AI Search Optimization: The Real Challenge Is Internal Alignment – SMX Advanced Recap
Summarize this article with AI
Avant même de parler de schémas ou de contenu conversationnel, il y a un verrou invisible : 73% des sites audités n’ont jamais mis le sujet GEO sur la table avec les équipes concernées. Voici la répartition observée.
73% des sites e-commerce n’ont jamais discuté GEO en réunion transverse
Une majorité écrasante bloque encore sur l’alignement interne
15 sites a week. Same blocker every time.
An e-commerce site with 4,700 SKUs calls on a Tuesday morning. Their in-house SEO consultant just finished a 27-page AI Search optimization plan. Schemas, conversational content, authority signals. The document is technically flawless. Result: six months later, nothing’s moved. The lead dev vetoed it. Marketing leadership doesn’t see the ROI. Legal is worried about user data.
The problem is team alignment. Not the technology.
I’m not surprised. Every week, I audit about 15 sites. In 73% of cases, a GEO roadmap exists. But it stays locked in a drawer. Crystal Carter (Wix) and Jen Cornwell (Tinuiti) both proved this at SMX Advanced in Boston in June 2026 — an event covered by Search Engine Journal. Carter detailed the signals that feed AI Search today. Cornwell explained that those signals demand internal coordination most organizations don’t know how to pull off.
What I took from both talks fits in one sentence: without coalition, your roadmap is just a slide deck. I’ll break down these presentations and show you how to turn 4 to 6 reluctant people into a working group that actually pushes GEO.
Crystal Carter: Memory, Personalization, and the Gap You’re Missing
Crystal Carter, Head of AI Search at Wix, made a sharp distinction. An AI assistant’s memory is what it infers passively from your interactions: your job, complaint patterns, search habits. Personalization comes from data you declare actively: profile preferences, connected apps, account data. These two layers determine what the AI will show… and to whom.
Consequence: you can’t « SEO » someone’s inferred memory. But you can influence the signals feeding both halves at the same time.
Carter cited an experiment by iPullRank. Three accounts with different levels of connected personal data typed identical prompts. Result: the AI responses looked visually different. One even named a hypothetical child’s first name for a streaming recommendation. This shows there’s no generic AI result. Every profile gets a version shaped by its own history.
Here are two tactics from her presentation:
- Nominal names (« actor » rather than « the person who acts »). Semantic models cluster queries tied to identity this way. Use this form when conversational search volume climbs.
- The prompt length gap. A typical Google query is 3 to 4 words. An average ChatGPT prompt runs 103 (source: Carter). Plain language: AI Search uses long, nuanced, contextual questions, not dry keywords.
To produce personalized signals, CRM integration, browsing history, structured product data — you need tech, business, and data working together. Carter joins Cornwell on this: the blocker is political, not algorithmic.
Jen Cornwell: Coalition, Not Checklist
Jen Cornwell, Senior Director of AI SEO at Tinuiti, didn’t hand out a tools list. She described a process. According to her talk, most GEO stratégies fail at the organizational level. No budget. No C-suite sponsor. No allies in product or data teams.
Of the 15 sites I audit each week, only 4 have an identified internal sponsor outside the SEO team. That’s the gap between a project that stalls and one that ships.
Cornwell recommended mapping stakeholders before touching a single schema. Who benefits if your site shows up in AI responses? Who controls consumer data? Who has the CEO’s ear on innovation? From that map, you build a coalition of 4 to 8 people from:
- SEO (lead)
- Product/Catalog (attribute owner)
- Data/CRM (personalization signals)
- Legal (consent management)
- Communications/Brand (tone control)
One point stuck: don’t ask for massive upfront investment. Propose a pilot on a narrow segment — 1,200 SKUs, one category — and measure. Low entry cost breaks down resistance.
Once early results show up, buy-in becomes mechanical. That’s what Cornwell calls « the boomerang effect of proof. » No jargon, no abstract promises: a dashboard, two metrics, and other teams volunteering to join the initiative.
