# LUMA·E > LUMA·E is an AI-first ecommerce consultancy led by Leo Nguyen. We run AVOS — an > AI Visibility Operating System that makes ecommerce brands appear, parse, and > convert across ChatGPT, Perplexity, Google AI Overviews, Claude, and agentic > shopping — on top of the Shopify Plus and Magento 2 builds we ship. We work with > ecommerce brands directly and white-label for agencies serving them. > We also offer a free AI Ecommerce Audit tool that grades stores on 5 dimensions. ## Services - [AVOS — AI Visibility Operating System](https://luma-e.com/avos): Audit → Report → Implement → Operate. 2-week fixed-scope audit, not 8. Covers AI visibility — making your brand citable and parseable across AI search and agentic shopping. - [AVOS Care — Quarterly Correctness Retainer](https://luma-e.com/avos/care): Keeps a store reading right to AI agents as the catalog changes. Every quarter: re-run the 4-agent panel, check schema, feed, redirects and stock signals, fix the drift, and send a dated before/after report. Correctness only, not ranking or PR. Care Core 45h/$1,125 per quarter; Care Full 60h/$1,500. Shopify Plus & Magento 2. - [B2B Ordering Blueprint — Shopify Plus & Magento](https://luma-e.com/b2b): Fixed-fee $850 B2B ordering blueprint (2 weeks) → Shopify Plus / Magento build with ERP sync, per-customer pricing and quick order. Maps the real order flow, price rules and integration spec, then gives a fixed quote and dated timeline. Fee credited to the build; run from $750/month. - [B2B Ordering Blueprint (Tiếng Việt)](https://luma-e.com/vi/b2b): Blueprint đặt hàng B2B phí cố định $850 (2 tuần) → build Shopify Plus / Magento có đồng bộ ERP, giá theo từng khách, đặt hàng nhanh. Phí trừ vào build; vận hành từ $750/tháng. - [AI Ecom Build](https://luma-e.com/services/ai-ecom-build): Ground-up Shopify Plus + Magento builds with an AI-first delivery model. - [Lifecycle + CRO](https://luma-e.com/services/lifecycle-cro): Klaviyo lifecycle, retention, and conversion-rate optimization. - [AI Correctness Sprint](https://luma-e.com/ai-readiness): Fixed-price fix for stores where AI agents read the rendered text wrong (e.g. an in-stock item reported as backordered). Measure across 4 AI agents, ship five fixes verified in raw HTML, re-measure in 10 days. Correctness, not rankings. Quick Check free; Sprint $1,500 up to 5,000 SKUs; agency $850/store (min 3). Shopify Plus & Magento 2. - [Agentic Storefront Readiness Playbook (self-score)](https://luma-e.com/playbook): Free, ungated 8-point self-score for whether AI agents read a Shopify Plus or Magento 2 store's product record correctly — price, B2B/gated-catalog exposure, agent answerability, schema, stock truthfulness, cross-engine consistency, entity, and off-site carry. Five checks run in any browser; interactive agent simulator plus engineer-grade Shopify Plus and Magento 2 fixes. Fix the source of truth first, carry it off-site last. - [Free AI Audit](https://luma-e.com/audit): 5-minute AI visibility grader for your store. - [AI Agent Commerce Readiness Audit](https://luma-e.com/services/ai-agent-commerce-readiness-audit): See exactly what ChatGPT, Gemini, Perplexity and Claude say about your products — availability, price, attributes — scored on a fixed rubric, with a re-test after fixes. - [AI Visibility Consultant for Shopify Stores](https://luma-e.com/services/ai-visibility-consultant-shopify): Get your Shopify store cited in AI answers and read correctly by shopping agents. AVOS shipped as engineering on your existing stack. Oroca: 0 → 45% share-of-voice. - [Headless Shopify + Next.js — Vietnam engineering team](https://luma-e.com/services/headless-shopify-nextjs-vietnam): Headless Shopify and Magento storefronts on Next.js from a Vietnam engineering team — built for hard LCP targets and heavy B2B carts, with an honest take on