Agentic Commerce Optimization by Industry: Playbooks for Fashion and Luxury Ecommerce, Jewelry, Beauty, Home, and Food
What AI shopping agents need to see in five categories, from luxury ecommerce and jewelry SEO to beauty, home goods, and food, with real customer results.
The short answer
An ACO industry playbook is the list of product attributes and shopper intents an AI shopping agent needs to match and recommend products in one category, along with the compliance rules that apply there. Fashion and luxury ecommerce get matched on fit, fabric and provenance, jewelry on metal and stones, beauty on ingredients for a skin type, home on dimensions, food on origin and allergens.
ACO by industry in 30 seconds
AI shopping agents read the same kinds of fields in every store. Which fields decide the sale depends on what you sell, from fit and provenance in fashion and luxury ecommerce to allergens in food.
- 01
Agents filter before they rank. Miss the deciding attribute and good copy won't save you.
- 02
The deciding attributes change by category. Fashion turns on fit and fabric. Jewelry turns on metal purity and the stone. Beauty is ingredients against skin type, home is centimeters, food is origin plus allergens.
- 03
An industry playbook is the short list of attributes and shopper questions an agent needs in your category, written into every product page and every feed.
- 04
In regulated categories, what the law makes you disclose (lab-grown stones, cosmetic claims, food allergens, pharmacy accreditation) is also what agents trust most.
- 05
The gap is rarely knowledge. The detail sits in a spec sheet or a buyer's head and never reaches the product detail page (PDP).
What differs by industry: attributes, intents, rules
The mechanics of Agentic Commerce Optimization don't change between categories. What changes is which attributes work as hard filters, where shopper demand sits, and what the law or the platform makes you disclose.
Agentic Commerce Optimization works on the product data that search engines, ads and agents read. That data model is category specific. Google requires gender, age_group and color for apparel offers, plus size for shoes (Google Merchant Center Help). Meta lists more than 20 jewelry-only fields such as metal_stamp_or_purity (Meta for Developers).
When we audit a catalog, we read those specs before the copy, then run each page through our product page checklist for AI readiness.
of top-tier luxury customers used AI in recent purchase journeys
Luxury shoppers ask AI firstSource: Bain & Company and Comité Colbert, 2026of luxury-related AI prompts do not mention a brand
The product has to match on factsSource: Bain & Company and Comité Colbert, 2026of shoppers use generative AI for product recommendations
Beauty led FMCG in AI-driven searchesSource: Euromonitor Internationalhigher conversion for AI-referred retail visitors in July 2026
AI traffic arrives ready to buySource: Adobe AI Traffic Trends, Aug 2026AI citation readability of apparel versus grocery sites
Readiness varies by industrySource: Adobe AI Traffic Trends, Aug 2026of desktop apparel sites lack sufficient size information
The top fashion gapSource: Baymard InstituteThe five categories side by side
| Industry | Hard-filter attributes | Top shopper intents | Rules to respect |
|---|---|---|---|
| Fashion and luxury | Size, size system, fit, fabric composition, color, gender, age group | Occasion, fit for body type, season, care | Google apparel requirements, fiber composition labeling |
| Jewelry and accessories | Metal, purity, gemstone, carat weight, ring size, chain length | Gifting, occasion, budget, everyday wear, allergies | FTC Jewelry Guides on lab-grown stones and precious metals |
| Beauty and pharma | Ingredients, skin or hair type, size in ml, format, SPF | Concern (acne, dryness), routine step, sensitive skin | Cosmetic claims criteria, pharmacy certification, OTC rules |
| Home and garden | Dimensions, weight, material, finish, capacity, assembly | Room, space constraints, outdoor use, style | Product safety information, consistent units |
| Food and beverages | Origin, ingredients, allergens, dietary labels, weight or volume | Dietary need, pairing, occasion, gifting | EU food information rules for distance selling, allergens before purchase |
We'd rather see a category's five hard filters filled on every product than fifty optional fields filled on half. An agent that can't confirm a ring size drops the ring. Once the filters are filled, a product description template for each category keeps the copy consistent across thousands of SKUs.
