Skip to content

Prompts · Marketing

AI prompts for marketing, tested with every reply

10 marketing prompts, each run on GPT-6 Luna, Claude Sonnet 5.5, Gemini 3.1 Pro (preview), and Grok 4.7: 40 of 40 replies passed our checks. Below, each prompt with its replies side by side: whether each passed, what it counted as on Pro and what it cost to run. No prompt split the models.

A real reply from these prompts' runs

GPT-6 Luna

Write 5 headlines for a meal-prep delivery service at $9 a meal aimed at nurses working night shifts, one from each angle: price, speed, pain, social proof and curiosity. Label each with its angle. Don't invent numbers or awards.

GPT-6 Luna · passed our check · September 28, 2026

- **Price:** Night-shift meals for $9 each—no prep required  
- **Speed:** Get dinner ready in minutes after a long shift  
- **Pain:** Skip the post-shift scramble for something to eat  
- **Social proof:** A practical meal option for nurses working nights  
- **Curiosity:** What if your next night-shift meal were already taken care of?

Ask your own question: sign up free for 5 messages on 14 of the models, no card.

Start free

Test runs checked . Updated .

Short answer

All 4 models passed 10 of these 10 prompts, in our runs on September 28, 2026. No prompt split the models: every reply passed its check, so these prompts work across all of them.

What these prompts are for

Prompts for the marketing work you do every week: headlines from different angles, a persona grounded in real reviews, A/B tests with hypotheses, a week of posts, features turned into benefits, subject lines with a character limit, a positioning statement, the claims in an ad that need proof, a landing page's first screen and customer acquisition cost.

Marketing prompts fail in two ways: they break a limit (a subject line too long for the inbox) or they invent proof. Our checks look for both.

The prompts at a glance

Every prompt on every model, as our check scored the reply. Tap a prompt to jump to it and read the replies.

Every prompt on every model: passed or failed
PromptGPT-6 LunaClaude Sonnet 5.5Gemini 3.1 ProGrok 4.7
Five headlines, five angles, no invented proofPassedPassedPassedPassed
A buyer persona from real reviewsPassedPassedPassedPassed
A/B tests with a hypothesis eachPassedPassedPassedPassed
A week of social posts leading up to a launchPassedPassedPassedPassed
Features into benefits, without hypePassedPassedPassedPassed
Email subject lines under 45 charactersPassedPassedPassedPassed
A positioning statement in the classic formPassedPassedPassedPassed
Find the claims you'd have to provePassedPassedPassedPassed
A landing page's first screenPassedPassedPassedPassed
Customer acquisition cost, worked outPassedPassedPassedPassed
Passed10 of 1010 of 1010 of 1010 of 10
Each model on these prompts
ModelPassedEach reply on ProCost per replyTime per reply
GPT-6 LunaOpenAI10 of 101 message on Pro$0.00012.5 s
Grok 4.7xAI10 of 101 message on Pro$0.00459.0 s
Gemini 3.1 Pro (preview)Google10 of 101 message on Pro$0.00456.6 s
Claude Sonnet 5.5Anthropic10 of 101 message on Pro$0.00463.7 s
Passed: of the prompts each model answered, how many replies passed their check. Each reply on Pro: what one of these messages counts as on llmwise's Pro plan. Cost: what OpenRouter charged us per reply, on average; in llmwise you pay per message, not per token.

Where the replies differed most

No prompt here split the models: every reply passed its check. These are the prompts where the grader still rated one model's reply well above another's.

  • Features into benefits, without hype

    Graded highest: Claude Sonnet 5.5, 4.7 of 5. Lowest: GPT-6 Luna, 4.0 of 5. Both passed.

    The grader on GPT-6 Luna's reply: “Each line gives a mostly customer-framed benefit and stays accurate and restrained; small flaws are the battery line restating the spec and 'up to' adding a hedge, and mild marketing phrasing such as 'Enjoy' and 'just 1.1 kg'.”
  • Five headlines, five angles, no invented proof

    Graded highest: Claude Sonnet 5.5, 5.0 of 5. Lowest: GPT-6 Luna, 4.3 of 5. Both passed.

    The grader on GPT-6 Luna's reply: “All five angles are present and labelled, but the social-proof line cites no peers or endorsement; nurses and night shifts appear in most lines while speed and pain stay generic, and only the given $9 price is used with no invented figures or awards.”
  • A positioning statement in the classic form

    Graded highest: Grok 4.7, 5.0 of 5. Lowest: Gemini 3.1 Pro, 4.3 of 5. Both passed.

    The grader on Gemini 3.1 Pro's reply: “The reply follows the two-sentence form with every bracket filled and names one clear customer, but the difference mostly restates the benefit, and 'seamlessly' and 'inbox clutter into organized family events' embellish slightly beyond the description.”

The 10 prompts, with every reply

Copy a prompt, replace the parts in [brackets] with your own, and send it to any model. Below each, the replies we got with our example filled in.

1. Five headlines, five angles, no invented proof

Hard · 4 of 4 passed

Naming the angles gets variety; banning invented proof matters most for the social-proof headline, where models reach for fake numbers.

The prompt: replace the parts in [brackets]

Write 5 headlines for [your product] aimed at [your audience], one from each angle: price, speed, pain, social proof and curiosity. Label each with its angle. Don't invent numbers or awards.

Open in llmwise
How we filled it in and checked the replies
Write 5 headlines for a meal-prep delivery service at $9 a meal aimed at nurses working night shifts, one from each angle: price, speed, pain, social proof and curiosity. Label each with its angle. Don't invent numbers or awards.
  • Five angles: One headline per angle, each labelled.
  • Fits the audience: Speaks to night-shift nurses.
  • No invented proof: No made-up numbers of customers, ratings or awards.
  • GPT-6 Luna

    Passed

    Graded 4.3 of 5 on average (lowest 4).

