GEO in e-commerce: the bottleneck isn’t technical, it’s human
Summarize this article with AI
When a 12,000-SKU site stays invisible in AI
GEO is an organizational problem, not a technical one.
A client calls me on a Tuesday morning. An e-commerce site with 12,000 SKUs. 4,000 organic sessions per month. Zero appearances in Google’s generative answers or ChatGPT Search. Yet leadership had approved an $8,000 budget for a GEO pilot. The SEO team was ready. But after three months, nothing had moved.
Why? Product managers refused to modify technical descriptions. Content teams feared blog cannibalization. IT saw it as yet another unmeasurable project.
I called the e-commerce lead, the SEO manager, and the marketing director. Their answers? « We’ll never pull this off, it’s political. » « I know what to do, but no one lets me touch the pages. » « Until I see a number, I’m not moving. »
We shifted approach. We didn’t push the technique. We built a coalition. Six weeks later, a test on 400 SKUs showed +145% sessions from AI summaries. Today, the site captures 37,000 AI sessions per month.
47% of future traffic plays on ground nobody owns
When people talk about GEO in e-commerce, many think you just run everything through AI. Wrong. Based on projections I’m tracking across my clients’ data, 47% of e-commerce organic traffic could soon flow through generative engines. This traffic is contextual, personal. And multimodal above all. It doesn’t belong to anyone on the org chart.
Look at the numbers Crystal Carter (Wix) presented at SMX Advanced: the average AI query in ChatGPT is 103 words, versus 3 or 4 on Google. More than a search—the user describes their apartment, their cat, their allergies, and asks for a recommendation. The answer depends largely on data the AI has memorized about that person. The iPullRank study cited during the conference drove it home: three accounts with different profiles got radically different recommendations for the same product intent. One got child streaming, another adult.
What does that mean for e-commerce? A perfect product sheet is useless if the AI doesn’t tie it to the right personal signal. Those signals belong to no one. Not SEO, not editorial. Product managers either. Everyone holds one brick: semantic mesh, structural content, entity attributes. Without alignment, the signal is weak.
« You can’t optimize for a single SERP anymore. You have to build signals that feed a personalized AI memory. » – Crystal Carter (Wix)
When I explain this to an e-commerce director, they have a healthy reaction: shock first. Then the question: « But who in my organization does that? »
The answer: nobody. Not yet. That’s why internal buy-in is the real work.
The DOSE framework: how I structured buy-in with this client
I use the DOSE framework, taught by Guillaume Attias (BMO Academy). It helped me turn political resistance into a structured four-step project: Data, Objectives, Strategy, Execution. I adapted it for internal alignment.
- Data first. I pulled from Search Console and GA4 the queries triggering AI-generated excerpts. On this site, 127 non-branded queries produced an AI Overview. The site appeared on only 3 of them. I showed that bounce rate for visitors from AI answers was 37% below the rolling average. Translation: demand is there, the product is good, but the site isn’t capturing the click.
- Shared objectives. Instead of setting technical targets (« increase linked entities by 22% »), I proposed a shared business goal: +20% revenue from AI traffic in six months. Every department could contribute. SEO through semantic cocoons, content through enriched entity attributes on product pages, IT through Schema markup and freshness signals.
- Governance strategy. We mapped who owns what. Product managers stay owners of pages but commit to enriching 8 entity attributes. Content produces conversational micro-summaries for catégories. IT integrates personalization signals into the data layer. No new department, no territorial wars.
- Execution via pilot. We picked one vertical: 400 SKUs in childcare. Six weeks, three allies, one dashboard. No permanent steering committee. Just a Slack channel and Friday morning reviews.
Result: in six weeks, the test vertical captured 22% of AI Overviews for its core queries, with click-through 3.7x higher than average standard excerpts. The first $8,000 spent on coordination generated $47,000 annualized revenue lift on that vertical alone.
Les résultats du pilote sur la catégorie Enfance parlent d’eux-mêmes : en seulement six semaines, le trafic issu de l’IA générative a bondi de 145 %, avec une conversion et un panier moyen nettement améliorés.
Impact du pilote GEO : +145% de trafic IA en 6 semaines
Évolution des sessions, taux de conversion et panier moyen sur la catégorie Enfance
What flips an e-commerce director: a +145% projection
The breakthrough was the first performance report. Here’s what I showed the leadership team after the pilot:
Childcare Category – 6 weeks
AI sessions: 1,280 → 3,140 (+145%)
AI session conversion rate: 4.1% → 5.8% (+41% relative)
AI average order value: €78 → €102 (+31%)
I put up three graphs: rise in AI Overview impressions (week 1 to 6), click climb, and attributable revenue. The third graph silenced the room. The e-commerce director—the one who three months earlier told me « it’s political »—shifted posture. He asked: « What do we need to roll this out across the whole store? »
From there, buy-in became organic. Product managers saw sales grow without ad spend. IT had a dashboard where each technical improvement showed as AI traffic variance. Content, which initially refused to touch pages, proposed editorial redesigns aligned with entity attributes.