How to build your industry playbook
- 1
List the filter attributes
Start from your channels' category specs: Merchant Center, Meta catalog categories, OpenAI's feed spec, Shopify category metafields.
- 2
Map the intents
Pull real questions from site search, support tickets, reviews and keyword data.
- 3
Write the category FAQ
Answer the top questions briefly on every product page where they apply.
- 4
Add the compliance layer
Put what you must disclose, and must never claim, into your writing rules.
- 5
Fill and publish
Complete the attributes in your PIM or store and push identical values to the PDP and every feed.
- 6
Measure per category
A fashion catalog can be complete on color and empty on fit.
To find your gaps, put your attributes next to the questions your shoppers actually ask.
Fashion and luxury: the ACO playbook for fashion ecommerce SEO and AI shopping
Fashion brands get ready for AI shopping by writing down what a screen can't show: fit, fabric, how sizes run, when you'd wear it. Luxury ecommerce adds the provenance that justifies the price.
Take a prompt like "a navy wool blazer for a summer wedding, slim but not tight, under $600." A PDP that says "Navy Blazer, timeless elegance" gives the agent a color. It can't confirm wool, the cut, or a cloth light enough for June.
Adobe rates apparel sites highest for AI citation readability, at 76% in July 2026 (Adobe). Depth is the problem. Baymard finds that 83% of desktop apparel sites do not give enough size information (Baymard Institute). We see it constantly: a size chart saved as an image, no model measurements.
Attribute table for fashion
| Attribute | Feed or field name | Example value |
|---|---|---|
| Color | color | Navy |
| Size and size system | size, size_system, size_type | 50, IT, regular |
| Gender and age group | gender, age_group | male, adult |
| Material | material | Wool/silk/elastane |
| Fit | product_detail or metafield | Slim fit, true to size |
| Occasion and season | Description, FAQ, metafield | Wedding guest, spring and summer |
| Pattern | pattern | Pinstripe |
| Care | Description, FAQ | Dry clean only |
Google's material attribute is required when a product varies by material, and it takes a primary material plus up to two secondary ones (Google Merchant Center Help). On your site, mark variants with ProductGroup and variesBy for size, color, material or pattern (Google Search Central). In Shopify, choosing the Shirts category exposes category attributes like neckline or fabric, plus target gender (Shopify Help Center).
Set size_system too, since an Italian 42 and a US 42 are different garments. Also watch for sizes sharing the parent's GTIN, and for Shopify category metafields your feed app never sends to Merchant Center.
Intents and FAQs agents need in fashion
- Fit and sizing: "Does it run small?", "What size is the model wearing?", "Will it work with broad shoulders?"
- Occasion: office, wedding guest, travel, black tie, weekend.
- Season and climate: fabric weight, breathability, warmth.
- Care and styling: washing instructions, pilling, what to wear it with.
What changes in luxury ecommerce
The usual advice is to keep luxury pages spare. We think that has aged badly. Some 82% of top-tier luxury customers used AI in recent purchase journeys, and around 70% of luxury prompts start without a brand name (Bain & Company and Comité Colbert). The same Bain study found that 90% of the URLs language models cite come from outside brand websites (Bain & Company and Comité Colbert).
Your page should be the most complete source an agent finds: where it's made, the origin and grade of materials, construction (hand-stitched lapels, full canvas), edition, repairs.
You keep the tone. You just stop hiding the facts behind it.
Case: Bomboogie
Italian outerwear brand Bomboogie optimized its product pages with AndromedAI and saw +1,800% clicks from AI chats such as ChatGPT, Gemini, and AI Mode. See the full Fashion and Luxury playbook.
