1 / 38
E-Commerce Warsaw Expo · April 23, 2026
AI Search for
E-commerce
How to get your products cited, recommended
and bought through ChatGPT, Perplexity & Gemini
Nathan Fenina
Founder & CEO, Decupler · Top-20 SEO Expert France
Société Générale · Decathlon · Le Point · Pluxee
Step 1 — Prompts
Marketers vs
how buyers actually ask
✗ How marketers think
❌"waterproof running shoes B2B"
❌"premium athletic footwear collection"
❌"omnichannel sports retail solution"
✓ How buyers ask AI
✅"best trail shoes for wet terrain under €100"
✅"which shoes don't slip on rainy days"
✅"trail vs road shoes for beginners"
💡
Think like a customer with a problem — not a marketer with a product
Step 1 — Prompts
5 sources to
mine real prompts
Ask the AI directly — "What questions would someone ask about [category]?" → instant list
Google People Also Ask — real questions = real prompts. Expand every PAA on your top keywords.
Reddit & forums — community language = LLM language. Same words, same intent.
Your customer support logs — real questions from real buyers. The highest quality source.
Competitor reviews (Amazon, Trustpilot) — what buyers say about competitors = what they'll ask AI.
Reverse Prompting
Ask the AI what it needs
to cite you instead
The 3 prompts to use
→"Why do you recommend [Competitor]?"
AI tells you exactly which signals it's reading
AI tells you exactly which signals it's reading
→"What would [YourBrand] need to be recommended instead?"
Direct content gap list from the model itself
Direct content gap list from the model itself
→"What sources would you need to see [YourBrand] cited in?"
Your off-page hit list — straight from the AI
Your off-page hit list — straight from the AI
// Run this in ChatGPT today
Q: "Why do you recommend Nike for trail running?"
AI: "Nike is cited in Runner's World, has
detailed drop/stack specs documented, and
appears in community comparisons on Reddit..."
// Now you know what to build:
→ Detailed spec pages (drop, stack, weight)
→ Get into specialist running publications
→ Structured comparison content
💡
Run across ChatGPT, Perplexity & Gemini — prioritise gaps that appear in all 3
On-Page — Content Types
8 formats
LLMs love to cite
🏆
Top / Best
"Best hiking boots wide feet"
Numbered lists with criteria. Highest citation rate for commercial queries.
⚔️
VS / Comparison
"Trail shoes vs road shoes"
Structured tables with clear winner per use case. AI extracts these directly.
🧮
Calculator / Tool
"Shoe size converter EU/US"
Utility content = high trust signal. Gets cited as "useful resource".
📖
How-To / Guide
"How to choose trail shoes"
Most cited by AI. Hugely underused by e-commerce brands.
❓
FAQ
"Do I need waterproof shoes?"
Direct Q&A is what LLMs quote verbatim. One question = one answer <150 chars.
📊
Data / Stats
"47 waterproof models, 38–48"
Specific numbers = credibility. AI treats quantified claims as citable facts.
📍
Use-Case Specific
"Shoes for nurses on hard floors"
Niche specificity = low competition + high intent. Almost zero brands do this.
⭐
Review / Verdict
"Our 6-month test of trail shoes"
First-person testing + specific data. Strong E-E-A-T signal for AI.
On-Page — Key Technique
The Capsule Summary —
what LLMs quote verbatim
✗ Generic — ignored by AI
"Discover our running shoes collection. Find the perfect pair for your next adventure with our quality products."
❌No specific facts
❌No answerable data
❌Nothing citable
✓ Capsule — AI quotes this directly
"Trail running shoes. 47 models, heel drop 0–12mm, sizes EU 38–48. Waterproof options from €89. Updated April 2026."
✅Under 150 characters
✅Specific, factual, current
✅Directly answerable
On-Page — Actions
What to add to every
product & category page
Capsule summary — 1 bold paragraph, <150 chars, answers the exact query
Comparison table — headers with specs; RAG systems extract tables directly
Numbered list with specs — "1. Gore-Tex membrane (10,000mm)" not "1. Waterproof material"
FAQ block (3–5 Q&As) — one direct answer per question, match PAA phrasing
Freshness signal — update date + current price range + stock count
Demo — Product Page AI
Input: product name →
Output: full AI-ready page
Product name
+ category
+ category
→
Claude Code
+ DataForSEO
+ DataForSEO
→
Title · Description
Capsule · FAQ · Schema
Capsule · FAQ · Schema
→
CMS push
auto
auto
Generated title
Salomon Speedcross 6 — Best Trail Shoes for Beginners 2026
Capsule summary
Trail shoe, 6mm lugs, 6mm heel drop, 280g. Ideal for muddy terrain. EU 38–48.
FAQ generated
Are Speedcross 6 good for beginners?
Yes — 6mm drop eases road-to-trail transition. Aggressive lugs provide grip without technical skill.
Yes — 6mm drop eases road-to-trail transition. Aggressive lugs provide grip without technical skill.