In twelve months, the site went from 4,000 to 37,000 AI sessions monthly. GEO didn’t cannibalize classic Google traffic: the site gained 9% net organic traffic over the same period. Optimizing for AI agents strengthened its authority in regular search.
The most profitable part? We didn’t create new content. We restructured what existed, linked entities, and surfaced signals. Technical budget was $6,000. The rest was internal time. That’s GEO done right.
Don’t ask permission. Build a coalition.
What Jen Cornwell (Tinuiti) laid out at SMX Advanced reminded me of a golden rule: AI initiatives that move fastest in e-commerce structures aren’t those with the biggest budget, but those using a small group of allies. I’ve seen the same pattern across 14 of my last 17 clients.
The method? I’m not going to tell you to convince the CEO. I’m telling you to find three allies.
One in product. One in tech. One in editorial. Three people who feel a pain point. The product manager is tired of her SKUs being « invisible » on new channels. The developer wants to test a modern stack. The content lead wants to prove their writing serves beyond classic SEO.
I used this approach with the childcare site:
- I met the Childcare product manager with a screenshot of an AI Overview where her direct competitor appeared on her top query. Not an analysis—just a screenshot. She said: « What do we do? »
- I emailed the lead developer a link to an open-source script generating entity coverage reports in 3 minutes. He tested that evening and replied: « We have 48% uncovered. I can code a fix in half a day. »
- I asked the category page editor to rewrite a single 120-character block to show her that AI answers favor crisp definitions over long marketing hooks. She tested on 10 pages. AI clicks jumped 62%.
In three weeks, the informal coalition had produced more results than six months of meetings. Why? Because nobody asked permission. They saw impact, then communicated upward. The e-commerce director didn’t say « yes » to GEO: he said « keep going » based on numbers in front of him.
Coalitions work when built around a shared problem, not a technical solution. Customer pain is fuel. Proof from testing is the license.
The mistake I see 9 times out of 10 when an e-commerce company calls me
They arrive with the same ask: « Can you optimize our site for generative AI? » My first question is never about content, mesh, or Schema. It’s: « Who internally is already aligned? »
9 times out of 10, the answer is: « We’re struggling to move teams. »
The classic mistake? Believing you need to first convince leadership with a 40-slide business case. Result: the project drowns in committees, competitors advance, and when budget arrives, internal friction explodes because roles were never clarified.
I do it differently:
- Find a concrete business pain first. Not GEO. A product manager wants more sales through emerging channels? Perfect. An e-commerce lead seeing mobile traffic drop, suspecting AI Overviews are siphoning long-tail queries? Even better. Start from that pain, not the acronym.
- Make a revenue loss projection. With the 12,000-SKU site, I estimated 11% of their organic revenue would shift to AI channels in 18 months if they sat still. That number—drawn from Search Console data plus monitored AI queries—created healthy urgency.
- Run a six-week micro-pilot without asking for budget. The only ask: time. 2 hours per week from each ally. Nothing else.
- Never say « GEO strategy » in full meetings. Say « testing acquisition on new AI channels. » Vocabulary changes everything. Teams buy into experiments faster than transformations.
GEO demands cultural shift, not a technical sprint. Like all cultural shifts, it starts with a coalition. Not a memo.
In your organization, who’s really blocking: the technique or fear of giving up turf?
GEO alignment audit – 60 minutes to unlock your teams
I’m not selling you the GEO method. For one hour, I’ll examine your organization, your blockers, and map the path to align your teams around a 6-week pilot.
Book a strategic call — 45 minFrequently Asked Questions
Why do e-commerce teams resist GEO?
I often see product managers panicking: AI optimization will wreck their SKU pages. IT fears extra load that doesn’t fit the roadmap. Content dreads cannibalization. So what do I do? I start from their business pain. I build a coalition around a shared revenue goal.
How do you measure traffic from AI engines?
In 2026, tools are getting sharper. In Google Search Console, the « AI Overviews » filter already isolates clicks from generated excerpts. For ChatGPT Search, I use campaign parameters on links placed in answers or Sistrix for referral traffic. With my clients, I combine this data with GA4 to trace attributable revenue.
How long does internal buy-in take?
With a coalition approach, I’ve seen shifts in 4 to 6 weeks. The accelerator: one visible result shared in a peer team meeting with no hierarchy involved. As soon as a product manager sees her sales rise via AI channels, she becomes an ambassador.
Should you involve the C-suite from day one?
Not necessarily. In 8 out of 10 cases, the strongest pilots started without EXCO sign-off. Show results on one tight vertical, then use that to get resources. Leadership believes numbers, not concepts.
Where do you start for a GEO initiative?
I map my 10 main category queries and check if they trigger AI Overviews. Next, I find one person in each function (product, content, tech) with direct interest in improving performance on those queries. I pitch them a 6-week test with one clear revenue target.