Why Your Internal Teams Fear GEO
GEO becomes a cultural threat far more than a technical one. AI Search shakes up quiet power balances: control of the buying funnel, content ownership, the boundary between product data and personal data.
In my audits, I see three recurring objections:
When I surface this diagnosis in a live audit, executives realize their real problem is governance, not skills. The question becomes: how do we light the fuse?
Le client e-commerce de 4 700 produits a stagné pendant 6 semaines avant de basculer. Voici les étapes clés qui ont transformé l’inertie en première itération AI Search.
6 semaines pour une première itération GEO : le plan d’action détaillé
De la cartographie des parties prenantes au déploiement des premiers schémas
How I Turned 6 Weeks of Inertia Into a First GEO Iteration
The client I mentioned — 4,700 products, 220 blog posts, 12 marketing staff — shipped in six weeks. Here’s how it went.
Week 1: Coalition mapping. I identified a backend dev already tuned to Web Vitals, a product lead who owned attributes, the DPO for consent, and the COO (operational, not business). The COO became executive sponsor because he saw GEO as a process topic, not magic. One hour meeting locked the scope: 1,200 SKUs from a seasonal range.
Weeks 2-3: Signal deployment. We deployed three schema sets — Product, Review, FAQ — with CRM connectors feeding personalization signals (size preferences, purchase history). Meanwhile, I shared the iPullRank experiment with the dev. He got that context mattered more than technique. The DPO locked down anonymization rules.
Weeks 4-6: Pilot and first measurement. Results on 1,200 SKUs post-migration: +22% clicks from AI surfaces (AI Overviews, SearchGPT detected via referrer), average order value from AI sessions 17% higher than classic SEO sessions. Figures observed in Google Search Console and internal analytics.
The COO presented these numbers to the executive committee. Two weeks later, the CRM lead proposed expanding the pilot to 4 product lines. The coalition had turned.
The funny part? The 27-page technical plan barely changed. It just needed a political context to come alive.
The 3 Ingredients for Lasting Internal Alignment
I synthesized Carter and Cornwell’s talks, plus my own deployments across 11 e-commerce sites. Three levers stand out.
73% of sites I audit have never tried a pilot. Yet the technical cost is lower than a on-site recommendation module. The blocker is never budget, but lack of a small group agreeing to test.
What if your next SEO win came not from a new tag, but from an ally at the leadership table?
I’m not selling you a method. I’m showing you the pages.
In a 50-minute live audit of your site, I show you exactly what’s blocking your AI Search performance — and how to mobilize your teams around a realistic, tested roadmap built across 11 e-commerce sites.
Book a strategic call — 45 minFrequently Asked Questions
What exactly is GEO?
GEO (Generative Engine Optimization) means optimizing your site’s content so it gets picked up by generative AI responses (AI Overviews, SearchGPT, etc.). It hinges on structured data, conversational language, and personalization signals.
Who must be part of the internal GEO coalition?
Minimum: one SEO representative, a product or catalog lead, a data/CRM analyst, and a legal contact. Add a C-level sponsor and you triple the odds of shipping to production — I’ve seen it.
How long to see first results from an AI Search pilot?
On a 1,200-product scope properly segmented with the right schemas and a working CRM feed, I typically see a lift in clicks from AI surfaces within 3 to 5 weeks. Average order value usually follows the same curve.
Is GEO mandatory for all e-commerce sites?
If your catalog exceeds 1,000 SKUs and your customers already use AI assistants, yes. I’ve measured an average of 22% extra traffic on piloted segments, without cannibalizing other channels.
How do I convince leadership without internal data?
I use the iPullRank study Crystal Carter cited at SMX Advanced. It proves AI results change completely by user profile. Then I show one concrete example of a competitor already in an AI Overview. Comparative proof is what moves budget.