when not to. ## AVOS Capabilities - [Answer engine optimization](https://luma-e.com/avos): Your brand appearing in ChatGPT, Perplexity, Google AI Overviews, and Claude search. Content, entity, and citation strategy engineered for LLM ingestion. - [Agent commerce readiness](https://luma-e.com/avos): Your products parseable and shoppable by AI shopping agents. Structured feeds, agent-friendly checkout signals, schema, knowledge graph. - [LLM brand monitoring](https://luma-e.com/avos): Track brand mentions, sentiment, and share-of-voice across AI answers. Competitive intelligence on the AI surface, not the SERP. - [Entity & structured data engineering](https://luma-e.com/avos): Schema, JSON-LD, knowledge graph, and product feed engineering treated as a first-class engineering problem — not a tag-manager job. ## Case Studies - [Kangar Wear — fashion brand built from zero on Shopify](https://luma-e.com/work/kangar-wear-shopify-build): A US founder with a brand idea and no store. We built the full Shopify storefront from scratch — men's and women's lines, collections, checkout — and shipped it live. - [Puff Puff Pass It — built from scratch, then scaled](https://luma-e.com/work/puff-puff-pass-it-shopify): Built the Shopify store from the ground up, then kept enhancing it — large catalog, age-gating, free-shipping logic, and 3,000+ verified reviews live on site. - [Bird Pick Tea — migrated off Magento onto Shopify](https://luma-e.com/work/bird-pick-tea-magento-shopify): Replatformed an established tea brand from Magento to Shopify — full catalog migration, multi-currency, and a cleaner storefront the team can actually run. - [Mrs. Tea — store enhancement for a functional tea brand](https://luma-e.com/work/mrs-tea-shopify-enhancement): Enhanced an existing Shopify store for an Australian functional-tea brand — sharper storefront, wholesale portal, and subscription, tuned around how customers buy. - [Homelegance USA — built from scratch + SEO](https://luma-e.com/work/homelegance-usa-build-seo): Built a large-catalog furniture store from the ground up and laid the SEO foundation — structure that holds up across thousands of products. - [Home Elegance USA — four AI agents said in-stock furniture was backordered. It wasn't.](https://luma-e.com/work/home-elegance-usa-ai-agent-commerce-readiness-2026): Every AI shopping agent — ChatGPT, Gemini, Perplexity, Claude — quoted this 30,855-product Shopify store's furniture as backordered when it wasn't. Ten days, five fixes, all four agents stopped. Three now confirm "In stock." - [Oroca — built AI-visible from day one: 0 → 45.0% share of voice](https://luma-e.com/work/oroca-ai-visibility-0-to-41-share-of-voice-2026): A B2B dental-materials distributor, built from scratch to be AI-visible, now holds 45.0% share of voice in its category — the outright #1, more than the next three rivals combined and about 3.3× the nearest — measured weekly on a dashboard we built. - [Due Toscani — a Prato cashmere factory, engineered to be read by AI](https://luma-e.com/work/due-toscani-b2b-cashmere-ai-visibility-2026): A B2B private-label cashmere manufacturer in Prato, engineered to be readable by AI shopping assistants. Measured across ChatGPT, Google AI Mode and Gemini, it's named #1 for core buyer prompts and appears in 61–67% of category answers — described accurately, with zero hallucinated facts. - [DreamFly Edu: from invisible to named by ChatGPT](https://luma-e.com/work/dreamfly-edu-vietnam-australia-education-ai-visibility-2026): A Vietnamese Australia-focus study-abroad and migration consultancy, invisible to AI assistants at baseline — 0 of 40 buyer-question cells. Six weeks later ChatGPT names and correctly describes it for 6 of 18 core buyer questions — while Google's AI answers haven't caught up yet, and we say so. Measured the same way, before and