    - **Price:** Night-shift meals for $9 each—no prep required  
    - **Speed:** Get dinner ready in minutes after a long shift  
    - **Pain:** Skip the post-shift scramble for something to eat  
    - **Social proof:** A practical meal option for nurses working nights  
    - **Curiosity:** What if your next night-shift meal were already taken care of?

    1 message on Pro · under $0.0001 to run · 1.8 s · 83 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    **Price:** Real Meals for $9 Each, Delivered to Your Door Before Your Next Night Shift
    
    **Speed:** Heat, Eat, Sleep: Dinner Ready in Minutes When Your Shift Won't Wait
    
    **Pain:** Tired of Vending Machine Dinners at 3 A.M.? Your Break Deserves Better
    
    **Social Proof:** The Meal Plan Night-Shift Nurses Are Passing Around the Break Room
    
    **Curiosity:** What Do Night Nurses Eat at 3 A.M. When They Have a Choice?

    1 message on Pro · $0.0031 to run · 3.7 s · 199 tokens out

  • Gemini 3.1 Pro

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    **Price:** Fuel Your Night Shift With Healthy, Chef-Prepared Meals for Just $9 Each.
    
    **Speed:** Zero-Prep Meals You Can Heat and Eat Before Your Short Break is Over. 
    
    **Pain:** Stop Surviving Your 12-Hour Shifts on Stale Vending Machine Snacks at 3 AM.
    
    **Social Proof:** The Meal Delivery Service Trusted by Fellow Night Shift Nurses.
    
    **Curiosity:** Here is How Night Nurses Are Finally Eating Well Without Ever Cooking.

    1 message on Pro · $0.0045 to run · 7.1 s · 672 tokens out

  • Grok 4.7

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    **Price:** Night-shift meals for $9 — ready when you get home.
    
    **Speed:** Meals delivered before your next night shift starts.
    
    **Pain:** Too exhausted after nights to cook? Dinner’s already done.
    
    **Social proof:** The meal night-shift nurses pass along to the next crew.
    
    **Curiosity:** What do nurses eat when the cafeteria is closed?

    1 message on Pro · $0.0037 to run · 7.9 s · 516 tokens out

2. A buyer persona from real reviews

Everyday · 4 of 4 passed

Quoting a review for each point keeps the persona grounded; without it, models invent ages, jobs and incomes.

The prompt: replace the parts in [brackets]

From these customer reviews, write a one-paragraph persona of our typical buyer: who they are, what they want and what worries them. Quote a review for each point, and invent nothing the reviews don't say. [paste your reviews]

Open in llmwise
How we filled it in and checked the replies
From these customer reviews, write a one-paragraph persona of our typical buyer: who they are, what they want and what worries them. Quote a review for each point, and invent nothing the reviews don't say.

"Bought it for my daughter's first flat. Easy to assemble, and she loves the colour."
"Finally a desk that fits our tiny spare room. Wish it came with cable clips."
"Assembly took 20 minutes on my own. Solid for the price."
"Worried it would wobble on our old floorboards but it's steady."
"Second one I've bought, this time for my home office."
  • Grounded: Each point is backed by a quote from the reviews.
  • Nothing invented: No age, income or job the reviews don't mention.
  • Useful: Captures small spaces, easy assembly and worry about wobbling.
  • GPT-6 Luna

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    The typical buyer is furnishing a home office or a small living space, sometimes buying for a family member: one customer says, “Bought it for my daughter's first flat,” while another wanted “a desk that fits our tiny spare room.” They value easy assembly, a pleasing colour, and sturdy, good-value construction—described as “Easy to assemble,” “Solid for the price,” and “it's steady.” They may buy more than one, as one reviewer says, “Second one I've bought.” Their concerns include whether the desk will wobble on uneven floors and whether it has useful extras such as cable clips: “Worried it would wobble on our old floorboards” and “Wish it came with cable clips.”

    1 message on Pro · $0.0001 to run · 2.4 s · 150 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    **Persona:** Our typical buyer is a practical home-furnisher, often buying for a small or newly set-up space, whether for themselves or for a family member: "Bought it for my daughter's first flat" and "Finally a desk that fits our tiny spare room." They want furniture that is quick and painless to put together and good value: "Assembly took 20 minutes on my own. Solid for the price." They're also happy with the product enough to come back, as one reviewer did: "Second one I've bought, this time for my home office." Their worries are mostly about stability and small practical gaps: "Worried it would wobble on our old floorboards but it's steady," and they'd like a few extras included, such as "Wish it came with cable clips."
    
    **Note on limits:** The reviews don't say anything about buyers' age, income, or job, so I've left those out. The "daughter's flat" and "home office" mentions are the only clues about who the desk is for.

    1 message on Pro · $0.0046 to run · 3.1 s · 299 tokens out

  • Gemini 3.1 Pro

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    Our typical buyer is someone outfitting a small living space or working environment, purchasing the desk for a "home office" or a "daughter's first flat." They want an affordable, compact, and easy-to-build solution, valuing a desk that "fits our tiny spare room" and is "solid for the price." However, they are primarily concerned about the furniture's stability in their environment, noting they were "worried it would wobble on our old floorboards."

    1 message on Pro · $0.0054 to run · 8.0 s · 821 tokens out

  • Grok 4.7

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    Our typical buyer is furnishing a small space for someone else or themselves—a daughter’s first flat, a tiny spare room, or a home office—and may already own one (“Bought it for my daughter's first flat.”; “Second one I've bought, this time for my home office.”). They want a desk that fits a tight room, goes together quickly on their own, looks good, and feels solid for the price (“Finally a desk that fits our tiny spare room.”; “Assembly took 20 minutes on my own. Solid for the price.”; “Easy to assemble, and she loves the colour.”). They worry it will wobble on old floorboards (“Worried it would wobble on our old floorboards but it's steady.”) and that it won’t come with cable clips (“Wish it came with cable clips.”).

