Buyers now ask an assistant before they ask a search engine.
Ask for the best linen shirt under 120€ and an assistant answers from a handful of sources it retrieved and trusted. Being one of those sources is a different job from ranking on Google, and most stores have done nothing about it.
A new channelyou cannot bid on, cannot buy your way into, and cannot see in your analytics.
Not a ranking. A citation.
Search hands the visitor a list and lets them choose. An assistant reads the sources, picks, and gives one answer. There is no second page, so the difference between being cited and being absent is the entire result. The work is making your store legible to something that reads rather than crawls: clear answers to the questions buyers actually ask, structured data that says what a page is, and enough mention elsewhere that you are corroborated somewhere other than your own domain.
Four steps, and you can only move two.
Retrieval and trust are the two you can move. Everything here aims at those.
“Best linen shirt under 120€ for hot weather?”
The question
Conversational, specific, usually with a budget attached
Your leverage
None
Retrieval
The model gathers sources it can read and parse
Your leverage
This is where the work happens
Trust
It weighs corroboration: reviews, mentions, consistency
Your leverage
Earned, slowly
The answer
One recommendation, no second page
Your leverage
None
The work, concretely.
- 01
Answer the buying questions in text
Sizing, fit, materials, delivery, returns, comparisons. If the answer lives in an image, a PDF or a tab rendered by script, it is not in the answer the assistant gives.
- 02
Structured data that is actually correct
Product, offer, review and organisation markup that matches what the page says. Wrong markup is worse than none: it teaches the model something false with your name on it.
- 03
Let the AI crawlers in, deliberately
A robots file and a sitemap that decide which agents may read what, plus a plain-text summary of the catalogue for the ones that prefer it.
- 04
Be corroborated off your own site
A claim only you make is a claim. The same claim on a review platform, a marketplace listing or a press mention is evidence.
- 05
Measure by asking
There is no ranking report here. We track a set of real buying prompts over time and record whether you appear, and how you get described when you do.
What we will not tell you about this.
Nobody can promise a position in a generated answer, and anyone who does is selling something they cannot deliver. The mechanisms are young, they change without notice, and there is no ranking dashboard to point at. What we can do is make your store readable and corroborated, then measure your presence across real prompts so you can watch it move rather than take it on faith.
We did this to our own site first.
This page, and every other page here, ships structured data, a machine-readable sitemap, an explicit policy for AI crawlers and a plain-text summary of what the studio does. It is the cheapest possible demonstration: you are reading the thing we would build for you.
Start with a free teardown.
We look at your store, name the five fixes worth the most, and send them over. No call needed to get it, and nothing owed if you take it and leave.