after. - [Glamour Luxe Beauty — salon site build (VN)](https://luma-e.com/work/glamour-luxe-beauty-build): A nail & lash salon in Ho Chi Minh City needed a real website. LUMA-E hand-built it ground-up on Next.js, in Vietnamese — service menu, price list, gallery, and a clear booking/contact path — shipped live. - [VA Glamour Beauty — new Australian beauty store, built from scratch](https://luma-e.com/work/va-glamour-beauty-build): A second beauty store built from the ground up for the Australian market — storefront, catalog, and checkout, shipped live. - [NexAuto — full build cycle for a UAE car-service platform](https://luma-e.com/work/nexauto-uae-platform): Owned the full build cycle for a Dubai car-service brand — UI/UX through integration, a booking system, and service-contract flows. Pickup-to-delivery car care, online. - [Diviners — design system + build with CRM and LLM under the hood](https://luma-e.com/work/diviners-platform-crm-llm): Designed and built the platform from scratch — a full design system on top, with CRM and LLM integrations underneath supporting events, content, and community across EN/VI. - [ANLAP — large corporate / B2B build for a Vietnamese manufacturer](https://luma-e.com/work/anlap-b2b-vietnam): A large corporate B2B build for one of Vietnam's established manufacturers — structured to present decades of capability, services, and capacity to industrial buyers. - [Ranpolaris — travel-booking site for a Da Nang tour operator](https://luma-e.com/work/ranpolaris-travel-booking): A travel-booking site for a Da Nang tour operator — tours, private transfers, car rental and activities, with multi-currency and multi-channel contact built for international travellers. - [TL Portraitz — Wix site for a Canadian wedding photographer](https://luma-e.com/work/tl-portraitz-wix-wedding): A Wix site for an Edmonton wedding photography and videography studio — portfolio, packages, and booking built to win couples planning their day. - [Siam BioLife — launch site for a Thai cannabis B2B startup](https://luma-e.com/work/siam-biolife-thailand): A launch site for a new Thai medical-cannabis B2B startup — built to present cultivation, compliance, and capacity credibly to partners and buyers in a regulated category. - [Medicinal cannabis group — one corporate site for a multi-entity, seed-to-sale operation](https://luma-e.com/work/medicinal-cannabis-group-wordpress-corporate-site): A plant-based medicine group runs cultivation, GMP production and EU distribution through several operating entities. We built the corporate site that carries the whole seed-to-sale story in one place — for distributors, partners and investors, in a category that cannot advertise or sell online. - [Elkicken — WooCommerce build & customization for a Swedish e-mobility brand](https://luma-e.com/work/elkicken-woocommerce-build): Built and customized the WooCommerce store for a Swedish e-scooter and hoverboard brand — product storytelling, spec-led product pages, and a checkout tuned for the Nordic market. - [Powrplus — Shopify brand built from UI/UX through maintenance](https://luma-e.com/work/powrplus-shopify-lifecycle): Built a Shopify brand from the ground up — UI/UX, store build, and ongoing maintenance. The brand performed so well the owner sold it, and is now back to build the next one. - [Vivian — Multi-Brand Beauty Web Build + Social Retainer Across AU and Vietnam](https://luma-e.com/work/vivian-multi-brand-beauty-web-social-au-vn): Australia-based multi-brand operator — one-off Astro plus Decap web build for a launch beauty brand, plus an ongoing social retainer across a beauty salon brand in Ho Chi Minh City and an education brand. Three brand tracks, one shared memory layer. - [Explore To Win — one tangled catalog rebuilt into a clean, trackable system](https://luma-e.com/work/explore-to-win-shopify-catalog-architecture): A