Beyond Rufus: How ChatGPT, Perplexity and Google AI Decide Which Products to Recommend

A growing number of shoppers no longer start on Amazon or Google. They start by asking an AI assistant: “What’s the best lifeguard rescue tube for a public pool?” or “Which peanut gift tin ships well?” The assistant answers with a few named products. Then the shopper goes to buy one, often on Amazon.

If your brand isn’t in that answer, you lose the sale before your listing ever gets a chance. This post covers what general AI assistants draw on and what established brands can do about it. For Amazon’s own assistant, see why Rufus doesn’t recommend your product.

Where AI assistants get their product answers

Assistants like ChatGPT, Perplexity, Claude and Google’s AI results generally build product answers from some mix of:

• What’s published about your product on the open web: reviews, buying guides, comparison articles, forums, industry publications.
• Your own website, especially pages that clearly state what the product is, who it’s for and how it compares.
• Retail and marketplace listings, including Amazon and Walmart pages, where they can access them.
• Live search results, for assistants that search the web before answering.

The pattern that matters: AI assistants recommend products that are described clearly and consistently in many places. One great listing isn’t enough. They’re looking for agreement across sources.

Why established brands get left out

This surprises a lot of owners. You have a real product, real customers and years of reputation. A newer brand with a fraction of your sales shows up in AI answers and you don’t. Usually it’s one of these:

1. Your reputation lives offline. Dealers, trade shows, catalogs and word of mouth built your business. None of that is text an AI can read.
2. Your website describes the company, not the products. A beautiful brand site with thin product pages gives an assistant little to quote.
3. Nobody has written about you. Comparison articles and buying guides are where AI assistants learn which products are “best for” something. If you’re not in those, you’re invisible.
4. Your information disagrees with itself. Different specs on your site, Amazon and a dealer’s page make you a risky recommendation.
5. Your site blocks AI crawlers. Some sites unintentionally block the bots these assistants use to read the web.

What to do about it

1. Make your own site answer product questions. Each key product should have a page that says plainly what it is, who it’s for, what it’s best at, key specs and how it compares with alternatives. Write it so a person could get a clear answer in thirty seconds. That’s also what an AI needs.

2. Make everything match. Specs, names, sizes and claims should be identical across your website, Amazon, Walmart and dealer pages. Start with your top sellers.

3. Let AI crawlers in. Check your site’s robots.txt with whoever runs your website. Unless you have a specific reason to block them, allow the major AI crawlers to read your public pages. Some sites also publish a plain-language summary file (llms.txt) to help assistants understand what they offer.

4. Get into the places AI reads. Trade publications, independent reviews, “best of” roundups in your category, and reputable industry directories. This is closer to PR than advertising, and it compounds.

5. Keep your Amazon reviews and rating strong. Assistants often cite marketplace reviews, and shoppers who click through land on your listing. A strong rating supports both.

6. Test it like a customer. Ask several assistants the questions your buyers ask. Note who gets named and what sources they cite. Repeat monthly.

How this connects to Amazon

Many shoppers who get a recommendation from an AI assistant still buy on Amazon. That means your Amazon listing has to close the sale the assistant started. If the assistant says “best for commercial pools” and your Amazon title doesn’t mention commercial use, the shopper wonders if they’re on the right page.

That’s why we treat AI visibility and Amazon as one job, not two. The work that helps Rufus, clean attributes, consistent information and use-case language, is the same work that helps every other assistant.

Where to start

If you do one thing this month: pick your three best-selling products, ask three AI assistants who makes the best version of each, and see if you’re named. If not, the list above is your plan.

If you’d like help, our free channel scan is a good first step on the Amazon side, or book a 20-minute call.

Frequently asked questions

Written by

David Ledbetter, Chief Growth Officer

Read bio →