    1 message on Pro · $0.0050 to run · 9.3 s · 755 tokens out

3. A/B tests with a hypothesis each

Everyday · 4 of 4 passed

Asking for a hypothesis and one metric per test gets experiments you can run, not a list of ideas.

The prompt: replace the parts in [brackets]

Suggest 3 A/B tests for [the page]. For each: what to change, the hypothesis, and the one metric to watch.

Open in llmwise
How we filled it in and checked the replies
Suggest 3 A/B tests for our pricing page, where 40% of visitors leave within 10 seconds. For each: what to change, the hypothesis, and the one metric to watch.
  • Three real tests: Three distinct changes that could plausibly keep visitors on the page.
  • Hypotheses: Each has a testable hypothesis tied to the early exits.
  • One metric: Each names one metric to watch.
  • GPT-6 Luna

    Passed

    Graded 4.3 of 5 on average (lowest 4).

    1. **Clarify the price upfront**
       - **Change:** Show the starting price and billing cadence near the headline, rather than lower on the page.
       - **Hypothesis:** Visitors who can quickly judge affordability will be less likely to leave immediately.
       - **Metric:** **10-second bounce rate** (share leaving within 10 seconds).
    
    2. **Make plan differences easier to scan**
       - **Change:** Replace dense feature lists with a concise comparison table that highlights the recommended plan and key differences.
       - **Hypothesis:** Less effort to compare plans will help visitors understand their options and continue exploring.
       - **Metric:** **Pricing-page-to-checkout click-through rate**.
    
    3. **Explain the value behind the price**
       - **Change:** Add a short, specific value statement beside the plans—for example, the main outcome customers get or what’s included.
       - **Hypothesis:** Connecting price to a clear benefit will make the offer feel more relevant and reduce quick exits.
       - **Metric:** **10-second bounce rate**.

    1 message on Pro · $0.0001 to run · 3.0 s · 219 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 4.3 of 5 on average (lowest 4).

    **1. Above-the-fold clarity: value-led headline and visible plan summary**
    - **Change:** Replace the generic headline with an outcome-focused one (e.g., "Cut reporting time in half, pick the plan that fits"). Show all plan names, prices, and a one-line "best for" descriptor without scrolling.
    - **Hypothesis:** Many 10-second bounces come from visitors who can't tell within a glance whether the page answers their question. Making the value and price range immediately visible will reduce confusion-driven exits.
    - **Metric:** Bounce rate (or % leaving within 10 seconds).
    
    **2. Plan presentation: highlight a recommended plan and simplify the comparison**
    - **Change:** Mark the mid-tier plan as "Most popular," reduce the feature comparison to 5-6 key differentiators (with a "See full comparison" expander), and make CTAs visually dominant.
    - **Hypothesis:** Visitors face choice overload and a dense feature grid. A clear default and less clutter will help them decide faster and engage with a plan instead of leaving.
    - **Metric:** Plan CTA click-through rate.
    
    **3. Risk reduction: add trust and low-commitment cues near pricing**
    - **Change:** Add a "Start free / 14-day trial, no credit card" line and a money-back guarantee under each price. Add one customer logo strip or short testimonial beside the plans.
    - **Hypothesis:** Some visitors leave because price feels like a risky commitment. Lowering perceived risk will keep more of them on the page and moving toward signup.
    - **Metric:** Trial/signup start rate (per pricing page visitor).
    
    **Tips:**
    - Run one test at a time, or use separate traffic splits, to avoid muddying results.
    - Segment by traffic source and device. Fast bounces are often driven by mismatched ad or email traffic, or by slow mobile load times, so check page speed before testing.

    1 message on Pro · $0.0075 to run · 6.4 s · 645 tokens out

  • Gemini 3.1 Pro

    Passed

    Graded 4.3 of 5 on average (lowest 4).

    Here are 3 A/B tests designed to reduce quick exits on your pricing page:
    
    **1. The "Recommended" Tier Highlight**
    *   **What to change:** Visually emphasize one specific pricing tier (e.g., make it slightly larger, add a "Most Popular" badge, and use a contrasting button color) while slightly muting the others. 
    *   **Hypothesis:** Highlighting a default choice reduces choice paralysis and cognitive overload, making visitors more likely to stay and evaluate the primary option rather than abandoning the page out of confusion.
    *   **Metric to watch:** 10-second bounce rate.
    
    **2. Benefit-Driven Headline and Subheadline**
    *   **What to change:** Replace a generic headline like "Pricing Plans" with a value-driven headline (e.g., "Everything you need to grow, at a price that makes sense") and add 3 bullet points of core product benefits right above the pricing cards.
    *   **Hypothesis:** Reminding visitors of the immediate value of your product *before* forcing them to look at costs will capture their attention and give them a reason to spend time reading the tiers.
    *   **Metric to watch:** Average time on page.
    
    **3. Monthly/Annual Toggle Simplification**
    *   **What to change:** If you currently show monthly and annual prices side-by-side or require scrolling, hide the annual prices behind a single, highly visible "Monthly / Annual (Save 20%)" toggle at the top of the cards. Default to the monthly view.
    *   **Hypothesis:** A cleaner, less cluttered interface minimizes the initial information density, making the pricing easier to digest at a glance.
    *   **Metric to watch:** Click-through rate (CTR) on the plan selection buttons.

    1 message on Pro · $0.0054 to run · 7.9 s · 825 tokens out

  • Grok 4.7

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    Three tests aimed at the first 10 seconds, not later-page conversion.
    
