The WisWes blog
Practical writing on conversational commerce, AI chat assistants, and turning store traffic into sales.
Margeen
An LLM that hunts margin, in public. Can a free model on free SaaS find resellable products with real margin and ship them through Shopify? One chunk a day, for forty days.
- Day 4Day 4: Moving the search into a real Shopify app — and giving it a brain
- Day 3Day 3: One prompt, twenty-five searches — keyword expansion with Gemini
- Day 2Day 2: Building AI search across AliExpress
Magento
Guides for Magento 2 and Adobe Commerce teams — interview prep, performance and EAV deep-dives, and how to plug modern AI into a Magento storefront without rewriting the checkout.
- Magento Security Problems: The 2026 Bugs, and the Pattern Behind Them
- 45 Magento Frontend Developer Interview Questions — The Interactive Quiz
- 45 Magento Interview Questions — The Interactive Quiz
Shopware
Guides for Shopware 6 teams — interview prep, the DAL and plugin/app architecture, performance and caching, and how to plug modern AI into a Shopware storefront without rewriting the checkout.
- 45 Shopware Interview Questions — The Interactive Quiz
Vibe Code
Build real software with AI, even if you don’t code. You describe what you want in plain English; Claude Code writes and runs it. Step-by-step guides for founders and managers — starting with apps for your own store.
- How to Set Up Claude Code to Write Shopify Code for You (No Coding Background)
Shopify
Plain-language Shopify guides for merchants and founders — what a Shopify app actually is, how to build your first one even if you’ve never written code, and how to add AI to your store.
- 45 Shopify Backend Developer Interview Questions — The Interactive Quiz
- Life after checkout.liquid: building with Checkout UI Extensions
- 45 Shopify Frontend Developer Interview Questions — The Interactive Quiz
All articles
How to Add UCP to Shopware (There Is a First-Party Plugin)
Shopware does speak UCP. There is a first-party plugin that publishes the discovery document and serves the shopping capabilities over four transports, scoped per sales channel. That inverts the Magento story — the work is configuration and edges, not construction. Here is exactly what it gives you and what it does not.
Read article →
Where AI chat loses the sale: the WisWes conversation funnel
"The chatbot is working" is not a number. WisWes Reports breaks every conversation into eight steps, shows which one shoppers fall out of, and refuses to print a rate it cannot honestly compute.
Read article →
We Ran Chrome’s New AI-Agent Audit on the 15 Biggest Ecommerce Sites
Chrome quietly shipped a fifth Lighthouse category that scores your site for AI agents rather than humans. We pointed it at Amazon, Walmart, Temu and twelve others. Six of them returned 403 and refused to be measured at all — and not one of the fifteen has implemented the standard everyone is writing about.
Read article →How to Add UCP to Magento (There Is No Native Support — Yet)
Magento does not speak UCP. There is no module to enable and no checkbox in the admin — Shopify co-authored the Universal Commerce Protocol and ships it first-party, and everyone else adds it by hand. This is the bare minimum to make a Magento store buyable by an AI agent, written from the real 2026-04-08 spec.
Read article →How to Evaluate AI Assistant Quality Across Languages
An assistant that scores 9/10 in English is often a 5/10 in German, Portuguese, or Ukrainian — and an English-only dashboard will never show it. Here is how to measure quality per language, in production, with remote LLMs you do not control.
Read article →How to Generate AI Content in Shopify Using Context and Memory — at Minimum Agentic Cost
The cost of AI content is set by how many times the model runs, not which model you pick. A no-code n8n build, with the numbers.
Read article →How Shopify Auth Actually Works (4 Flows, One App)
"Shopify auth" isn't one thing. A real Shopify app authenticates in four different ways, each with its own token, its own signature scheme, and its own failure mode. Mix them up and you get the classic symptoms: all webhooks fail HMAC, the embedded admin shows a blank iframe, guest order lookup 403s forever. Here is the map, written from production code.
