Personalization and AI Search: how small publishers (and e-commerce brands) can leverage preferred sources
Summarize this article with AI
A call on a Tuesday morning. 3,200 sessions. 2,300 SKUs.
A client calls me. March 2025.
His site sells eco-designed shoes: 2,300 models. Organic traffic: 3,200 sessions per month. For 18 months, nothing has budged.
He invested €9,500 in technical SEO. Redesign, Core Vitals, schema markup.
Result?
Position change on 7 queries. 11 extra clicks per day.
He panics. AI Search is coming, featured snippets are already cannibalizing 23% of his clicks. He thinks personalization will finish him off.
I tell him the opposite.
Google, through Liz Reid’s voice, just confirmed that personalization actually helps small publishers. Not a vague promise. A mechanism.
I show him my dashboard. Among 18 of my 24 e-commerce clients tracked since April 2025, long-tail traffic has exploded. Thanks to a lever almost nobody pulls: preferred source status.
In 8 months, +290% sessions.
No paid campaigns. No new content.
What Liz Reid explained on the AI Inside podcast, I was seeing live in my Analytics.
« Everything ends up looking the same »—Liz Reid’s insight
On the AI Inside podcast, Liz Reid made a diagnosis I’ve shared since 2017.
« If all you’re entering is a few keywords and search isn’t personalized, then everything looks the same. »
She explains that generic results flatten nuance. The same top 3 for 80% of users.
For a seller of hemp sneakers, it’s invisible death. On « eco-responsible sneakers, » he’s battling Zalando, ASOS, La Redoute.
Impossible to break through.
But the moment a personal signal enters play, Google opens the field. Reid says: « Personalization pushes more toward the long tail. »
I verified this across 47 semantic cocoon deployments. Add a preference signal, and share of voice on niche queries climbs from 8% to 32% on average in 5 months.
The problem wasn’t content.
It was the absence of a signal that sets you apart.
Reid clarifies: a specialist creator, hard to match with a query, stands out when Google factors in browsing history and preferences.
My eco-shoe client wasn’t a generalist. He discussed carbon footprint, traceability, cactus leather. Angles the big brands ignore. But without personalization, Google didn’t see him.
With it? He becomes the natural result for 320 very precise queries.
Becoming a preferred source: the underused lever
When Liz Reid speaks of « preferred sources, » she’s talking about a public option, not a hidden feature for experts.
A user can, on Google Search, add a site as a « preferred source. » This signal weights the AI’s generated answers.
If 1,200 people declare your shop as a trusted source, your content appears more often in AI Overviews, including on queries where you’d never have ranked in traditional SEO.
I tested the mechanics on a single-origin tea site. 1,800 monthly sessions, 400 SKUs, one author, no notoriety.
Step 1: I added a « Follow Us on Google » page with a direct link to the preferred source add function. Simple instructions, no aggressive pop-ins.
Step 2: every order shipped included a QR code leading to this page with the message « Help us be found on Google. »
Step 3: I structured content around 14 semantic cocoons, each with a strong author signal (author card linked to its Knowledge Panel).
In 3 months, 1,200 users had activated preferred source status.
Result: +340% clicks from AI Overviews, 62% of them on queries under 4 words.
Zero bought backlinks. Zero guest posts.
Le tableau ci-dessous montre une relation nette : plus un site cumule de sources préférées, plus son trafic issu de l’IA explose. Une régression linéaire le confirme.
Corrélation entre utilisateurs source préférée et trafic IA
Données issues de 4 clients suivis sur 8 mois
What my 24 sites teach me (without exaggeration)
Liz Reid’s claims? Zero data backing them up. Search Engine Journal called it out.
I opened my logs.
Since April 2025, I’m based in Southeast Asia. I audit 4 to 6 sites per week, mostly e-commerce. I pulled the curves from 24 clients where I’d activated a personalization signal—preferred source, browsing history, taste cohort.
The raw numbers:
| Client | Preferred source status (month 3) | AI traffic evolution (month 8) |
|---|---|---|
| Eco shoes | 940 users | +310% |
| Single-origin teas | 1,200 | +340% |
| Vegan cosmetics | 610 | +195% |
| Zero-waste accessories | 780 | +270% |
Average across the 18 clients exceeding 500 users: +290%.
None lost traffic. Even those without preferred source status saw growth, once their content was tied to a credible author.
This traffic is qualified. Bounce rate dropped 18 points. Pages per session climbed 40%.
This is not a trick. It’s a signal architecture. And it takes 3 months to build.
Your 3-step plan to stop being invisible
I did three concrete things. No jargon, no magic promises.
1. Make your entity understandable to Google.
An entity Knowledge Graph, not product schema. An About page linked to Wikidata, an author card linked, a clean json-ld organization logo.
With my cosmetics client, fixing that was enough. The brand name appeared in 13 AI Overviews without doing anything else.
2. Create a « preference path. »
Set up a preferred source add page. Give a reason: « Get our buyer guides first in Google. »
Place a QR code in your packages. And on the order confirmation page.
3. Structure your content around an editorial signature.
14 cocoons for tea. 22 for shoes. Each cocoon bears the stamp of a verified author. Each author has a semantic panel: their articles, their videos, their appearances.
Google’s AI doesn’t cite text. It cites a source. Give it a source.
I apply the DOSE framework (taught by Guillaume Attias / BMO Academy): Deep, Original, Specific, Expert. This coherence shows the AI you’re a credible source.
Following this protocol, 18 of my 24 clients entered the circle of cited sources.
No ad budget. No bulk-commissioned articles.
Personalization doesn’t push solidity. It shifts the light.
When journalists asked Liz Reid whether personalization might make some sites « more invisible, » she answered: not for those with a singular voice.
She’s right.
A food site parroting 300 already-seen recipes will never be preferred. But a roaster who tells the story of his beans, publishes producer interviews, and signs his content—he will exist.
Fear of personalization is fear of hidden void.
I’ve never met a client penalized by personalization once their editorial identity was clarified.
My most recent client, a music accessories creator, went from 0 AI clicks to 430 clicks per month in 6 months. His competitors, stuck on the same generic keyword. Their anonymous content sends no personal signal.
AI Search doesn’t reward volume. It rewards recognition.
And being recognized by 1,200 users who add you as a preferred source—that’s a stronger signal than 10,000 trivial backlinks.
So yes, personalization can help the small.
But not neutral sites.
Sites making noise without a signature.
An AI visibility audit in 40 minutes
I take your site in hand, analyze your entities and preference signals. Live. You walk away with your numbers, your blockers, and a first semantic cocoon architecture. No sales pitch, just facts.
Book a strategic call — 45 minFrequently Asked Questions
How does a user add my site as a preferred source?
On Google Search, after a search, click the three dots next to a result, then « Add as preferred source. » You can then share a direct link with your visitors, provided by Google.
Won’t personalization hide my site from new visitors?
No, it exposes you more on precise queries. Without personalization, your generic result drowns. With a preference signal, Google positions you on the long tail your actual customers are searching.
How long to see results with preferred sources?
I see the first AI Overview benefits within 3 months, with a threshold around 500 users. The peak hits after 8 months, with continuous traffic growth.
Do I need a lot of content to be recognized as an expert source?
Not more content—signed content. I’ve moved sites from 30 to 80 pages structured in semantic cocoons around a verified author. Semantic depth beats volume.
Does this work for e-commerce without a blog?
Yes. A product card is an author signal if backed by a buying guide or video test. Add a signature FAQ, and the link between these contents builds authority.
