five parts: task, context (your facts), audience, format and rules. No technical skills are needed. What is prompt engineering? covers the basics.

Can AI write SEO content that ranks?

AI can plan and draft. Whether a page ranks depends on whether it's helpful and original, not on who or what wrote it. Google's guidance on AI-generated content (February 2023) says it rewards high-quality content "however it is produced", and that using AI mainly to manipulate rankings breaks its spam policies. Add your own data and expertise, and check every fact.

Is it safe to paste customer data into ChatGPT for marketing?

Only in a tool your company has approved for that data, with your work account. For most marketing tasks you don't need personal data at all. Summaries and totals are enough. Ask IT which tools are approved.

Can AI replace A/B testing?

No. AI can suggest variants and explain why each might work, which helps you design better tests. Only real people reacting to real messages tell you what performs.

• Learn the method behind these prompts: AI prompts for work

• Catch wrong facts before you publish: how to check AI answers

• Know what you can paste: is it safe to use ChatGPT at work?

• Show the model your voice: few-shot prompting

• The other role playbooks: AI for sales, AI for HR and AI for finance