Jewelry and accessories: jewelry SEO and the product attributes agents need
Jewelry ecommerce needs what a photo can't carry: metal and purity, the stone and its origin, carat weight, ring size or chain length, with lab-grown or plated materials disclosed plainly. Jewelry SEO is won mostly on gifting intents.
Here is the kind of prompt a title like "Aurora Necklace" can't answer.
It chose the page that stated purity and nickel content in text. "Gold tone" wouldn't make the list.
Product attributes for jewelry ecommerce
Meta's catalog spec is the fullest public list of jewelry fields we know. It recommends material, size and gemstone for every item and supports specific fields such as metal_stamp_or_purity, plating_material, gemstone_cut, gemstone_clarity, gemstone_creation_method, total_gemstone_weight, chain_length, clasp_type and occasion (Meta for Developers).
| Attribute | Example value | Shopper question it answers |
|---|---|---|
| Metal and purity | 18k yellow gold, 750 hallmark | "Is it solid gold or plated?" |
| Gemstone and creation method | Diamond, laboratory-grown | "Is the diamond natural or lab-grown?" |
| Cut, clarity, color, carat | Round brilliant, VS1, F, 0.50 ct total | "How big and how good is the stone?" |
| Size and size system | Ring size 6 US, 52 EU | "Will it fit? Can it be resized?" |
| Dimensions | Chain 45 cm, pendant 12 mm | "How does it sit on the neck?" |
| Certificates | Grading report, hallmark | "How do I know it's authentic?" |
| Care and allergies | Nickel-free, avoid chlorine | "Can I shower in it?" |
Sizing breaks across markets. US ring sizes are numbers, UK and Irish sizes are letters, many continental shops use millimeters. Store the measurement once, convert per market. And gold vermeil has a defined meaning under the FTC Guides, so never let it share a material value with solid gold.
Jewelry SEO and AI intents
- Gifting: anniversary, birthday, graduation, push present, "for her", "for him".
- Occasion and style: engagement, wedding band, everyday, stacking, minimalist.
- Budget: "under $300", "affordable diamond", "lab-grown alternative".
Gift-guide posts get most of the jewelry SEO attention. Yet many queries are category-level ("lab-grown engagement rings under $2,000"), and a filterable category page serves them better. Build those pages around actual search demand with the category page optimizer.
See the full Jewelry and Accessories playbook.
Beauty and pharma: ingredients, skin types and compliance
Beauty and pharmacy products get matched on ingredients and on who they're for (skin or hair type, concern). Every claim has to be true and substantiated, so compliance and AI readiness pull the same way.
Euromonitor says beauty and personal care led all FMCG categories for AI-driven product searches in 2025, that 28% of shoppers already use generative AI for recommendations, and that online reached 37% of global skin care sales in 2026 (Euromonitor International).
A typical prompt: "a fragrance-free moisturizer for sensitive, acne-prone skin." The agent checks what's in the jar, and what's left out, against a skin type. Sensorial copy sells to people, but agents don't feel textures.
Attribute table for beauty and pharma
| Attribute | Example value |
|---|---|
| Key ingredients and INCI list | Niacinamide 5%, ceramides |
| Free-from (only when true) | Fragrance-free, alcohol-free |
| Skin or hair type | Sensitive, combination, curly |
| Concern | Acne, redness, dryness |
| Format and size | Gel cream, 50 ml |
| Routine step and usage | Morning and evening, after serum |
| SPF | SPF 50, broad spectrum |
| Certifications | Dermatologically tested, vegan |
Google's product_detail attribute holds up to 100 technical details per product as section, name, and value, and warns against submitting a value that has not been confirmed (Google Merchant Center Help).
Two details we check on every beauty audit. The INCI list should be text, in pack order (descending concentration, with ingredients under 1% in any order at the end). And sun care wording needs localizing: "broad spectrum" is US labeling language, while EU packs use the UVA circle logo.
Intents and FAQs agents need
- "Is it safe for sensitive skin?", "Can I use it with retinol?" (answer only with what you can substantiate).