language-learning publisher's Shopify catalog had 12 products with messy tracking. We're re-architecting it into 4 lines × 3 formats behind a variant-style selector, with per-line reviews and a rebuilt ad-tracking layer so every format has its own identity. Build in progress. - [Due Toscani — Italian cashmere brand, lifecycle email + storefront redesign](https://luma-e.com/work/due-toscani-shopify-lifecycle-redesign): An Italian cashmere D2C on Shopify had no lifecycle marketing and dated pages. We built bilingual (IT/EN) abandoned-cart + checkout flows in Klaviyo, new-product push campaigns, cut duplicate native automations, and redesigned the homepage + PDP with a colour-swatch selector. - [EU B2B wholesaler — moving off two WordPress sites onto one Shopify platform](https://luma-e.com/work/eu-b2b-wholesaler-wordpress-shopify-replatform): A Dutch/German B2B wholesaler ran two separate WordPress storefronts. We're rebuilding onto a single Shopify platform — unified structure, cleaned menus/footers, full 301-redirect mapping, multi-market NL/DE storefronts and wholesale signup — staged on subdomains ahead of cutover. - [B2B distributor: 15,000 SKUs, ERP orders without re-typing](https://luma-e.com/work/b2b-distributor-erp-ordering-15000-skus-2026): A B2B supplies distributor moved trade ordering onto one storefront that prices per customer and hands every order to its ERP and accounting software, checked before it lands. ~15,000 SKUs, 5 price tiers, ~500 orders a month. ## Giải pháp (Vietnamese solution pages) - [Hệ thống đặt hàng B2B tích hợp ERP cho doanh nghiệp phân phối](https://luma-e.com/vi/giai-phap/he-thong-dat-hang-b2b): Hệ thống đặt hàng B2B tích hợp thẳng SAP/MISA và CRM: giá riêng theo đại lý, đơn xuất sang ERP, tách đơn khi thiếu hàng. 8 lớp vận hành đang chạy thật. - [Tích hợp Magento 2 với SAP ERP — từ CSV thủ công đến API, theo chi phí tối thiểu](https://luma-e.com/vi/giai-phap/tich-hop-magento-2-voi-sap): Tích hợp Magento 2 ↔ SAP S/4HANA theo bậc chi phí: CSV thủ công hết nhập tay ngay, rồi mapping có preview health, rồi API. ~500 SO/tháng đang chạy thật. - [Xuất đơn hàng từ website sang SAP — có mapping SKU, khách, đơn vị tính](https://luma-e.com/vi/giai-phap/xuat-don-web-sang-sap): Xuất đơn web → SAP với mapping SKU/khách/đơn vị tính, Incoterms 2 bước, phí ship, VAT công ty và preview health chặn đơn sai trước khi đẩy. - [Giá riêng theo từng đại lý trên website đặt hàng B2B](https://luma-e.com/vi/giai-phap/gia-rieng-theo-dai-ly): Shared catalog nhiều tier + giá riêng theo từng công ty: đại lý đăng nhập thấy đúng giá hợp đồng, ẩn giá với khách chưa duyệt, nhất quán PDP/PLP/cart/checkout. - [Đơn thiếu hàng → tách đơn, giao một phần, giữ phần chờ hàng về](https://luma-e.com/vi/giai-phap/don-thieu-hang-tach-don): Xử lý đơn B2B thiếu hàng: tách đơn giao phần có, giữ phần backorder, import tồn top-up khi hàng về, giao một phần khớp công nợ, khách thấy rõ trạng thái. - [Xuất đơn hàng từ website sang MISA để làm hoá đơn (kèm VAT GTGT)](https://luma-e.com/vi/giai-phap/xuat-don-web-sang-misa): Xuất đơn web → MISA làm hoá đơn GTGT: gộp dòng theo sản phẩm cha, thuế per item cho bundle/sản phẩm con, MST từ hồ sơ khách. ~500 đơn/tháng đang chạy. - [Dịch vụ AI Visibility cho ecommerce — xuất hiện đúng chỗ người mua đang hỏi](https://luma-e.com/vi/giai-phap/dich-vu-ai-visibility-cho-ecommerce): Agency kỹ thuật chuyên AI visibility cho ecommerce: xuất hiện đúng trên ChatGPT/Perplexity/Gemini, đọc đúng dữ liệu sản phẩm. Oroca 45% SoV, HE 4/4 sai → 0. - [Dịch vụ xây dựng ecommerce — Shopify B2B, Magento 2, headless NextJS](https://luma-e.com/vi/dich-vu/xay-dung-ecommerce): Dịch vụ xây dựng và migration ecommerce: Shopify B2B, Magento 2, headless NextJS. Một kỹ sư senior cộng AI production. Hệ B2B 8 lớp, 15.000 SKU cho một nhà phân phối. Từ $5.000. - [Dịch vụ Lifecycle và CRO — Klaviyo ra doanh thu, thử nghiệm có kiểm chứng](https://luma-e.com/vi/dich-vu/lifecycle-cro): Dịch vụ Lifecycle và CRO trên Klaviyo: kiến trúc 12 flow, phân khúc RFM, CRO có kiểm chứng. 