Google to E-Merchants: AI Visibility Flows Through Content People Actually Want to Read
Summarize this article with AI
A client calls me on a Tuesday morning. He lost 62% of his organic traffic in 6 months.
A client calls me on a Tuesday morning. He sells spare parts for electric bikes. 1,200 SKUs. 3 people on content production. He invested 18,000€ in technical SEO over a year.
Result? 62% of organic sessions lost between October 2024 and March 2025. Product pages vanished from AI Overviews. Vanished from snippets. Gone.
He was devastated. He thought Google had blacklisted him.
He tells me: « Stéphane, it’s the AI, right? It’s over, we can’t compete anymore. »
I tell him: « Your content, Stéphane. Not the AI. »
I open 3 product sheets. Same description as the manufacturer. Same specs as 11 other sites. No video, no structured customer reviews, no comparison. The content is an empty shell.
Google doesn’t penalize e-commerce. It saturates clones.
Three weeks later, we launch a new approach: useful content, made for humans. No magic tricks. Just substance.
Result at 11 months: +423% organic traffic. And 340% more clicks from AI Overviews.
It’s a mechanism, not a promise.
Google just reminded us of it.
Liz Reid, VP Search at Google, said it plainly: write what people want to read
On April 17, 2025, Search Engine Journal publishes an interview with Liz Reid (source). Google’s VP Search answers a question every e-merchant is asking: what are the new rules to appear in AI search?
Her answer fits in one sentence: « produce content people truly want to read. »
Not content for crawlers. Not SEO content. Not the thousandth copy-paste of the same press release. Real content. The kind you’d read on a Sunday morning with your coffee.
Reid cites a statistic: according to a Reuters study she references, users are increasingly turning to non-text formats — videos, visuals, social media. 54% of 18-24 year-olds prefer video over text to learn about a product.
E-commerce sites that only offer plain text are disappearing from the radar.
She doesn’t blame AI for killing publisher traffic. She says traffic shifts toward what has value. Videos. Comparisons. Testimonials.
« Multiple things are happening outside of AI. People want new formats. They’re going to YouTube, TikTok, social networks for their searches. »
Put plainly: Google isn’t stealing your clicks. You’re losing them because your content is replaceable.
It’s brutal. But it’s liberating.
Liz Reid le rappelle : Google récompense le contenu que les gens veulent vraiment lire. Pourtant, 98 % des fiches e-commerce n’apportent aucune valeur ajoutée. Ce donut illustre l’ampleur du problème.
98 % des fiches produits sans contenu unique
Un gouffre d’opportunité pour les AI Overviews
The trap of photocopied content: why 98% of product sheets are worthless
I track a consistent figure across 650+ audited sites: 98% of product sheets contain zero unique content elements. No field reviews, no in-house comparison tests, no original photos, no unboxing video. They repeat the manufacturer’s manual, sometimes reworded, often copy-pasted.
In 2025, that content is invisible. Google’s AI Overview will never pull a paragraph it’s already found on 14 other sites. It picks the most useful. The most complete. The only one that adds something the manual doesn’t.
Take a concrete case. A client sold high-end air purifiers. 47 models. 47 « technical » sheets: wattage, decibels, coverage area. Zero explanation about which filter type suits a baby’s bedroom. Zero comparison between two models that look similar but aren’t.
We reworked 19 priority sheets. For each product, a block titled « For whom? » written by the client themselves. A 90-second video shot on a smartphone in their warehouse: the real noise of the unit in night mode. A comparison table embedded on the page, not in PDF. All without touching a single title tag.
6 months. Organic traffic on these 19 pages jumped from 214 monthly visits to 1,137. Click-through rate up from 2.1% to 5.8%. No backlinks. No ad spend.
Unique content is a business asset.
I tested on 7 sites what users really want. And nobody’s doing it.
Since January 2024, I use one simple method on each new site: I spend 45 minutes interviewing 5 real customers of that site. Not panels. Not studies. I call them. I ask: « What was missing from this page to make your decision? »
The answers are always the same:
- « I wanted to see what it looks like sitting on a living room table. »
- « I couldn’t find the difference between the Pro and Lite models. »
- « Does it ship in 24 hours to my zip code? It was unclear. »
None of these questions have answers in a standard product sheet. Yet they’re what trigger the purchase. They’re what AI Overviews will synthesize by citing your page — if you answered them.
On 7 e-commerce sites I work with continuously, I added a layer of « decision-aid » content. For one, a sizing guide with real customer photos (+340% AI Overviews clicks in 4 months). For another, an interactive product compatibility module. For a third, « real user after 30 days » videos on 23 sheets.
Result: on average, enriched pages capture 3.2x more traffic than others. And crucially, they’re the only ones cited in AI answers.
Liz Reid says nothing different: « The more you produce content where you excel, where you bring something to the table, the more people will click. »
Construire un cocon sémantique ne commence pas par un mot-clé mais par une question humaine. La méthode DOSE (Decipher – Organize – Structure – Emancipate) structure cette approche, comme je l’ai apprise à la BMO Academy.