    **1. Outcome headline, plans second**  
    Change: Replace the current hero with one line on who it’s for and the outcome (“For [role]: [result]”), then three plan names and prices. Move the feature grid below the fold.  
    Hypothesis: People leave because they can’t tell what they’re buying before the page feels like a spreadsheet.  
    Metric: Share of visits that last longer than 10 seconds.
    
    **2. Annual price as the default**  
    Change: Show the annual price (with “$X/mo, billed yearly”) as the default; keep monthly behind a toggle. Same plans, same copy.  
    Hypothesis: The monthly number reads as expensive and people bounce before comparing plans.  
    Metric: Plan-select rate (click on a plan CTA).
    
    **3. One recommended plan above the fold**  
    Change: Highlight a single “Most teams start here” plan with price, 3 bullets, and a primary CTA in the first screen. Other plans stay, but smaller and lower.  
    Hypothesis: Too many equal options with no cue causes an immediate “I’ll decide later” exit.  
    Metric: Click-through rate on the recommended plan’s CTA.

    1 message on Pro · $0.0056 to run · 15.1 s · 924 tokens out

4. A week of social posts leading up to a launch

Everyday · 4 of 4 passed

A table with a caption limit gets a plan you can schedule, and a date gives the week a shape.

The prompt: replace the parts in [brackets]

Make a 7-day Instagram plan for [your business] as a table with the day, the post idea, a caption under 20 words, and how many hashtags to use.

Open in llmwise
How we filled it in and checked the replies
Make a 7-day Instagram plan for a plant shop opening its second store on Saturday as a table with the day, the post idea, a caption under 20 words, and how many hashtags to use.
  • Seven days: A row for each of seven days, building up to Saturday's opening.
  • Short captions: Every caption is under 20 words.
  • Varied: The posts vary rather than repeating one idea.
  • GPT-6 Luna

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    | Day | Post idea | Caption (under 20 words) | Hashtags |
    |---|---|---|---:|
    | Monday | Teaser photo of the new storefront | A new little jungle is growing. Our second shop opens Saturday! | 5 |
    | Tuesday | Sneak peek of plants being unpacked | Meet the leafy arrivals waiting to find their forever homes. | 5 |
    | Wednesday | Short video of the shop setup | Shelves are filling, pots are landing, and opening day is getting closer. | 5 |
    | Thursday | Introduce the new location and share directions | Find us at our new spot this Saturday—we can’t wait to welcome you! | 5 |
    | Friday | Staff picks: favorite plants for the new shop | Our team picked their opening-week favorites. Which one would you take home? | 5 |
    | Saturday | Opening-day photo or Reel with customers and ribbon cutting | We’re open! Come celebrate our second store with us today. | 8 |
    | Sunday | Customer photos or a thank-you post | Thanks for making opening day so special. We’re happy to grow with you! | 5 |

    1 message on Pro · $0.0001 to run · 3.2 s · 242 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    Here's a 7-day plan leading up to and including the Saturday opening (Day 7 = Saturday).
    
    | Day | Post Idea | Caption (under 20 words) | Hashtags |
    |---|---|---|---|
    | Sun (Day 1) | Teaser Reel: close-up of new storefront covered in leaves, "something's growing" text | Something green is sprouting across town. Store #2 opens this Saturday. 🌱 | 5 |
    | Mon (Day 2) | Carousel: sneak peek of the new store interior, one shelf per slide | Shelves are stocked, lights are on. Which corner would you visit first? | 6 |
    | Tue (Day 3) | Behind-the-scenes Reel: team unpacking and arranging plants | Unboxing hundreds of plants takes teamwork. Four days until doors open! | 5 |
    | Wed (Day 4) | Poll Story + feed post: "Which plant should we feature at opening?" | You pick the star of our grand opening display. Vote in Stories! 🗳️ | 4 |
    | Thu (Day 5) | Announcement graphic: opening day hours, address, and giveaways | Saturday, doors open at 10am. First 50 visitors get a free plant! | 7 |
    | Fri (Day 6) | Countdown Reel: "Tomorrow!" with quick clips of finished store | Last night before opening. Tomorrow, come say hi and take a plant home. | 6 |
    | Sat (Day 7) | Live opening Reel/Stories: ribbon cutting, crowd, happy customers | We're open! Thank you for helping us grow. Come explore Store #2. 🌿 | 8 |
    
    **Tips:**
    - Add a location tag and the new store's address in every post.
    - Repost customer Stories on Saturday and the days after.
    - Pin the Thursday announcement to your profile until Saturday.

    1 message on Pro · $0.0074 to run · 6.2 s · 633 tokens out

  • Gemini 3.1 Pro

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    | Day | Post Idea | Caption (<20 words) | Number of Hashtags |
    | :--- | :--- | :--- | :--- |
    | **Sunday** | Teaser/Announcement | Big news! Our plant family is growing. Store #2 is opening this Saturday. Stay tuned! 🌱 | 5 hashtags |
    | **Monday** | Behind the Scenes | Unboxing fresh green beauties for the new shop. We can't wait for you to see! 📦🪴 | 7 hashtags |
    | **Tuesday** | Location Reveal | We’re officially joining the neighborhood! Come find our new green oasis this Saturday. 🌿🏘️ | 8 hashtags |
    | **Wednesday** | Sneak Peek (Rare Plants) | Spotted: rare Monsteras hitting the shelves at the new location. Get here early Saturday! 🪴✨ | 6 hashtags |
    | **Thursday** | Meet the Team | Meet Sarah, our new store manager! She’s ready to help you find your perfect plant match. 👋🌿 | 5 hashtags |
    | **Friday** | 1-Day Countdown | 1 day left! Set your alarms. Our doors open tomorrow at 10 AM. See you there! ⏰🌱 | 10 hashtags |
    | **Saturday** | Grand Opening Day | We are OPEN! Come celebrate our second location with us today. Let’s get planting! 🥳🌿🪴 | 12 hashtags |

