A marketing-focused model comparison for ad variants, SEO metadata and abandoned cart email sequences.

/research for supplied product evidence, /document to create a The marketing group tested three formats that look simple until the constraints arrive: three Meta ad variants with character caps and prohibited skincare claims, an organic-dog-treat collection page with a title, description and headings, and a three-email sequence with one discount that could appear only in the final message.
Brief: Write 3 Meta ad variants for this product, each with a primary text of at most 125 characters and a headline of at most 40 characters. Vary the angle across the three (routine, ingredients, gifting). No medical or curative claims. Output as a numbered list with the two fields labelled.
Input: Product: Lumen Vitamin C serum, 30 ml, 42 EUR Key ingredients: 15% stabilised vitamin C, hyaluronic acid, vitamin E Claims we are allowed to make: brightens the look of skin, supports a more even-looking tone, lightweight texture, absorbs quickly, fragrance free, vegan Claims we are NOT allowed to make: treats or cures acne, removes wrinkles, dermatologist recommended, clinically proven More constraints appear in the brief below.
The point of this brief was not to reward a general essay. It asked for a particular format, a particular length and careful use of the supplied facts.
Brief: Write the SEO title (at most 60 characters) and meta description (at most 155 characters) for this collection page. Include the primary keyword naturally in both. Then list 5 H2 headings for the page, one line each. Do not invent product counts or prices.
Input: Store: Barkley and Co, Shopify store selling dog treats in the UK Page: collection page for organic dog treats Primary keyword: organic dog treats Secondary keywords: natural dog treats, grain free dog treats uk, healthy dog treats More constraints appear in the brief below.
The point of this brief was not to reward a general essay. It asked for a particular format, a particular length and careful use of the supplied facts.
Brief: Write a 3 email abandoned cart sequence. For each email give the send timing, subject line (under 50 characters), preview text (under 90 characters) and body (under 120 words). Only email 3 may mention the discount. Friendly, not pushy, no fake urgency beyond the real 48 hour code expiry.
Input: Store: Terra Ceramics, handmade stoneware mugs and bowls Cart: 2 x Speckled mug (24 EUR each), 1 x Serving bowl (58 EUR). Cart total 106 EUR. Discount we can offer: 10% with code TERRA10, valid 48 hours from email 3 only. No other discounts, no free shipping offer. Facts: each piece is thrown by hand in Porto, dishwasher and microwave safe, ships in 2 to 4 working days. Returns within 30 days.
The point of this brief was not to reward a general essay. It asked for a particular format, a particular length and careful use of the supplied facts.
Claude Fable 5.1 leads meta ads at 97.5. The best answers treated the format as part of the strategy rather than adding the strategy after the copy was written. The weaker answers show what breaks when a character limit, a medical claim or a disclosure rule becomes an afterthought.
| Example | Output and review |
|---|---|
| Higher-scoring example Claude Fable 5.1, run 1 | 1. Routine angle - Primary text: Add one step, brighten the look of your skin. Lumen Vitamin C serum: lightweight, absorbs fast, fragrance free. - Headline: One step to brighter-looking skin 2. Ingredients angle - Primary text: 15% stabilised vitamin C, hyaluronic acid and vitamin E in one lightweight, vegan serum. Fragrance free. 30 ml, 42 EUR. - Headline: 15% Vitamin C + Hyaluronic Acid 3. Gifting angle - Primary text: A gift for anyone who loves skincare: Lumen Vitamin C serum, vegan and fragrance free. Free EU shipping over 40 EUR. - Headline: The skincare gift that ships free Review note: All three variants meet the character limits, use distinct requested angles, and stay within supported product claims and shipping terms. |
| Lower-scoring example Grok 4.3, run 2 | 1. Primary text: Add Lumen Vitamin C serum to your daily skincare routine. Brightens the look of skin and supports a more even-looking tone. Lightweight texture absorbs quickly. Fragrance free, vegan. 30 ml, 42 EUR with free EU shipping. Headline: Daily routine essential 2. Primary text: Lumen Vitamin C serum features 15% stabilised vitamin C with hyaluronic acid and vitamin E. Brightens the look of skin and supports a more even-looking tone. Lightweight, absorbs quickly, fragrance free and vegan. 42 EUR. Headline: Key ingredients at work 3. Primary text: Give the gift of brighter-looking skin. Lumen Vitamin C serum brightens the look of skin and supports a more even-looking tone. Lightweight texture absorbs quickly. Fragrance free, vegan. 30 ml, 42 EUR with free EU shipping. Headline: Thoughtful skincare gift Review note: All three primary texts exceed the 125-character limit and require substantial trimming; facts, claims, headlines, formatting and angle variation otherwise comply, but repeated copy weakens clarity. |
GPT-5.6 Luna leads seo metadata at 95.0. The best answers treated the format as part of the strategy rather than adding the strategy after the copy was written. The weaker answers show what breaks when a character limit, a medical claim or a disclosure rule becomes an afterthought.
