"What's the best cordless vacuum for pet hair under £300?" More shoppers now ask that question to ChatGPT, Gemini, Perplexity or Google's AI Mode instead of typing two words into a search box. The AI answers with a shortlist of specific products. If yours aren't on it, you don't get considered.
Here's how AI assistants choose products, and what you can do to be chosen.
Where AI assistants get product information
| Source | Used by |
|---|---|
| Google's Shopping Graph (built partly from Merchant Center feeds) | Google AI Mode, AI Overviews, Gemini |
| Shopping catalogues and partner channels (e.g. Shopify Catalog) | ChatGPT, Copilot, Perplexity and others |
| Your product pages and structured data | All of them, through crawling and live search |
| Reviews on your site and review platforms | All of them |
| Comparison articles, "best of" lists, forums and Reddit | All of them, heavily |
| Marketplaces (Amazon, eBay, Etsy) | Many assistants for availability and pricing |
The assistant combines these to decide which products fit the request and which are trustworthy.
The checklist
1. Make every product fully describable
AI matches detailed requests to detailed data. For each product, make sure your page and feed state:
- What it is, in plain words, in the first sentence.
- Key specs: dimensions, weight, materials, capacity, battery life, compatibility.
- Who it's for and what it's best at ("best for small flats with pets").
- Price, stock and delivery time.
- Returns and warranty.
If a shopper's question includes a requirement your page doesn't mention, the AI can't confirm you meet it.
2. Clean up your feeds and channels
- Keep your Google Merchant Center feed accurate and complete. See product feed optimisation.
- On Shopify, check Agentic Storefronts. See our guide.
- List on relevant marketplaces if your category is shopped there.
3. Add Product structured data
Mark up name, brand, GTIN, price, availability, shipping, returns and aggregate ratings. It helps Google and other systems read your pages reliably. Our free schema generator and JSON-LD checker help.
4. Earn reviews that say something
Assistants read review content, not just star ratings. "Picks up dog hair from rugs brilliantly, battery lasts about 45 minutes" gives the AI evidence for a recommendation. Ask customers what they used it for. Zelvorm Loop sends review requests after delivery.
5. Get into the comparison articles
When an assistant answers "best X for Y", it often summarises the top comparison articles. Find the articles that rank for your key product searches and:
- Offer review samples to genuine reviewers.
- Make sure your products have the information reviewers need (spec sheets, high-quality images).
- Pitch your product's specific strength for a specific use.
Read brand mentions are the new backlinks.
6. Create buying guides on your own site
Write honest buying guides and comparisons: "How to choose a cordless vacuum for pet hair", "Our X vs Y: which should you buy". Include your products and, where fair, alternatives. Assistants cite clear, specific guides. Our guide to query fan-out and topic clusters explains how AI search uses pages like these.
7. Be consistent everywhere
Product names, prices and specs should match across your site, feeds, marketplaces and retailers. Conflicting data lowers confidence.
8. Make the click worth it
When an assistant links to you, the landing page must confirm everything quickly: price, stock, delivery date, returns, reviews and an easy checkout. See cart abandonment fixes.
How to check where you stand
Once a month, ask several assistants the questions your customers ask:
- "Best [product type] for [use] under [price]"
- "[Your product] vs [competitor]"
- "Is [your brand] any good?"
Record whether you appear, how you're described and which sources are cited. Those sources are your to-do list. Our AI visibility checker gives a quick technical check of your site.
Track AI referrals in GA4 too, using our guide to tracking ChatGPT, Gemini and Perplexity traffic.
What doesn't work
- Keyword-stuffed descriptions. AI reads for meaning; stuffing reads as low quality.
- Fake reviews. They break platform rules and consumer law, and AI systems are trained to discount suspicious patterns.
- Hiding prices or delivery costs. Assistants favour products with clear, verifiable details.
- Thin, copied manufacturer descriptions shared with dozens of other shops. Add your own detail.
The short version
AI assistants recommend products they can describe confidently and trust. Give every product complete, specific information on the page and in your feeds, add structured data, collect detailed reviews, get into comparison articles, publish honest buying guides and keep data consistent. Then check monthly which sources the assistants cite and work on those.
For the wider shift, read agentic commerce explained. Our e-commerce team can audit your products for AI visibility.