    1 message on Pro · $0.0058 to run · 8.1 s · 893 tokens out

  • Grok 4.7

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    | Day | Post idea | Caption | Hashtags |
    |---|---|---|---|
    | Sunday | Countdown teaser | Something green is growing. Second store opens Saturday. | 5 |
    | Monday | New-space sneak peek | Shelves up, light pouring in. Almost ready for you. | 6 |
    | Tuesday | What you’ll find | New store, same plant obsession. Rare finds land Saturday. | 8 |
    | Wednesday | Setup behind the scenes | Soil, pots, and a lot of excitement. One shop becoming two. | 5 |
    | Thursday | Opening details | Doors open Saturday at 10. Come say hi and take a plant home. | 7 |
    | Friday | Final reminder | One more sleep. Grand opening tomorrow. First 20 guests get a cutting. | 6 |
    | Saturday | We’re open | Second store, first day. Come grow with us. | 10 |

    1 message on Pro · $0.0050 to run · 11.2 s · 797 tokens out

5. Features into benefits, without hype

Everyday · 4 of 4 passed

"Without exaggerating" keeps benefits believable; the hype words are the first thing a model reaches for.

The prompt: replace the parts in [brackets]

Rewrite these product features as customer benefits, one line each, without exaggerating. [paste your features]

Open in llmwise
How we filled it in and checked the replies
Rewrite these product features as customer benefits, one line each, without exaggerating.

- 20-hour battery
- Weighs 1.1 kg
- Two USB-C ports
- Aluminium body
- 14-inch matte screen
  • Benefits: Each line says what the feature does for the customer.
  • Accurate: Nothing beyond what the features support.
  • No hype: No superlatives or buzzwords.
  • Automatic rules: exactly 5 bullet points; doesn't use “revolutionary”; doesn't use “game-changer”; doesn't use “game changer”.
  • GPT-6 Luna

    Passed

    Graded 4.0 of 5 on average (lowest 4).

    - Enjoy up to 20 hours of battery life between charges.
    - Carry it easily at just 1.1 kg.
    - Connect two USB-C devices at once.
    - Get a durable-feeling design with an aluminium body.
    - Work on a 14-inch matte screen with reduced glare.

    1 message on Pro · under $0.0001 to run · 1.7 s · 63 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    - **20-hour battery:** Get through a full workday and more without hunting for an outlet.
    - **Weighs 1.1 kg:** Easy to carry in a bag all day without adding much weight.
    - **Two USB-C ports:** Charge and connect a display or accessory at the same time.
    - **Aluminium body:** A sturdy, solid feel that holds up well to daily travel.
    - **14-inch matte screen:** A comfortable size for working, with less glare in bright rooms.

    1 message on Pro · $0.0027 to run · 1.9 s · 161 tokens out

  • Gemini 3.1 Pro

    Passed

    Graded 4.3 of 5 on average (lowest 4).

    - **20-hour battery:** Work or play through the entire day without needing to plug in.
    - **Weighs 1.1 kg:** Carry it effortlessly in your bag without straining your shoulder.
    - **Two USB-C ports:** Charge your device and connect an accessory or external monitor at the same time.
    - **Aluminium body:** Protect your device against everyday wear and tear with a durable casing.
    - **14-inch matte screen:** Work comfortably in bright rooms or outdoors without distracting screen glare.

    1 message on Pro · $0.0033 to run · 5.3 s · 477 tokens out

  • Grok 4.7

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    - Work through a full day without stopping to charge.
    - Light enough to carry between home, work, and meetings.
    - Charge the laptop and plug in an accessory at the same time.
    - A rigid shell that holds up to everyday handling.
    - A compact screen that’s easier to read in bright light, with fewer reflections.

    1 message on Pro · $0.0036 to run · 7.0 s · 486 tokens out

6. Email subject lines under 45 characters

Hard · 4 of 4 passed

Inboxes cut subject lines short, so a character limit matters, and it's easy for a model to miss on one line of five.

The prompt: replace the parts in [brackets]

Write 5 email subject lines for [the campaign], each under 45 characters, as a numbered list with nothing else.

Open in llmwise
How we filled it in and checked the replies
Write 5 email subject lines for a 48-hour sale with 30% off running shoes, each under 45 characters, as a numbered list with nothing else.
  • Under 45 characters: Every subject line is under 45 characters.
  • Says the offer: Most make the sale, the discount or the deadline clear.
  • Only the list: The numbered list and nothing else.
  • Automatic rules: exactly 5 bullet points.
  • GPT-6 Luna

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. 48 Hours: 30% Off Running Shoes
    2. Run Into 30% Off—Ends in 48 Hours
    3. 30% Off Running Shoes for 48 Hours
    4. Your Fastest Deal: 30% Off Running Shoes
    5. 48-Hour Sprint: Save 30% on Running Shoes

    1 message on Pro · under $0.0001 to run · 1.4 s · 74 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. 48 Hours Only: 30% Off Running Shoes
    2. Run Faster, Pay Less: 30% Off Ends Soon
    3. Lace Up! 30% Off Shoes for 2 Days Only
    4. Your Next Run Is 30% Cheaper
    5. Flash Sale: 30% Off Running Shoes, 48 Hrs

    1 message on Pro · $0.0023 to run · 1.7 s · 124 tokens out

  • Gemini 3.1 Pro

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. 30% Off Running Shoes! 48 Hrs Only
    2. 48-Hour Sale: Save 30% on Running Shoes
    3. Run Faster: 30% Off All Shoes for 48h
    4. 30% Off Running Shoes Ends in 48 Hours
    5. 48h Flash Sale: 30% Off Running Shoes

    1 message on Pro · $0.0043 to run · 5.5 s · 645 tokens out

  • Grok 4.7

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. 48 Hours Only: 30% Off Shoes
    2. 30% Off Running Shoes for 48 Hrs
    3. Last Call: 30% Off Running Shoes
    4. 48-Hour Flash: 30% Off Runners
    5. Save 30% on Running Shoes — 48 Hrs

    1 message on Pro · $0.0084 to run · 14.3 s · 1,509 tokens out

7. A positioning statement in the classic form

Everyday · 4 of 4 passed

A fixed form makes you choose one customer, one benefit and one difference, which is the point of positioning.