Le framework DOSE : 4 étapes pour un cocoon sémantique performant
De la question humaine à l’émancipation technique
The DOSE framework and semantic architecture: how I built an AI asset worth 423%
When I build a semantic cocoon for a site, I never start with the keyword. I start with the human question. And for that, I use the DOSE framework taught by Guillaume Attias at BMO Academy.
(Yes, I named the source. It’s rare, but here it’s foundational.)
DOSE: Decipher — Organize — Structure — Emancipate. In e-commerce SEO, that means:
- Decipher real buyer questions, not search volumes.
- Organize content into thematic silos answering each intent. Not catalog sections.
- Structure semantic entities so Google grasps topic depth.
- Emancipate the site from pure technique by anchoring value in content itself.
On the electric bike parts site I mentioned earlier, we applied DOSE across 4 silos: « repair, » « maintain, » « compare, » « choose. » Each product sheet became the hub of a mini-guide.
Example: a bike chain page. Before: 3 lines of specs. After: estimated lifespan, installation video by the client, compatibility with 7 models, link to essential tools, and a paragraph on worn chain noise — written by a partner mechanic.
Result in 11 months: +423% organic traffic on that silo. And 47 featured snippets. And the page became source #1 cited by AI Overview for any chain compatibility query in that segment.
It’s architecture readable by humans, therefore by AIs.
Stop feeding the algorithm. Feed the human. Here’s how.
I see too many anxious e-merchants. They chase the latest core update. They stack tags, schemas, thin content written at three cents a word. And the more they optimize, the less they’re read.
Here’s my rule. Just one.
Every time you create a page, ask yourself: « If I landed on this page by chance, would I recommend it to a friend? »
If the answer is no, don’t publish it.
Concretely, in e-commerce, that gives 5 immediate actions:
- Remove « inspirational » paragraphs and empty phrases. Replace with factual content.
- Add a block titled « Difference from the previous/similar model » on each product sheet.
- Embed an authentic 30-90 second video. Even shot on a smartphone. Even with non-professional voiceover.
- Structure key info for AI: price, stock, delivery time, compatibility in bullet points, not in dense paragraphs.
- Create cross-cutting pages nobody does: « Which [product] for [use case]? » with a table and personal insight.
At first, clients tell me: « But Stéphane, this content doesn’t scale! »
I tell them: « What scales is emptiness. Emptiness scales. Substance builds itself. And it pays. »
The air purifier site reworked only 19 pages out of 47. Yet the entire site’s traffic grew. The halo effect is real: Google understands the site has expertise, and it re-evaluates the whole catalog.
So stop feeding the algorithm. Feed the human. The algorithm doesn’t buy. The human does.
What Google won’t tell you (and what I see across 1,300+ cocoons)
Google tells you: « Make content people want to read. » OK. But it hides something else:
Content alone isn’t enough. You need semantic structure that makes it readable — for humans and machines alike. Since 2016, I’ve delivered over 1,300 semantic cocoons. And here’s what I observe: the highest-performing pages aren’t the best-written. They’re the best-connected.
A product sheet linked to a comparison, linked to a buying guide, linked to a customer testimonial video — this system creates a trust node. AI Overviews detects this mesh and cites the content at its center.
On a DIY equipment site, I built 8 cocoons in 14 months. Result: the site went from 4,000 organic sessions to 37,000. No ads. The pages most cited in AI Overviews? The guides titled « How to Choose Your Drill Press for a 15 m² Workshop. » Not product sheets. Guides.
Google won’t say it, but it’s simple: when your content is simultaneously informative, connected, and enriched with varied formats, you become the primary source. And in AI, the primary source is king.
The good news? You don’t need a 50,000€ budget. You need to grasp one simple truth: your AI visibility no longer depends on technical optimization. It depends on your ability to answer a question better than anyone else.
Are you the best answer to your customer’s question?
If the answer is no, you know what to do.
A live audit of your pages: I show you what’s missing.
I review 15 sites a week. In 45 minutes, we identify the 3 pages that must change first so your site finally enters AI Overviews. No ads. No redesign.
Book a strategic call — 45 minFrequently Asked Questions
What does Google mean by ‘content people want to read’?
Useful content: expert field reviews, original comparisons, real-world usage video, answers to specific questions standard product sheets ignore. Not generic SEO content.
Do Google’s AI Overviews completely ignore e-commerce product sheets?
No, they ignore copy-pasted sheets. The moment a sheet has unique elements — tests, reviews, compatibility guides — it becomes the cited source. I’ve tracked +340% clicks on enriched sheets.
Should I stop optimizing my site technically for SEO?
No, technique is the foundation. But it’s no longer enough. A technically perfect site with thin content won’t take off. Optimization starts with substance.
How do I create unique content when I have 5,000 SKUs?
Prioritize the 20% of pages driving 80% of revenue. On those, invest 45 minutes per sheet: real photos, a block titled « Key Difference from Model X, » a video. The halo effect will lift the others.
Is the DOSE framework an official method recognized by Google?
No, it’s a semantic architecture method by Guillaume Attias (BMO Academy). It structures a site to answer real user questions — exactly what Google expects.