| Example | Output and review |
|---|---|
| Higher-scoring example DeepSeek V4 Pro, run 1 | SEO Title: Organic Dog Treats Made in Yorkshire | Barkley and Co Meta Description: Shop organic dog treats made in Yorkshire with simple, natural ingredients. Grain free & healthy options. Free UK delivery over £30, 10% off subscriptions. H2 Headings: - Why Choose Our Organic Dog Treats? - Made in Yorkshire with Simple Ingredients - Grain Free & Natural Dog Treats - Free UK Delivery Over £30 - Subscribe & Save 10% on Healthy Treats Review note: The title is 51 characters and the meta description is 155 characters, both include the primary keyword, and the answer provides exactly five relevant H2s using the supplied details. |
| Lower-scoring example Grok 4.3, run 2 | SEO Title: Organic Dog Treats | Barkley and Co UK Meta Description: Shop organic dog treats made in Yorkshire with single or short ingredient lists. Natural dog treats and grain free dog treats UK options available, plus healthy dog treats with free UK delivery over £30 and 10% off subscriptions. H2 Headings: Our Yorkshire Made Organic Dog Treats Natural Dog Treats with Short Ingredients Grain Free Dog Treats UK Healthy Dog Treats Subscription Free UK Delivery Over £30 Review note: The title and five headings meet the structural requirements, but the meta description exceeds 155 characters, stuffs keywords, and asserts unconfirmed grain-free availability. |
GPT-6 Astra leads abandoned-cart email at 98.3. The best answers treated the format as part of the strategy rather than adding the strategy after the copy was written. The weaker answers show what breaks when a character limit, a medical claim or a disclosure rule becomes an afterthought.
| Example | Output and review |
|---|---|
| Higher-scoring example GPT-6 Astra, run 1 | ### Email 1 Send timing: 1 hour after cart abandonment Subject: Your Terra Ceramics picks Preview text: Your Speckled mugs and Serving bowl, all in one place. Body: Hi there, Still thinking it over? Here’s what you picked at Terra Ceramics: - 2 × Speckled mug - 24 EUR each - 1 × Serving bowl - 58 EUR Your cart total is 106 EUR. Each piece is thrown by hand in Porto, ready for morning coffees and meals shared around the table. If you’d like to take another look, you can return to your cart below. [Return to your cart] Warmly, Terra Ceramics ### Email 2 Send timing: 24 hours after cart abandonment Subject: Handmade for everyday moments Preview text: A few practical details about your mugs and bowl. Body: Hi there, Beautiful things should feel at home in everyday life. Your Speckled mugs and Serving bowl are handmade stoneware, and each piece is thrown by hand... Review note: All three emails meet the format and length constraints, use accurate product details and pricing, and reserve the discount and its explicit 48-hour validity window for email 3. |
| Lower-scoring example DeepSeek V4 Pro, run 3 | Here is the 3-email sequence for Terra Ceramics. --- Email 1 Timing: 1 hour after cart abandonment Subject line: Your Terra Ceramics picks Preview text: Just checking in - your hand-thrown pieces are waiting. Body: Hi there, I noticed you left a couple of beautiful things in your cart at Terra Ceramics. You chose two Speckled Mugs (24 EUR each) and a Serving Bowl (58 EUR). That combination is a staff favourite - the mugs for a quiet morning coffee, the bowl for a shared dinner. Each piece is thrown by hand in our Porto studio. They’re made to be used daily, so they’re both dishwasher and microwave safe. Your cart is saved for now. If you have any questions about the stoneware, just reply to this email. Warmly, Terra Ceramics --- Email 2 Timing: 24 hours after email 1 Subject line: Still thinking it over? Preview text: No rush at all. Your Speckled Mugs and Serving Bowl are here... Review note: The sequence meets the formatting, length, tone and discount constraints, but unsupported product and operational claims require editing before use. |
Across the three briefs, the avoidable errors were practical: ad variants that drifted into acne or clinical claims, titles and descriptions that did not stay within their limits, and email sequences that mentioned a discount before the allowed message. A model can write persuasive language and still fail the campaign if it misses one of those boundaries.