The prompt: replace the parts in [brackets]

Write a positioning statement for [your product] in this form: For [target customer] who [need], [product] is a [category] that [key benefit]. Unlike [main alternative], it [difference].

Open in llmwise
How we filled it in and checked the replies
Write a positioning statement for Tandem, a shared family calendar app that reads school emails and adds the dates for you in this form: For [target customer] who [need], [product] is a [category] that [key benefit]. Unlike [main alternative], it [difference].
  • The form: Follows the two-sentence form, with every bracket filled.
  • Sharp: One clear customer, benefit and difference.
  • Credible: Claims nothing the description doesn't support.
  • Automatic rules: at most 2 sentences; includes “Unlike”; includes “Tandem”.
  • GPT-6 Luna

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    For busy families who need to keep everyone in sync without manually tracking every school date, Tandem is a shared family calendar app that reads school emails and adds important dates for you. Unlike traditional calendars that rely on you to enter events yourself, it turns school announcements into ready-to-go calendar entries.

    1 message on Pro · under $0.0001 to run · 1.5 s · 63 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    For busy parents of school-age children who struggle to keep track of dates buried in school emails, Tandem is a shared family calendar app that automatically reads those emails and adds events for you, so everyone in the household stays in sync without manual entry. Unlike general-purpose calendars such as Google Calendar or Apple Calendar, it finds and adds school dates on its own instead of relying on you to copy them over by hand.

    1 message on Pro · $0.0024 to run · 1.9 s · 134 tokens out

  • Gemini 3.1 Pro

    Passed

    Graded 4.3 of 5 on average (lowest 4).

    For busy parents who struggle to keep track of school events, Tandem is a shared family calendar app that automatically extracts dates from school emails and adds them to your schedule. Unlike standard digital calendars, it eliminates manual data entry by seamlessly turning inbox clutter into organized family events.

    1 message on Pro · $0.0027 to run · 4.7 s · 387 tokens out

  • Grok 4.7

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    For busy parents who miss school dates buried in email, Tandem is a shared family calendar that reads school messages and adds the events for you. Unlike a regular shared calendar, it fills itself from the inbox instead of waiting on someone to type everything in.

    1 message on Pro · $0.0023 to run · 3.5 s · 231 tokens out

8. Find the claims you'd have to prove

Hard · 4 of 4 passed

Advertising rules require proof for claims like these; a list of what needs evidence is the useful review.

The prompt: replace the parts in [brackets]

Here's our draft ad copy. List every claim we'd need evidence for, quoting it, and say what evidence would back it up. Don't rewrite the ad. [paste your ad]

Open in llmwise
How we filled it in and checked the replies
Here's our draft ad copy. List every claim we'd need evidence for, quoting it, and say what evidence would back it up. Don't rewrite the ad.

The UK's most-loved coffee subscription! Our beans are 100% organic and roasted within 24 hours of shipping. Customers save £200 a year compared with café coffee. Join 50,000 happy subscribers. Better for the planet, guaranteed.
  • Finds them all: Finds all six: most-loved, 100% organic, roasted within 24 hours, £200 saving, 50,000 subscribers, better for the planet.
  • Right evidence: Suggests fitting evidence for each (certification, survey, calculation, records).
  • No rewrite: Doesn't rewrite the ad.
  • Automatic rules: includes “organic”; includes “50,000”; includes “£200”.
  • GPT-6 Luna

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    - **“The UK's most-loved coffee subscription”** — Evidence that “most-loved” is a fair, measurable comparison with UK competitors, such as a representative independent consumer survey using a clear definition of “loved.” Define the market and comparison period.
    - **“Our beans are 100% organic”** — Current organic certification covering all beans sold, plus supplier and batch records to substantiate the claim. If relying on UK organic rules, verify that the relevant certification and labeling requirements are met.
    - **“roasted within 24 hours of shipping”** — Dated roasting and dispatch records for orders, showing the time between roasting and shipment, including how exceptions are handled.
    - **“Customers save £200 a year compared with café coffee”** — A robust, current calculation supporting the £200 figure: subscription and café prices, serving sizes, usage assumptions, and a representative customer basis. The comparison must be like-for-like and make clear whether £200 is typical, average, or otherwise qualified.
    - **“Join 50,000 happy subscribers”** — Records showing at least 50,000 subscribers under a clearly defined count (for example, current active subscribers rather than cumulative sign-ups). “Happy” also needs support, such as reliable customer satisfaction data; the number alone does not substantiate it.
    - **“Better for the planet”** — Evidence identifying what the product is better than and in what environmental respects, backed by a sound comparison—such as a relevant life-cycle assessment using comparable products and methods. The broad wording risks implying an overall environmental benefit.
    - **“guaranteed”** — Evidence of a genuine, accessible guarantee with clear terms, scope, duration, and remedy. As written, it is unclear what environmental benefit is being guaranteed.

    1 message on Pro · $0.0001 to run · 5.3 s · 364 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    Here are the claims that need substantiation, in order of appearance:
    
    **1. "The UK's most-loved coffee subscription!"**
    - This is a superlative comparative claim. Evidence: independent survey or ranking data comparing named competitors (e.g., Trustpilot/Feefo scores, NPS, or a commissioned poll with a representative UK sample), with methodology, dates, and sample size. "Most-loved" is vague, so you'd need to define the metric. Without one, it's hard to defend.
    