3 năm cho nhà bán lẻ top 3 VN, +38% doanh thu email 90 ngày. Từ $3.000/tháng. - [Cổng đại lý riêng cho từng nhà phân phối, và cho sale tạo đơn hộ khách](https://luma-e.com/vi/giai-phap/cong-dai-ly-sale-tao-don): Cổng đại lý có workspace riêng cho từng dealer, sale tạo đơn hộ khách trên CRM, Magento tính đúng giá hợp đồng. Đồng bộ CRM ↔ Magento ~1.000 bản ghi/ngày. - [Khuyến mãi B2B mà Magento không làm sẵn: mua X tặng Y, BOGO, quà theo đơn](https://luma-e.com/vi/giai-phap/khuyen-mai-b2b): Engine khuyến mãi B2B trên Magento: mua X tặng Y, BOGO, mua 4 tặng 1, quà theo mốc đơn, áp đúng cả đặt hàng nhanh và đặt lại, đo từng chương trình. 3–5 chương trình/tháng. ## Partners - [Agency Partner Program](https://luma-e.com/partners): White-label AI visibility audits and implementation for agencies with an existing ecommerce install-base. Partner keeps the client relationship; LUMA·E runs the back end. Sample audit report available on /partners — no email gate. Also shipped for: BioLife Group, Medivita, Dragonfly APAC. ## Writing - [Writing — index](https://luma-e.com/blog): Long-form on AI search, ecommerce replatforming, and the operating model behind AVOS. - [Legibility Gets You Read. Proof Gets You Picked.](https://luma-e.com/blog/legibility-vs-corroboration-ai-commerce): An AI shopping agent first has to read your product page correctly, then find outside sources that back it up. Most stores fail the first test and pay to fix the second. - [Why B2B orders break at the ERP export (and fix)](https://luma-e.com/blog/b2b-order-erp-export-health-check-2026): The hardest part of B2B ordering isn't the catalog. It's getting each web order into the ERP without re-typing. Here's the health-check pattern that stops bad orders at preview instead of at month-end. - [StyleSmuggler: a 2-minute check to see if your Magento store is patched](https://luma-e.com/blog/magento-stylesmuggler-2-minute-check): CVE-2026-75650 has been exploited since 4 September 2026. Here is how a store owner can tell, without a developer, whether the Adobe hotfix VULN-39341 is live on their Magento 2 or Adobe Commerce store, what the signs of a compromise look like, and what to do in which order. - [How a Migration Makes You Invisible to AI](https://luma-e.com/blog/migration-redirect-map-ai-visibility): When you replatform, every old URL an AI model learned becomes a dead end unless you redirect it. Here is the redirect-map + structured-data gate that keeps an AI agent able to find and cite your store after the move — with real counts from a Woo to Shopify B2B migration. - [A B2B Catalog Built to Be Read by AI From Day One](https://luma-e.com/blog/b2b-catalog-built-agent-visible-from-day-one): Most B2B stores retrofit AI visibility after the agents already skip them. Here's what a catalog built agent-visible from day one looks like, and the share of voice it produced. - [The Migration Number Nobody Reconciles](https://luma-e.com/blog/migration-reconciliation-count-deltas-prestashop-shopify): A "successful" PrestaShop→Shopify import can still silently drop 85% of your customers. Here's the reconciliation gate — count deltas, archived-vs-active, and a clean-or-dirty verdict — that catches it before cutover. - [How to Measure Your Share of Voice in AI Answers](https://luma-e.com/blog/how-to-measure-share-of-voice-ai-answers-2026): Share of voice tells you whether AI assistants name you when buyers ask for a supplier — before Google. Here's the reproducible way to measure it: prompts, engines, scoring, and one real result. - [When Two Schema Graphs Collide, AI Drops Your