- "How long until I see results?", "How long does a 50 ml jar last?"
Compliance as a data rule
In the EU and UK, cosmetic claims must meet six common criteria: legal compliance, truthfulness, evidential support, honesty, fairness, and informed decision-making, in any medium (Regulation (EU) No 655/2013). For pharmacies, Google allows prescription drugs in Shopping only in the US and Canada, from accredited pharmacies, and many countries require pharmacy registration plus Google certification even for OTC products (Google Merchant Center Help).
Write these limits into your banned words, so nobody ships "heals eczema" at 6pm.
Case: Semprefarmacia
Italian online pharmacy Semprefarmacia optimized its catalog with AndromedAI and grew sales by 46.9% and organic traffic by 160%, saving 483 hours of catalog work. See the full Beauty and Pharma playbook.
Home and garden: dimensions, materials and room intents
Home and garden products are bought on fit and function. No agent can recommend a sofa for a small apartment unless the dimensions are written down in consistent units.
Adobe scores furniture and home sites at 64% for AI citation readability, below apparel and cosmetics (Adobe). Usually the data is trapped in PDF spec sheets.
For "an outdoor dining table for six that fits a 3 by 4 meter terrace," the agent must confirm it seats six, survives outdoors, and fits.
Attribute table for home and garden
| Attribute | Feed or field name | Example value |
|---|---|---|
| Dimensions | product_length, product_width, product_height; OpenAI dimensions | 220 x 100 x 75 cm |
| Weight | product_weight; OpenAI weight | 38 kg |
| Material and finish | material, product_detail | Teak, oiled finish |
| Capacity | product_detail | Seats 6 |
| Indoor or outdoor | Attribute, description | Outdoor, UV and weather resistant |
| Assembly | product_detail, FAQ | Assembly required, 30 minutes, 2 people |
| Style and room | Description, metafield | Mediterranean, terrace or garden |
Google's product dimension attributes accept cm or in, and weight in lb, oz, g or kg. Google asks for the same unit across dimension attributes, or the information will not be displayed, and for values consistent with product_detail (Google Merchant Center Help). OpenAI's feed prefers a single dimensions object with at least two of length, width and height plus a unit (OpenAI Commerce).
Common mistakes are mundane: shipping_weight and product_weight holding the same boxed weight, or a bare "220 x 100 x 75" where nobody can tell depth from height.
title
BeforeLido Table Teak
AfterCasavela Lido Outdoor Teak Dining Table for 6, 220 x 100 cm, Oiled Finish
description
BeforeBring Mediterranean style to your garden with our beautiful Lido table.
AfterSolid teak outdoor dining table for six, 220 cm long by 100 cm wide, 75 cm high. Stays outside all year with a yearly re-oil; two people assemble it in about 30 minutes.
product_detail
BeforeMaterial: teak
AfterMaterial: solid teak; Finish: oiled; Seats: 6; Use: outdoor, all year; Assembly: 2 people, 30 minutes
dimensions
Before(empty)
Afterproduct_length: 220 cm; product_width: 100 cm; product_height: 75 cm; product_weight: 38 kg
Intents agents look for in home
- Room and space: small apartment, studio, balcony, open-plan living.
- Use case: dining for six, home office, pets and kids, outdoor all year.
In home and garden, the dimension you didn't write down is the reason the agent picks someone else.
See the full Home and Garden playbook.
Food and beverages: origin, dietary attributes and pairing intents
Food and drink are chosen on origin and what's inside, then on dietary fit and occasion. In the EU most mandatory food information must be available before purchase, so a complete PDP is a legal requirement first.
Grocery is the least AI-readable retail sub-industry in Adobe's data, at 59% in July 2026 (Adobe). That leaves room for specialty brands, as agents already get asked for "an aged Pecorino for a wine tasting" or "a gluten-free panettone gift box delivered to Dublin."