    **2. "Our beans are 100% organic"**
    - Evidence: valid organic certification (e.g., Soil Association or another UK-recognised certifier) covering all beans sold, plus supplier certificates and batch traceability. "100%" means no exceptions across every product, blend, and time period. Certification must also cover your roasting and packing operations if you use the organic logo.
    
    **3. "roasted within 24 hours of shipping"**
    - Evidence: roasting and dispatch timestamps or batch logs showing this holds for every order, not on average. Also confirm it holds during peak periods and for all products.
    
    **4. "Customers save £200 a year compared with café coffee"**
    - Evidence: a documented calculation with stated assumptions: price per cup at cafés (sourced, e.g., market data on average UK café prices), price per cup of your coffee, and cups per week. Also needs to reflect a typical customer's actual consumption rather than a best case. Whether it's "customers" (all) or a typical/average customer also needs clarifying, and the comparison must include your subscription cost and any delivery fees.
    
    **5. "Join 50,000 happy subscribers"**
    - Two claims here:
      - *50,000 subscribers:* current, verifiable subscriber records (active subscribers, not cumulative sign-ups).
      - *"Happy":* satisfaction data such as survey results, review scores, or retention/churn figures showing most subscribers are satisfied.
    
    **6. "Better for the planet"**
    - A comparative environmental claim. Evidence: a life-cycle assessment or carbon footprint comparison against a clearly defined alternative (better than what?), covering packaging, transport, and sourcing. Organic certification alone doesn't prove this. Under the CMA's Green Claims Code, the claim must be clear, fair, and not omit significant trade-offs.
    
    **7. "guaranteed"**
    - Evidence: a defined, published guarantee (what's promised, remedy offered, e.g., refund) with terms available to consumers. It's unclear what is guaranteed here. If it attaches to the environmental claim, it's very hard to back up, as you'd need to be able to deliver a remedy if the claim proves untrue.
    
    **Regulatory note:** In the UK, these fall under the CAP Code (ASA) and the Consumer Protection from Unfair Trading Regulations / DMCC Act. The ASA expects you to hold substantiation *before* the ad runs, so gather it in advance.

    1 message on Pro · $0.0113 to run · 8.9 s · 983 tokens out

  • Gemini 3.1 Pro

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    Here are the claims from your ad copy and the evidence needed to back them up:
    
    *   **"The UK's most-loved coffee subscription!"**
        *   *Evidence needed:* Independent market research, consumer survey results, or comparative data from verified review platforms (like Trustpilot) showing higher satisfaction ratings than all major UK competitors. 
    *   **"Our beans are 100% organic"**
        *   *Evidence needed:* Valid certificates from a recognized organic regulatory body (e.g., Soil Association).
    *   **"roasted within 24 hours of shipping"**
        *   *Evidence needed:* Internal operational logs, timestamped roasting and dispatch records, or standard operating procedures proving this timeframe is consistently met.
    *   **"Customers save £200 a year compared with café coffee"**
        *   *Evidence needed:* A documented mathematical breakdown calculating the cost of your annual subscription versus the average annual cost of café coffee (using reputable, cited market averages for café pricing and typical consumption rates).
    *   **"Join 50,000 happy subscribers"**
        *   *Evidence needed:* Internal CRM or sales database records confirming a current, active subscriber count of at least 50,000.
    *   **"Better for the planet, guaranteed."**
        *   *Evidence needed:* Independent sustainability reports or a Life Cycle Assessment (LCA) proving a lower environmental impact compared to standard coffee supply chains. You also need clearly published terms and conditions explaining exactly what the "guarantee" means and how customers can claim it.

    1 message on Pro · $0.0064 to run · 9.3 s · 988 tokens out

  • Grok 4.7

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    **Claims that need evidence**
    
    1. **“The UK’s most-loved coffee subscription”**  
       Independent, recent UK research showing your service ranks highest for preference or affection among coffee subscriptions, with a clear sample, question, and competitor set. An award or high review score alone does not prove “most-loved.”
    
    2. **“100% organic”**  
       Current organic certification (for example Soil Association or an equivalent UK-recognised scheme) covering every bean in the subscription, not only some lots or the farm in general.
    
    3. **“roasted within 24 hours of shipping”**  
       Records showing roast date versus dispatch date for actual orders, and that the 24-hour window is met as a rule. Define whether “shipping” means dispatch or delivery.
    
    4. **“Customers save £200 a year compared with café coffee”**  
       A published calculation: cups per year, your price, the café prices and drinks used, locations, and that the comparison is like-for-like and typical—not a best-case example.
    
    5. **“50,000 happy subscribers”**  
       - **50,000:** a current count of active subscribers at the time the ad runs.  
       - **happy:** satisfaction data (survey scores, review ratings with volume) showing subscribers are happy, not just that they subscribe.
    
    6. **“Better for the planet, guaranteed”**  
       - **Better for the planet:** evidence of a real environmental benefit versus a stated alternative (for example lifecycle data on carbon, packaging, or sourcing). A vague green claim is not enough.  
       - **guaranteed:** written guarantee terms—what is promised and what the customer gets if it is not met—and proof you can honour it.

    1 message on Pro · $0.0069 to run · 16.6 s · 1,183 tokens out

9. A landing page's first screen

Everyday · 4 of 4 passed

Word limits per line are how landing pages are designed; they force one clear promise.

The prompt: replace the parts in [brackets]

Write the first screen of a landing page for [your product]: a headline under 8 words, a subheadline under 20 words, and a button label under 4 words. Label each.