Schema generated
Product FAQ BreadcrumbList AggregateRating
⚡
~40 seconds per product vs 45–60 minutes manually
Automation
Updating 10,000 products
while you sleep
Product DB
CSV / API
CSV / API
→
DataForSEO
SERP + intent
SERP + intent
→
Claude Code
Generate content
Generate content
→
Auto validation
→
CMS Push
Shopify / WC
Shopify / WC
→
Live + indexed
Schema embedded
Schema embedded
Task
Manual
Automated
1 product page
45–60 min
~40 sec
100 products
~75 hours
~1 hour
Schema deploy
Dev sprint
Overnight
Technical — Schema
Schema types that
move the needle
Schema type
Priority
What it signals to AI
Where
Product
🔴 Critical
Price, availability, rating — makes your product extractable
Every PDP
FAQPage
🔴 Critical
Q&A pairs that LLMs cite verbatim in answers
PDP + Category
ItemList
🟡 High
Signals "best of" authority — perfect for category pages
Category
AggregateRating
🟡 High
Social proof — AI weighs reviewed products higher
PDP
Organization
🟣 Medium
Entity disambiguation — links brand to knowledge graph
Homepage
⚠️
Most e-commerce sites have none of this beyond basic Product schema. Every row above is an untapped advantage.
Technical — Critical
Many sites accidentally
block AI crawlers
⚠ AI crawlers are separate bots
🤖GPTBot — OpenAI / ChatGPT
🤖PerplexityBot — Perplexity AI
🤖ClaudeBot — Anthropic
🤖Google-Extended — Google AI
Check your robots.txt today — many sites block them accidentally
✓ Correct robots.txt
# Allow all AI crawlers
User-agent: GPTBot
Disallow: /private/
User-agent: PerplexityBot
Disallow: /private/
User-agent: ClaudeBot
Disallow: /private/
# Everything else: allow
Technical — Critical
JS-rendered content
is invisible to AI
AI crawlers don't execute JavaScript. If your product descriptions load via React or JS after page load — AI cannot read them.
✗ What AI sees (JS-rendered)
<div id="product-description">
<!-- Loaded by JavaScript -->
<!-- AI sees nothing here -->
</div>
✓ What AI needs (server-side HTML)
<div id="product-description">
"Trail shoe, 6mm lugs, 280g.
Sizes EU 38-48. From €89."
</div>
<!-- In HTML = readable by AI ✓ -->
⚡
Critical content — descriptions, specs, prices — must be server-side rendered
Off-Page
Sources AI models
trust the most
Specialist editorial media — running magazines, gear review sites, niche blogs. One mention here = more weight than 50 generic backlinks.
Reddit & community forums — AI models trained heavily on Reddit. A top comment recommending your product is gold.
YouTube reviews — Perplexity and Gemini cite YouTube. A credible reviewer mentioning your brand = AI citation.
"Best of" comparison articles — being in "Top 10 trail shoes 2026" on authority sites is the highest-value AI citation signal.
Off-Page — Agent Workflow
Find exactly who
to contact at scale
Run prompts
in Perplexity
in Perplexity
→
Note every
source cited
source cited
→
DataForSEO
Find ranking articles
Find ranking articles
→
Claude Code
Extract contacts
Extract contacts
→
Personalised
outreach
outreach
✗ Generic — gets ignored
"Hi, I'd like to be featured in your article. Our brand has great products I think your readers would love."
✓ Specific, value-first — gets replies
"Saw your article ranks #2 for 'trail shoes beginners'. One gap: no option for wide feet. Our model covers EU 38–48 in 2E width, 4.7★ from 1,200 runners. Happy to send a pair for testing."
Agent — Monitoring
Know who AI cites
in your category — weekly
Prompt
You
Competitor A
Competitor B
Action
"Best trail shoes beginners"
Cited #2
—
Cited #1
Defend
"Waterproof trail shoes women"
—
Cited #1
Mentioned
Create content
"Trail shoes vs road shoes"
—
—
—
Quick win
"Running shoes under €100"
Cited #3
Cited #1
Cited #2
Improve rank
📊
DataForSEO also parses Google AI Overviews — track who's cited in both Google and LLM search simultaneously
The Full Stack
Your AI Search
automation stack
🤖
Claude Code + Claude / Gemini API
Generate AI-ready content, schema, FAQ, outreach emails
📊
DataForSEO
SERP data, AI Overview parsing, PAA extraction, competitor intelligence
🔍
SEO AI Systems
Monitor citations across ChatGPT, Perplexity, Gemini — weekly alerts
🌐
Your CMS
Shopify, WooCommerce or custom — receives content via API automatically
90-Day Roadmap
Month 1:
Foundations
Run AI visibility audit — ChatGPT + Perplexity + Gemini
Build your 50-prompt target list
Deploy Product + FAQ schema on your top 100 PDPs
Check robots.txt — allow GPTBot, PerplexityBot, ClaudeBot
Add capsule summaries to top 20 category pages
90-Day Roadmap
Month 2:
Content & Monitoring
Publish 4 AI-first buying guides (Top, VS, How-To, FAQ)
Automate category page freshness updates (price, stock, date)
Launch competitive monitoring agent — 50 prompts tracked weekly
Run reverse prompting on top 5 competitors
Build entity signals — Knowledge Panel, Organization schema
90-Day Roadmap
Month 3:
Scale
Bulk update all PDPs with AI-optimized copy via pipeline
Launch citation outreach campaign — 20+ editorial targets
Build DataForSEO → content brief automation
Review monitoring data — double down on winning content types
First consistent citations appearing in LLM answers
Questions
Q&A
Nathan Fenina
Founder & CEO, Decupler
Let's connect
Ready to appear in
AI answers?
🎯
Free AI Search Audit
30-min call · AI visibility map · Competitor analysis
+ full 90-day action plan
+ full 90-day action plan
📧 nathan@decupler.com
🔗 linkedin.com/in/nathanfenina
🌐 decupler.com
📱 +33 6 52 41 70 77
Nathan Fenina — Founder & CEO, Decupler
Top-20 SEO Expert France · Société Générale · Decathlon · Pluxee