×5
growth in AI traffic to retail sites over one year
All expertise
SEO/GEO & agentic commerce
Get cited and bought by AI agents
AI traffic to retail sites has grown 5x in a year. L'Oréal signed with OpenAI to place its brands inside ChatGPT, where people now buy without leaving the conversation. We get your catalog ready for this new channel: AI-Ready product pages with GEO copy and structured markup, plus an ACP-compatible catalog feed.
- Air France: 3,200 pages rebuilt across 10 languages
- Groupe Seb: 12,000 product pages with Amazon A+ content
Reality check
Product discovery already happens through AI.
Your customers no longer start their search on Google: they ask an assistant, which answers by citing three or four sources and, increasingly, offers to buy without leaving the conversation. The question is no longer whether this channel will matter — it's who the assistants name when they answer.
2% → 30%
ChatGPT's share as a product-search platform, in two years (Brazil)
54%
of shoppers have already bought a product recommended by an AI assistant (Brazil, 2026)
Real case
In a recent audit for a leading pharmacy chain (1,250 stores), we tested 5 non-branded queries: 0 citations out of 40 for the chain, even in its own home city where it's the #1 player. 17 citations went to competing pharmacies. What's at stake is your brand's place in the answer layer assistants actually read.
Audit
Measure before you act
Most "GEO audits" amount to asking ChatGPT three questions on a Tuesday morning. Ours produces a measured, repeatable baseline and quantifies the revenue at stake.
You walk away with a competitive snapshot and a backlog prioritized by exposed revenue — whether you continue with us or not.
The methodology, in 6 points
- 200 priority queries, defined from your real search logs and weighted by sales
- 5 engines tested: OpenAI, Copilot, Gemini, Anthropic, Perplexity — their coverage varies widely
- Each query repeated 3 to 5 times: the same question gets different answers on different days; we report a frequency, not a single run
- In your customers' language and geography: a query run from Paris or New York returns very different results
- Your competitors tracked in the same run: a score only means something in relative terms
- Test categories vs. control categories: market drift is excluded from results, and the model version is logged on every run
The business deliverable: three scenarios (conservative / base / high) of incremental revenue from a citation gain, built on your own numbers. Raw data is delivered to you every month — you can verify or replicate it yourself.
Produce
Write product pages the way AI assistants read them.
AI engines don't read your pages like Google does. They look for answers in the first sentence, headings phrased as questions, structured specs, trust signals (reviews, verifiable expertise) and deep links to your products.
We bring your catalog and editorial content up to that standard: GEO copy generated by AI under human review, JSON-LD markup, structured FAQs, content hubs linked to SKUs.
What we check and fix (from the checklist)
- Product pages readable without JavaScript (server-side rendering)
- Structured markup: Product, Offer, Review, FAQ in JSON-LD
- Unique meta description per product
- AI crawlers allowed in robots.txt (GPTBot, ClaudeBot, PerplexityBot) and a properly written llms.txt
- Reviews and ratings visible on product pages: the strongest "recommend this" signal that exists
- Editorial content in answer format, segmented by use case and profile, published under an identifiable hub
Air France: 3,200 pages rebuilt across 10 languages, AI writing platform with direct publishing into the CMS.
Sell
When the agent wants to buy, your catalog needs to answer.
Since 2026, people buy inside ChatGPT: the ACP protocol (OpenAI × Stripe) lets agents read your catalog, compare offers and place orders, with the merchant keeping control of payment. Etsy is connected, and a million Shopify merchants are following. If your product feed isn't up to standard, the agent recommends — and sells — elsewhere.
What we deliver
- ACP-compatible catalog feed: titles, descriptions, prices, stock, images and eligibility flags in the expected format, generated automatically from your PIM/DAM
- Multi-country, multi-language variants of the feed, aligned with your markets
- Connection to your e-commerce platform (VTEX, Shopify, Magento…): product and article page builders integrated into your stack
- Preparation for agent-driven checkout (merchant-controlled) on your D2C site
Groupe Seb: 40+ product-sheet fields written by AI and adapted per channel (Amazon, Darty, brand site) via a RAG connected to their product data.
Measure
Proof, not promises.
Four measurement instruments. And part of our fee at risk. Most AI influence produces zero clicks: a customer advised by ChatGPT walks into a store or types your brand name directly. That's why we measure at four levels — and why we're willing to tie part of our fee to the outcome.
The 4 instruments
- Weekly GEO test: citations, rank and share of voice across the query set, monthly written report
- Crawler & fetcher logs: are your pages actually retrieved by the engines? Real-time demand signal
- GA4: a dedicated channel group for AI referrals — a floor, never a total
- Post-purchase question: the only instrument that captures zero-click and in-store influence — the number your CFO will believe
The commercial model: up to 30% of production fees placed at risk against jointly set targets, after the audit, once the baseline is established.
Start with the audit.
A measured baseline across your queries and your competitors, with the revenue at stake quantified. Delivered in 8 to 10 days.