Brand](https://luma-e.com/blog/competing-schema-graphs-ai-drops-brand-2026): Your theme and your SEO app can each emit their own JSON-LD. Agents read one graph — and if it's the one without your brand, they describe you as someone else. - [Why AI Agents Say You're Out of Stock When You're Not](https://luma-e.com/blog/ai-agents-wrong-stock-availability-structured-data): Shopping agents read your availability from server HTML and JSON-LD — not the 'In stock' badge your JavaScript paints in. When those disagree, agents tell buyers you're sold out. Here's where availability actually lives and how to make it read right. - [0 to 45% Share of Voice in AI Answers: A Case](https://luma-e.com/blog/ai-visibility-share-of-voice-case-oroca): How a B2B dental-materials distributor, built to be AI-visible from day one, went from zero to 45.0% category share of voice — and how we measured every mention. - [The Cheapest AI-Readiness Fix Is a Redirect Map](https://luma-e.com/blog/redirect-map-cheapest-ai-readiness-fix-ecommerce): Before you touch schema or content, fix the URLs AI agents can't resolve. Broken and drifting links are the quietest reason an agent drops your store — and the cheapest to fix. - [Magento Shared Catalogs Confuse AI Shopping Agents](https://luma-e.com/blog/magento-shared-catalog-customer-group-pricing-ai-agents): Shared catalogs and customer-group pricing show the right price to logged-in B2B buyers — and the wrong one, or none, to the AI agent reading your public pages. Here's the no-rebuild fix. ## Viết (Vietnamese writing) - [Đọc được thì được nhắc. Có xác nhận thì được chọn.](https://luma-e.com/vi/blog/legibility-vs-corroboration-ai-commerce): AI shopping agent phải đọc đúng trang sản phẩm của bạn trước, rồi mới tìm nguồn bên ngoài xác nhận. Phần lớn store trượt bài đầu mà không biết, rồi lại chi tiền cho bài sau. - [Vì sao đơn B2B vỡ ở bước đẩy vào ERP — và cách chặn](https://luma-e.com/vi/blog/b2b-order-erp-export-health-check-2026): Phần khó nhất của đơn B2B không phải catalog, mà là đưa từng đơn web vào ERP mà không gõ lại. Đây là mẫu health check chặn đơn lỗi ngay ở preview thay vì cuối tháng. - [Khi migrate làm store "tàng hình" với AI](https://luma-e.com/vi/blog/migration-redirect-map-ai-visibility): Khi bạn đổi nền tảng, mọi URL cũ mà AI đã "học" trở thành ngõ cụt nếu không redirect. Đây là bước redirect-map + structured data giữ cho trợ lý AI vẫn tìm và trích dẫn được store sau khi migrate — kèm số thật từ một lần migrate Woo sang Shopify B2B. - [Một catalog B2B được dựng để AI đọc được ngay từ ngày đầu](https://luma-e.com/vi/blog/b2b-catalog-built-agent-visible-from-day-one): Phần lớn store B2B mới đi vá khả năng hiển thị với AI sau khi agent đã bỏ qua họ. Đây là một catalog được dựng agent-visible từ ngày đầu trông như thế nào, và share of voice nó tạo ra. - [Làm sao để website bán hàng được ChatGPT trích dẫn?](https://luma-e.com/vi/blog/lam-sao-website-duoc-chatgpt-trich-dan): ChatGPT, Perplexity và Gemini trích những trang đọc được rõ, tin được và khớp câu hỏi — không phải trang top Google. Ba thứ cần có, kèm số thật từ Oroca và Home Elegance. - [Con số không ai đối soát khi migrate](https://luma-e.com/vi/blog/migration-reconciliation-count-deltas-prestashop-shopify): Một lần import PrestaShop→Shopify "thành công" vẫn có thể im lặng mất 85% khách. Đây là bước đối soát — count delta, archived-vs-active, verdict sạch/bẩn — bắt lỗi trước cutover. - [Cách đo Share of Voice của bạn trong câu trả lời AI](https://luma-e.com/vi/blog/how-to-measure-share-of-voice-ai-answers-2026): Share of voice cho biết AI có gọi tên bạn khi khách hỏi mua hay không — trước cả Google. Đây là cách đo lặp lại được: prompt, engine, cách chấm, kèm một kết quả thật. - [Hai graph schema chọi nhau, AI đánh rơi thương hiệu](https://luma-e.com/vi/blog/competing-schema-graphs-ai-drops-brand-2026): Theme và SEO app của bạn