Read article →
How AI Chat Agents Process Product Images — Scoped to What Your Store Sells
A shopper photographs a product and drops it into the chat. A well-built agent answers three questions — is this one of ours, is it broken, what is like it — and refuses the fourth: a random image it declines to touch.
Read article →What is AEO? Why your best pages get skipped by AI answers
SEO gets you ranked. AEO gets you quoted. As buyers move from ten blue links to one synthesized answer, the pages that win are the ones a model can fetch, extract and trust — and most sites fail at step one without knowing it.
Read article →
How to A/B test a conversational AI agent
Your AI agent has dozens of knobs, and guessing which ones lift revenue is expensive. Here is how to test prompts and flows the disciplined way — and read the results without fooling yourself.
Read article →How to Price an AI Agent (So You Don’t Lose Money on Every Conversation)
The fastest way to lose money on an AI product is to charge a flat price for a cost that varies per conversation. Here’s how to find your cost floor, choose a pricing model, and keep your margin when usage spikes.
Read article →How to Build Custom AI Inside Magento (Without an External Service)
You trained a model. Now it has to run inside the store. This is the part nobody writes about: the two ways to build custom AI into Magento, when each one fits, and a complete worked example — a scikit-learn classifier ported to pure PHP and plugged into the admin reviews grid.
Read article →
Catching AI-Generated Fake Reviews — With a Model Small Enough to Ship Inside a Plugin
The newest review-spam problem is that the spam reads perfectly — a language model wrote it. The fix is not a bigger language model. We trained a 1.3 MB classifier that catches machine-written reviews at 94.6% accuracy and tells you why. Here is how, and the code.
Read article →Is Claude Fable 5 Actually Worth It? The Math Says Most Developers Should Wait
Fable 5 wins every benchmark. It also costs twice as much. So we did the thing the launch posts skipped — the actual math for a working developer. One documented day on Fable cost more than a whole month of Claude Max.
Read article →Yes, AI Is a Bubble. The Bubble Just Isn’t Where You Think.
“AI is a bubble” and “AI is worthless” are two different sentences, and almost everyone reads them as one. The bubble is in the money and the hype. The value is in a small, boring corner that keeps working no matter what the market does.
Read article →
The Question Nobody Answered: How Pre-Purchase Friction Quietly Kills Sales
A shopper on your product page likes what they see. Then a question forms — and there is no answer anywhere. So they leave to find a store that will answer, and they buy there. This is the costliest bounce of all.
Read article →
Why Customers Bounce: The 7 Reasons Shoppers Leave Without Buying
You read your bounce rate like it is one problem. It is seven. Each leaves the same footprint in analytics and each needs a different fix — and most stores spend a year fixing the wrong one.
Read article →
llms.txt is a sign nobody reads — and llm.pdf is even worse
You uploaded /llms.txt to your site. You felt productive. Here is the awkward part: every major AI provider has either said they will not read it, or quietly never started. The file is a placebo.
Read article →
AI for your product catalog, done right: 18 prompts that won't tank your Google rankings
Two years ago, Google quietly killed thousands of ecommerce sites overnight. The fix is not "stop using AI" — it is using it the way the survivors did. Here is the playbook.
Read article →
The Real Cost of One AI Conversation — and the Math Behind It
Everyone says AI chat is cheap. Cheap is not a number. This is the actual arithmetic behind one conversation — the tokens, the resend multiplier, and why two stores can pay 10x different for the same chat.
Read article →
30 Interview Questions for an AI Python Agentic Engineer
Hiring for agentic AI is hard: the field moves fast and a candidate who can recite the transformer architecture may still struggle to ship a reliable tool-calling loop. Thirty questions that test what actually matters on the job.
Read article →The Goal of MCP Tools in AI Ecommerce Agents
A language model alone can talk about your store. With MCP tools it can read your catalog, change a cart, and track an order — which is what shoppers actually want.
Read article →
What Is Conversational Commerce? A Practical Guide for Online Stores
Shoppers increasingly want to ask a question and get an answer — not dig through filters and FAQ pages. Conversational commerce turns that chat into the path to checkout.
Read article →