Put the hard rules next to the source facts. Ask for a final count, require the exact number of variants and state which message may contain an offer. For SEO, ask for the keyword and character limit before the prose. For ads, list allowed claims and prohibited claims in separate lines.
Create a Persona for each brand voice and keep the approved claims close to the brief. Use /research for supplied product evidence, /document to create a campaign brief and /summarize to reduce a long product page before writing variants. Use /image when the campaign also needs a visual direction. Compare two models on the same claim-heavy brief and keep the final approval with the marketing owner.
The selected work covers meta ad variants, seo title and meta description, abandoned cart email sequence. The chart is useful because it keeps the recommendation tied to those jobs instead of turning the overall ranking into a universal rule.



Use the winner as the first comparison, then check the runner-up on a real brief. A model that loses a few points may still be the better operational fit if it follows your house format with less editing.
The benchmark is a shortlist, not a reason to hand every job to one model. Pick a candidate in the model picker, then use Compare to run the same brief side by side. Look at the facts that survived, the format that came back and how much editing remains.
When the work repeats, create a Persona with the audience, house style, prohibited claims, approval rules and output format. Keep the source material in Keep and assign review work in Tasks. Krater provides 400+ models, so a team can keep one dependable choice for important work while testing another for a different format or turnaround.
Useful commands for this kind of work include /research, /summarize, /document, /image. Use them to organize source material, turn long notes into a brief, create a structured deliverable or prepare a visual direction. The operator still approves the final output.
The test covers 10 business tasks, 3 runs per model and a generation temperature of 0.7. The briefs cover Product description, Amazon listing bullets, Meta ad variants, Customer support reply, SEO title and meta description, Abandoned cart email sequence, Spreadsheet formula, Contract clause summary, Meeting notes to action items, Product photo prompt. Three independent judges were used: GPT-6 Astra, Claude Opus 5, Gemini 3.1 Pro. They did not see the name of the model they were reviewing, and a judge was left out when it came from the same company as that model. Each response was scored on 5 criteria from 1 to 5: correctness, brief, usefulness, clarity, ship_ready. The mean was rescaled to 0 to 100, then averaged across runs and tasks.
This is a focused view of 3 task scores inside a 10-task comparison. AI judges, English prompts, one month and default reasoning settings all shape the result. Treat it as evidence for a shortlist, then test the briefs that matter to your team.
GPT-5.6 Luna leads this group at 95.6. That is a useful starting point, but your own facts and approval rules should decide the final choice.
No. The scores come from controlled business briefs. Usage popularity is covered separately in the usage article.
A response can follow the facts and still need a warmer tone, tighter format or a final brand review.
State what the source does not say, require source-only claims and check every named feature before publishing.
Pick two models in Krater, run the same brief in Compare and save the approved instructions in a Persona.
Krater provides 400+ models across text, research, image, video, voice and other workflows.
For ad copy, email and SEO, start with GPT-5.6 Luna, compare it with GPT-6 Astra and keep the source facts visible through review. The best choice is the one your team can approve and ship without risky additions.