có thể mỗi cái phát ra một bộ JSON-LD riêng. Agent chỉ đọc một graph — và nếu đó là graph không có thương hiệu bạn, nó mô tả bạn thành một ai khác. - [Vì sao AI nói bạn hết hàng trong khi hàng vẫn còn](https://luma-e.com/vi/blog/ai-agents-wrong-stock-availability-structured-data): Shopping agent đọc tình trạng còn hàng từ HTML server và JSON-LD — không phải cái badge 'Còn hàng' mà JavaScript vẽ vào. Khi hai cái lệch nhau, agent nói với người mua là bạn đã hết hàng. Đây là nơi availability thực sự nằm và cách làm nó đọc đúng. - [Từ 0 lên 45% share of voice trong câu trả lời AI](https://luma-e.com/vi/blog/ai-visibility-share-of-voice-case-oroca): Một nhà phân phối vật liệu nha khoa B2B, build mới để AI đọc được từ ngày đầu, đi từ 0 lên 45,0% share of voice — và cách chúng tôi đo từng lượt nhắc. - [Cách sửa AI-readiness rẻ nhất là một redirect map](https://luma-e.com/vi/blog/redirect-map-cheapest-ai-readiness-fix-ecommerce): Trước khi động tới schema hay content, hãy sửa những URL mà AI agent không phân giải được. Link gãy và link trôi là lý do êm nhất khiến agent bỏ store của bạn — và cũng là cái rẻ nhất để sửa. - [AI agent đọc sai giá shared catalog Magento](https://luma-e.com/vi/blog/magento-shared-catalog-customer-group-pricing-ai-agents): Shared catalog và customer-group pricing hiện đúng giá cho buyer B2B đã đăng nhập — và hiện sai giá, hoặc không giá nào, cho AI agent đang đọc trang public của bạn. Đây là cách sửa không cần rebuild. ## News - [AI Commerce Watch — decision-ready briefings](https://luma-e.com/news): What changed in AI commerce and what to do about it. Each item: what changed (with a primary source), so-what for your product record, and one action. Updated when it matters, not on a schedule. - [2026-09-28 — Blocking "AI bots" is now blocking the ones that would cite you](https://luma-e.com/news): You can keep your content out of model training and still be eligible to get cited, but only if you let the search bots in. Roughly a quarter of ecommerce sites are accidentally blocking these at the CDN and don't know it. If the citation bot can't reach your page, you're not in the answer, full stop. - [2026-09-28 — Shopify Plus now exposes your catalog to AI agents by default](https://luma-e.com/news): The agent reads your live product record, not a marketing page. Whatever sits in the variant, the inventory field and the metafields is what it uses to answer and to build the cart. A gap there, like a missing size or a wrong stock count, becomes the agent's answer. - [2026-09-28 — Perplexity's merchant program charges 0% and funds the shipping itself](https://luma-e.com/news): A clean, current feed is the whole entry ticket. If your price, availability and variants are right, you're eligible to be the recommended product for free. If they're stale, you're either invisible or Perplexity quotes a wrong number to a buyer who's one click from paying. - [2026-09-28 — ChatGPT recommends from your structured feed now, through an open protocol](https://luma-e.com/news): Whatever your product record states in clean structured form is what an agent quotes and ranks on: price, stock, the variant, the answer to a question a buyer actually types. If that lives in description prose or a spec PDF, the agent hedges or picks the store that said it plainly. - [2026-09-28 — Google turned your Merchant Center feed into an AI answer source](https://luma-e.com/news): Your feed used to be an ads input. It's an answer input now. A feed carrying last week's price, or an availability that's already wrong, gets read back to shoppers at scale. Wider reach, wider wrong. ## About - [About Leo](https://luma-e.com/about): 10+ years on Shopify Plus + Magento, 50+ projects, runs LUMA·E with an AI-first operating model. - [Contact](https://luma-e.com/contact): Run a free 5-minute AI visibility audit.