New AI Search data: what they reveal about visibility and trust
Summarize this article with AI
A call that changes everything
A client calls me on a Tuesday morning. His dietary supplement site is losing organic traffic. Six months ago, it was 4,200 sessions per month. Today, 2,400. A 42% drop. The worst part? Pages still rank well on Google. But clicks have stopped following.
I ask him to type his three best queries into ChatGPT. Zero brand mentions. No citations. Nothing.
His problem: visibility in generative AI responses, not classic SEO. Few sites understand this.
A June 2025 Search Engine Land study shows that consumers validate information across multiple platforms before buying. AIs seek identifiable authority sources. Without citations, you disappear from the purchase journey.
When I look at numbers across my clients, I see 70% of e-commerce pages are never picked up by a generative AI tool. Yet they rank. The gap between Google visibility and AI visibility is widening.
We changed how we viewed SERPs. We analyzed where these AIs pull information. Then we restructured.
What the Search Engine Land study reveals about AI trust
The key figure from the Search Engine Land study isn’t a percentage. It’s a mechanism. Generative AIs don’t « read » your pages like Google does. They compare, cross-reference, and look at consistency between your site and other external sources.
A user types: « Best magnesium for sleep ». They don’t just take the first link. They use ChatGPT, Claude, or Google’s overview, and verify on Reddit or YouTube. If your brand appears on these channels, the AI cites you. If not, it cites your competitors.
74% of generative responses studied contained at least one brand mention. But these brands aren’t necessarily the best-ranked in traditional SEO. They’re the ones whose authority is confirmed by multiple independent sources.
The study identifiés three factors:
- 1. Co-citation. The more your brand is mentioned on trusted sites (forums, specialty media, comparison platforms), the more reliable the AI considers you.
- 2. Information freshness. A product sheet updated 14 months ago carries less weight than a page with recent reviews.
- 3. Usable structured data. Well-implemented schema.org doesn’t guarantee a citation, but without markup, the AI extracts your data poorly.
We were absent from all three on the supplement site. No external citations. No regular updates. Nearly empty schema. The AI wasn’t citing us.
Key takeaway: Cross-validation matters more than content alone.
Three trust signals I observe in the field
I analyze 15 e-commerce sites per week. I always see the same blind spots. Here are three signals that shift the needle for AI visibility.
Signal #1: recurring citations off-site. A single isolated Trustpilot review isn’t enough. You need a network: mentions on specialty blogs, answers on Quora or Reddit, local press articles. The more sources vary, the more the AI integrates your name into its trust graph. I verified this for a client: in three weeks, 12 citations on health forums and 4 external blog articles. Result: the brand appeared in 8% of ChatGPT responses about magnesium. Before, it was 0%.
Signal #2: semantic architecture in silos. A site with 800 product sheets with no logical links confuses the AI. I applied the DOSE framework (Guillaume Attias, BMO Academy): detect central entities, organize silos, structure internal links, evaluate gaps. I created silos around minerals, vitamins, sleep disorders. Each silo links the product sheet to a health guide and a marked-up FAQ. The AI now has a clear path.
Signal #3: enriched structured data. Product, Review, FAQ, HowTo. I marked up each product page with relevant attributes, especially verified reviews. The AI can now extract the rating, review count, and price. It often cites them.
These three pillars don’t cost more than standard content. They just demand consistency.
Découvrez le processus en trois étapes qui a permis à un site e-commerce de multiplier ses citations dans les réponses IA.
Les 3 étapes clés pour booster les citations IA
Un processus en trois phases testé et validé
How we restructured the site in 3 steps
The process is nothing mystical. Three steps are enough. No need to reinvent the wheel—precise execution does the job.
Step 1: citability audit. I listed every strategic query. Then verified, via ChatGPT and Claude APIs, if the brand appeared. Result: absent 93% of the time. I also analyzed the sources the AI cited for these queries. They came from sites with domain authority scores above 45, pages with at least 15 recent reviews, and active forums. Our site had none.
Step 2: mentions campaign. I contacted 22 health blogs and secured 7 guest posts, plus 5 substantive forum responses. Total cost: €2,400, of which €1,800 was writing. The rest was targeted press outreach. Then I synchronized site content with these external mentions: each silo contained links to the articles we secured, and vice versa.
Step 3: technical reinforcement. Updated 47 product sheets with recent customer reviews (at least 8 per sheet). Review and FAQ markup. Loading speed under 1.2 seconds. And crucially, created a detailed JSON-LD file for each database entity.
Three months later, the brand was cited for 32% of tested queries. Organic traffic bounced back 27%.
« Before, we were invisible in ChatGPT. Now they cite us for our customer reviews. » – Marc, founder of Compléments Nature
Voici les chiffres clés qui montrent l’efficacité de la stratégie mise en place sur le site e-commerce.
Avant vs Après : l’impact de la restructuration
Citations IA et trafic organique multipliés en 3 mois
The result: +32% AI mentions, and organic traffic climbs back
In six weeks, the site went from 0 to 14 citations in ChatGPT for main queries: magnesium, melatonin, omega-3. Claude started mentioning it three weeks later. Google SGE integrated it into its product overviews.
Organic traffic climbed from 2,400 to 3,050 sessions per month in three months. That’s +27%. The conversion rate on pages reached via AI hit 4.3%, versus 2.1% for classic SEO traffic. Visitors from AI citations are more engaged. They arrive pre-validated.
The silo structure tripled the number of pages indexed with enriched data usable by AIs. We went from 47 pages without schema to 189 pages with complete markup. The mechanical effect is clear.
All without spending a dime on extra advertising. Just a rebuild of trust architecture.
The trap: too much content, not enough proof
Sites multiply articles and product sheets. They think they’re covering the topic. But AIs don’t care. Without external sources, you don’t exist.
I saw a site with 400 « SEO-optimized » articles. Zero quality backlinks, zero reviews. Clean content, but isolated. The AI ignored it. The solution? Don’t write more—build proof.
Focus your effort on a few pillar pages, enriched with reviews, testimonials, cross-citations. One reference page with 25 recent reviews and 8 external mentions? More useful than 40 pages with no validation.
Don’t chase volume. Chase reputation.
If you sell products, every sheet must become a trust sheet, not just a technical description.
How many of your pages are being picked up by AIs today?
When I audit a site, I always ask one simple question. Open ChatGPT or Claude. Type your main keyword. Count mentions of your brand. If it’s zero, you’re losing customers every day who will never find you.
The Search Engine Land study confirms it: AI trust is built on external signals. Not on your word alone. So what are you waiting for to check where you stand?
Your AI visibility audit in 30 minutes
I’m not selling you the method. I’m showing you the pages. In 30 minutes, I pinpoint the 3 blockers preventing your site from being cited by generative AIs.
Book a strategic call — 45 minFrequently Asked Questions
What are the most important trust signals for generative AIs?
I look at external co-citations, recent reviews, structured data (Review, FAQ), and consistency between your site and reliable sources like forums and specialty media.
How do I know if my e-commerce site is cited by ChatGPT or Claude?
Test your best queries in the tool. See if your brand appears. There are also audits like the « AI Visibility Checker ».
Can I force AI citations with backlinks?
No. AI doesn’t just follow links. It analyzes overall reputation and review freshness. Better to diversify natural mentions.
How long until I see improvement?
With targeted restructuring, first citations arrive in 4 to 6 weeks. Full effect arrives after 3 months.
Can product sheets with few reviews still be cited?
Difficult. AIs mainly look at sources with at least 8 to 10 recent reviews. Below that, they consider it not reliable enough.
