US retail sites aren’t machine-readable to be visible to AI search at all
I'm BACK!!
Hey y’all,
It’s been a MINUTE and I have to deeply apologize. The last month or so has been bonkers (in the best way) and I had to make some prioritization choices. Even though I truly love writing this newsletter and I am so happy to be back, a founder only has so many arms.
A couple reasons why I’ve been off the Substack grid: we brought on x3 new brands in the last month to Praize for GEO and agentic-commerce work. Super excited to welcome Dame, Poppy & Peonies, and (soon to be announced) to the Praize roster. I also took my mom to Sweden and when we got back to Paris she helped me plant a bunch of flowers on my terrace. One of the highlights of my year so far.
We’re back to weekly now, and the timing is useful, because we got some real juicy updates in the AI Search landscape for ya. To set the scene before we dive in:
Harvard Business Review published the first credible empirical research showing that the merchandising tactics most DTC brands default to (scarcity badges, countdown timers, strike-through pricing, urgency vouchers, bundle nudges) don’t reliably move AI shopping agents, and the more advanced the model, the more it appears to resist them.
Amazon retired the Rufus brand and rolled the assistant into “Alexa for Shopping” — a unified, search-bar-default agent with one-year price history, scheduled-action carts, and cross-device memory — and pushed the whole thing live across the website, the app, and every Echo in the country, six weeks before Prime Day.
Read together, the two stories are one story: the buyer is changing, the surface where the buying happens is changing, and the merchandising playbook that ran 2018–2024 is now visibly losing leverage.
“Alexa for Shopping” is the moment agentic commerce stopped being a beta
Amazon did a major rebrand combining Rufus and Alexa+ — two assistants that mostly did similar things, sat in different shopping contexts, and never shared memory. Welcome “Alexa for Shopping,” which lives in the Amazon search bar on the web, in the app, and on every Echo device. No Prime required. No Echo required. Free to every U.S. customer, four to six weeks before Prime Day.
The capability stack rolled out:
A unified search bar that auto-routes natural-language questions (”good men’s skincare routine,” “Breville Barista Express vs Pro,” “where’s my order”)
Multi-product comparison directly from search results
AI overviews on PDPs and SERPs
One-year price history on hundreds of millions of products (up from 30–60 days)
Scheduled Actions — prompt-driven recurring carts and price-trigger purchases like “add this sunscreen to my cart if it drops below $10 and I haven’t bought it in two months”
One-click reorder from past purchases
Editable preference profiles
The previous Rufus surface drove an estimated $12B in incremental annualized sales in 2025, customers using it were 60% more likely to complete a purchase, and Alexa+ purchases ran at 3x the rate of original Alexa. Now Amazon has consolidated both behind a single, default brand in the most valuable real estate in e-commerce: the Amazon search bar.
ChatGPT, Gemini, Perplexity, and Claude all want to do product discovery directly. And those same agents are starting to absorb the adjacent jobs (trip planning, household management, health, meal planning, kids’ birthdays) that Amazon has historically been blind to, because Amazon’s been a search-and-buy surface, not a personal assistant. The Alexa rebrand is Amazon telling the market it plans to defend the context layer — and that’s the version of agentic commerce that gets monetized hardest, because the assistant is choosing what gets recommended before the user types anything that looks like a search query.