Open in llmwise
How we filled it in and checked the replies
Write the first screen of a landing page for a budgeting app for couples who share some bills but not all: a headline under 8 words, a subheadline under 20 words, and a button label under 4 words. Label each.
  • Within the limits: Headline under 8 words, subheadline under 20, button under 4.
  • Specific: Speaks to couples sharing some bills, not budgeting in general.
  • Labelled: Each part is labelled.
  • GPT-6 Luna

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    **Headline:** Budget Together, Stay Independent  
    **Subheadline:** Split shared bills fairly while keeping personal spending your own.  
    **Button:** Get Started

    1 message on Pro · under $0.0001 to run · 1.2 s · 34 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    **Headline:** Share the bills. Keep the rest yours.
    
    **Subheadline:** Split rent and utilities fairly, while keeping personal spending private. Budgeting built for how couples actually live.
    
    **Button label:** Start budgeting together

    1 message on Pro · $0.0019 to run · 1.4 s · 82 tokens out

  • Gemini 3.1 Pro

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    **Headline:** Budget together without combining every penny.
    
    **Subheadline:** Easily split shared expenses like rent and groceries while keeping your personal spending completely separate.
    
    **Button:** Get started together

    1 message on Pro · $0.0040 to run · 6.0 s · 596 tokens out

  • Grok 4.7

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    **Headline:** Share bills. Keep the rest.  
    **Subheadline:** Track joint expenses and personal spending in one place, without merging every dollar.  
    **Button:** Start together

    1 message on Pro · $0.0017 to run · 2.0 s · 105 tokens out

10. Customer acquisition cost, worked out

Hard · 4 of 4 passed

Asking for the definition first makes the model include every marketing cost, not just ads.

The prompt: replace the parts in [brackets]

Explain customer acquisition cost in two sentences, then work out ours: last month we spent [ad spend] on ads and [other costs] on marketing salaries, and got [new customers] new customers. End with a line "Final answer: " and the cost per customer.

Open in llmwise
How we filled it in and checked the replies
Explain customer acquisition cost in two sentences, then work out ours: last month we spent $4,500 on ads and $3,000 on marketing salaries, and got 150 new customers. End with a line "Final answer: " and the cost per customer.
  • Checked automatically: the reply's last “Final answer:” line must give 50.
  • GPT-6 Luna

    Passed

    Final answer 50 per customer: right.

    Customer acquisition cost (CAC) is the average amount spent to gain one new customer. Yours is ($4,500 in ads + $3,000 in marketing salaries) ÷ 150 new customers = $50 per customer.
    
    Final answer: $50 per customer

    1 message on Pro · under $0.0001 to run · 3.5 s · 117 tokens out

  • Claude Sonnet 5.5

    Passed

    Final answer 50 per customer: right.

    Customer acquisition cost (CAC) is the average amount you spend to win one new customer, calculated by dividing total sales and marketing costs by the number of new customers gained in the same period. It helps you judge whether your marketing spend is efficient, especially when compared to the revenue each customer generates over time.
    
    **Your calculation:**
    - Total spend: $4,500 (ads) + $3,000 (marketing salaries) = $7,500
    - New customers: 150
    - CAC = $7,500 ÷ 150 = $50
    
    Final answer: $50 per customer

    1 message on Pro · $0.0029 to run · 1.9 s · 183 tokens out

  • Gemini 3.1 Pro

    Passed

    Final answer 50: right.

    Customer Acquisition Cost (CAC) is the total expense a business incurs to gain a single new customer. It is calculated by dividing all sales and marketing costs by the number of new customers acquired during a specific period. 
    
    To work out yours, we add your total marketing expenses ($4,500 in ads + $3,000 in salaries = $7,500) and divide that by your 150 new customers. 
    
    Final answer: $50

    1 message on Pro · $0.0032 to run · 4.6 s · 460 tokens out

  • Grok 4.7

    Passed

    Final answer 50: right.

    Customer acquisition cost (CAC) is the average amount a business spends on sales and marketing to gain one new customer. It is calculated by dividing total acquisition spend by the number of new customers acquired in the same period.
    
    Last month’s acquisition spend was $4,500 on ads plus $3,000 on marketing salaries, or $7,500 total. Divided by 150 new customers, that is $50 per customer.
    
    Final answer: $50

    1 message on Pro · $0.0024 to run · 3.2 s · 251 tokens out

Getting more from these prompts

  • Name the audience precisely: "nurses on night shifts" beats "busy people".

  • Give limits in the units the channel uses: characters for subject lines, words for captions.

  • Ban invented proof explicitly, then check the social-proof lines first.

  • Ask for labelled variations (by angle, by tone) so you can test them against each other.

How we ran and checked them

Each prompt was sent the way llmwise sends a message: the app's own system prompt, each model's own settings, and Pro's reply size limit (8,000 tokens), through OpenRouter. Every reply is shown as it came.

Rubric (graded). The grader model scores the reply from 1 to 5 on each published criterion. It passes with an average of 4 or more and no criterion under 3, and only if it also meets the prompt's automatic rules (length, words it must or mustn't use).

Final answer. Automatic. The reply's last “Final answer:” line must hold the right value.

The grader is Claude Opus 5.5 at low reasoning effort; its own replies are graded by GPT-6 Astra, so no model grades itself. Its prompt is on our test runs page.

More tested prompts

Questions

Which AI writes the best marketing copy?

In our runs, all 4 models passed 10 of 10. Each reply was graded against a published rubric, and the ones with a limit (characters, words, a format) were also checked automatically.

What's the biggest risk with AI marketing copy?

Models invent proof: customer numbers, ratings and awards you never gave them. Say "don't invent numbers or awards", and check every claim before it runs; one prompt here lists the claims you'd need evidence for.

Can I try these prompts for free?

Yes. When the free-message box shows on this page, one message needs no account (GPT-6 Luna answers it); signing up gives 5 free messages, once, with no card.

Claude, GPT, Gemini, DeepSeek, Grok, Kimi, and GLM, in one chat.

See what a message costs before you send it. Free is 5 messages to try; sign in with an email link, no password or card.