Attribute table for food and beverages
| Attribute | Example value | Shopper question it answers |
|---|---|---|
| Origin and designation | Pecorino Romano PDO, Lazio | "Is it authentic?" |
| Ingredients | Sheep's milk, salt, rennet | "What's in it?" |
| Allergens | Contains milk | "Is it safe for my guest?" |
| Dietary labels | Vegetarian, gluten-free, organic | "Can I eat it on my diet?" |
| Net weight or volume | 1 kg, 750 ml | "How much do I get?" |
| Aging, vintage or ABV | Aged 12 months; 2021; 13.5% | "How mature or strong is it?" |
| Pairing | Pairs with Cesanese red wine, pears, honey | "What should I serve it with?" |
The legal baseline is also the data baseline
Under Regulation (EU) No 1169/2011, online sellers of prepacked food must show all mandatory food information before purchase, except the best-before or use-by date; non-prepacked food needs allergen information (European Commission). Ireland's food safety authority applies the same rule (FSAI).
Many shops still use a photo of the back label as the ingredient list. Agents can't read it. Use text.
Submit dietary attributes only when confirmed, never as a default, as Google's product_detail guidance says (Google Merchant Center Help). A wrong "gluten-free" is far worse than a missing one. For wine, remember alcohol is a restricted Merchant Center category with rules per country.
Intents agents look for in food
- Dietary needs: gluten-free, lactose-free, vegan, halal, low sugar.
- Pairing and recipes: wine pairing, cheese board, pasta recipe.
- Occasion: Christmas, aperitivo, corporate gifts, picnic.
Case: Matassa
Matassa used AndromedAI to generate complete product pages and got its first sales from ChatGPT within one week, with catalog generation costs down 95%. See the full Food and Beverages playbook.
Industry playbooks with AndromedAI
AndromedAI runs each industry playbook across the whole catalog: it scores every product against its category, fills the gaps in your brand voice and publishes to your store, PIM or feeds.
How the platform maps to the work
| Step | AndromedAI | What it does by industry |
|---|---|---|
| Find the gaps | AI Readiness Audit | Scores pages on Product Data Completeness, Keyword Coverage, Customer Intent Match and Shopping Metadata, showing missing fit, metal, ingredient, dimension or allergen data |
| Create missing pages | Creator | Builds complete product pages from brand or supplier data, including PDF spec sheets, in 12 languages |
| Rewrite for intents | Optimizer | Rewrites titles, descriptions, bullets, and FAQ, extracts attributes such as fabric, purity or dimensions, and adds use cases and intents |
| Cover category demand | Category page optimizer | Builds category pages around real demand, such as "lab-grown engagement rings" or "outdoor dining tables" |
| Publish | Integrations | Publishes to Shopify, Google Merchant Center, Salesforce Commerce Cloud, Adobe Commerce, Shopware, Akeneo, Plytix, and WooCommerce |
Each industry's rules live in the Brand Kit: tone of voice, examples, glossary, and banned words (such as "natural" next to a lab-grown stone). AI checks and approval workflows control what goes live, with auto-approval above an AI Checker score of 4.0. AndromedAI also generates Merchant Center conversational attributes, the product data Google introduced to match listings with conversational queries in AI Mode and Gemini (Search Engine Land). More than 500 catalogs have been optimized on the platform.
See also how to write product descriptions and product page optimization.
Industry checklist before you publish
- Category attributes complete
Every hard-filter field is filled
- Same values everywhere
Store, PIM, feeds, marketplaces all agree
- Intents covered
Occasions, use cases, audience named
- Disclosures written in
Lab-grown, plating, claims, allergens: all stated plainly
- Every market localized
Units, sizes, terminology adapted per country
Customer results
clicks from AI chats (ChatGPT, Gemini, AI Mode)
Bomboogiesales and +160% organic traffic, 483 hours saved
Semprefarmaciato the first sales from ChatGPT, and -95% catalog generation costs
Matassaadd-to-cart in two months, same marketing budget
Altaforma Milanobounce rate on product pages
AusiliumFAQ
Write down what shoppers can't see on screen: the fit, how sizes run, fabric composition and care, plus the occasion and season the piece suits. Fill the required apparel attributes (color, size, gender, age group), give each variant its own GTIN, and send the same values to your feed.
