Skip to content
Framework Daddy

How to get AI to recommend your product

Fill every field. Show full reviews. Answer your customer’s questions. Here is what Joe Kiernan learned from 5,000 AI shopping simulations.

Episode published Updated

What you can do in 1 hour

  1. Fill out every field

    Give AI useful information wherever it looks. Complete every relevant field your platform gives you.

    Amazon
    Images, A+ content, titles, bullets, descriptions, product attributes, and backend keywords.
    Shopify
    Images, descriptions, product details, and JSON-LD product schema.

  2. On Shopify, show reviews and add targeted Q&A

    Show the full review text on each product page. Add questions and answers for each customer persona and use case.

  3. On every platform, answer the 4 big questions

    Who is your customer? What is their problem? What are their options? Which one fits them best?

JK

Joe Kiernan · Vice President, Portfolio Strategy and Operations, Digital Fuel Capital

Joe ran the research below at Digital Fuel Capital. Previously, he built listing optimization systems at Perch and Razor Group.

Watch the full episode on YouTube

4 questions your listing must answer

Help Alexa and ChatGPT choose the right product for the shopper

  1. Who is your customer?

    Show the person, situation, or use case your product is made for.

  2. What is their problem?

    Name the job they need done and show how your product does it.

  3. What are their options?

    Explain the choices: sizes, materials, features, and alternatives.

  4. Which one fits them best?

    Give them a reason to choose your product for their specific needs.

Use these questions to choose what goes into your images, copy, and A+ content. Make the answers obvious to a shopper scanning the page.

About 50 product pages, 5,000 AI shopping simulations

Joe’s team built about 50 fictional product detail pages (PDPs) for men’s Oxford shirts. They changed one feature at a time, like an A/B test with more versions.

The features they tested:

  • Real product images
  • A one-line value proposition
  • JSON-LD product schema
  • Detailed brand copy
  • Repetitive, keyword-stuffed copy
  • Q&A for specific customer personas, shown or hidden
  • Full-text customer reviews

In each simulation, the AI saw a group of products and chose one for a specific customer.

For example, Joe showed a shirt with images alongside 4 competing shirts. Then he replaced it with the version without images, keeping the same 4 competitors. He repeated the comparison across 100 trials to see which version won more recommendations.

Across the experiment, the team ran 5,000 recommendation trials using about 50 page versions.

Explore Joe’s PDP test pages to compare the different page versions.

What matters for ChatGPT and other AI models

Results from 5,000 AI shopping simulations by Joe Kiernan at Digital Fuel Capital

  1. Full-text customer reviews

    Biggest lift

    The written reviews mattered more than a star rating and review count. Put the full text on the product page.

  2. Q&A for specific customers

    Big lift

    Answering “Should someone like me buy this?” helped AI match the shirt to the customer.

  3. Product schema / JSON-LD

    Helpful

    Adding structured product data increased the likelihood of a recommendation.

  4. Images, value proposition, and brand copy

    Small lift in this test

    Each helped a little. Keep investing in images: they show shoppers what words cannot.

  5. Repetitive keyword stuffing

    Hurt recommendations

    A 4,000-character block repeating that the shirt was “the best” made AI less likely to choose it.

Make your Q&A specific to the customer

For an Oxford shirt, “What is the fabric?” gives one fact. “Will this work for my first office job?” connects the product to a person and a situation.

Use the questions your customers ask before buying. Explain who the product suits, how to use it, and which version to choose.

Joe also tested hidden Q&A. It performed the same as visible Q&A. On your Shopify page, use expandable answers so shoppers can read them too.

Keep investing in images

Images gave a small lift in Joe’s test. His advice and Monte’s: keep building for the shopper as AI’s image understanding improves.

Show size, materials, use cases, and the details that make your product the right choice. Use the 4 questions above to plan the images.

3 ways to influence AI

  • Influence future models

    Publish useful content about your brand across the web. Give future models something worth learning.

  • Show up in the consideration set

    Fill your product fields and use clear product data so AI can find you when it searches.

  • Earn the recommendation

    Use reviews, targeted answers, and images to show why your product fits this shopper.

Joe’s experiment focused on earning the recommendation: once AI has found your page, give it reasons to choose your product.

Check whether your changes work

  1. Choose your searches. Write down the questions and shopping requests you want to appear for.
  2. See who shows up. Run those requests in the AI shopping tools your customers use. Record the products recommended and the reasons given.
  3. Repeat each month. Run the same requests after improving your listing. Track whether your product appears more and whether the reasons change.

At Perch and Razor Group, Joe’s team used changing Amazon searches to refine listings every night.

How Joe updated Amazon listings every night · 1:56 clipJoe explains how his team used changing searches to improve listings.

Watch the full episode

Open the full episode on YouTube (35:42)

← Back to The Amazon Formula More Impressions guides →

Put this framework into action with Pixii

Pixii turns these frameworks into Amazon-ready images and listings. Generate, test, and ship faster.

Use Pixii Grade your listing for free