Start with metal and purity, plus any plating, because that is the first thing shoppers and agents filter on. For stones, state the type, creation method (natural or lab-grown), cut, clarity, color and carat weight. Then add ring size with its size system, chain length, clasp type, certificates, and allergy notes such as nickel-free.
Most luxury prompts in AI assistants don't name a brand, so products have to match on facts rather than on reputation. State materials and their origin, place of manufacture, construction details, edition and aftercare services such as repairs. Keep the brand tone, but make your product page the most complete source an agent can find.
Jewelry SEO is the work of making jewelry product and category pages rank in search and get recommended by AI assistants. Shoppers usually search with a metal or stone plus a style, an occasion or a budget, such as a lab-grown engagement ring under 2,000 dollars. Pages win when they state those facts in text and category pages match that demand.
List the key ingredients with the full INCI list, and free-from claims only when they are true. Add skin or hair type, the concern the product addresses, format, size, usage and SPF, plus certifications such as vegan or dermatologically tested. Every claim needs substantiation, because cosmetic claims rules in the EU and UK apply in any medium.
Agents need exact dimensions and weight in one consistent unit, because a sofa that can't be confirmed to fit gets left out. Add materials and finish, capacity, indoor or outdoor use, and assembly details such as time and number of people. Then name the rooms or spaces the product suits, such as a small apartment or a balcony.
For prepacked food sold online, all mandatory food information except the best-before or use-by date must be available before purchase, under Regulation (EU) No 1169/2011. That includes the ingredient list, allergens and net quantity, written as text rather than a photo of the label. Non-prepacked food needs allergen information before purchase.
Glossary
- Industry playbook
- The attributes, intents, questions, and rules that AI agents need to match products in one category
- Hard-filter attribute
- An attribute an agent uses to include or exclude a product, such as size, metal, skin type or dimensions
- Metal purity
- The share of precious metal in an alloy, shown as karats or a hallmark such as 750
- Lab-grown diamond
- A diamond created in a laboratory, which must be disclosed as such in marketing
- INCI
- International Nomenclature of Cosmetic Ingredients, the standard naming system for cosmetic ingredient lists
- product_detail
- A Google Merchant Center attribute for technical details, written as section, name, and value
- Conversational attributes
- Merchant Center product data used to match products with conversational queries in AI Mode and Gemini
Keep reading
The framework behind every industry playbook
GuideProduct Page Optimization: The Complete Guide to PDPs That Rank, Convert and Get Recommended by AIBuild product pages agents can match and justify
GuideProduct Page Checklist for AI ReadinessCheck each PDP against what agents need before you publish
GuideProduct Description Templates by Category (Free)Category templates that turn attributes into copy
Sources (18)
- Bain & Company and Comité Colbert: Luxury and AI (2026)
- Adobe: AI Traffic Trends Report, August 2026
- Euromonitor: Digital beauty and skin care online sales
- Baymard Institute: Apparel size information
- Google Merchant Center Help: Apparel attributes
- Google Merchant Center Help: Material
- Google Merchant Center Help: Product length
- Google Merchant Center Help: Product detail
- Google Merchant Center Help: Healthcare and medicines
- Google Search Central: Product variant structured data
- Meta for Developers: Jewelry and watches catalog category
- OpenAI Commerce: Product feed specification
- Shopify Help Center: Product category
- European Commission: Distance selling of food
- FSAI: Labelling and distance selling
- Regulation (EU) No 655/2013: Cosmetic claims
- JCK: FTC new guides for the jewelry industry
- Search Engine Land: Merchant Center conversational attributes
Parlano di noi








