7-Day Image Test to improve Visual Recommendations for ChatGPT, Perplexity, and Gemini

By Steve Merrill | April 5, 2026

Your product images are part of your AI ranking signal, and most Shopify store owners don't know it yet. ChatGPT Shopping, Perplexity, and Google Gemini all display product images when they make recommendations. The image that shows up is pulled directly from your product feed. Which means if your images aren't set up for AI contexts, you're leaving real recommendations on the table.

I've run audits on 40+ Shopify stores over the past six months. The image quality gap is one of the most consistent findings. Stores with clean, high-resolution, correctly formatted images show up more in AI product recommendations and convert better when they do. The data on this is pretty clear.

This is a 7-day test you can run right now to figure out which image format actually performs for your products. No expensive tools. No agency required. Just a structured plan, a tracking sheet, and your existing Shopify backend.

Why Do Product Images Affect AI Recommendations?

AI shopping tools don't just read your text data. They pull the full product record from your feed, including your main image URL, and they display it to users alongside the recommendation. Google's Merchant Center image requirements set the minimum bar: 1200x1200px, clear product shot, no promotional overlays. ChatGPT Shopping and Perplexity both pull from similar structured data pipelines that flow through Google's ecosystem.

Here's the thing: AI recommendation interfaces are visual. When ChatGPT or Perplexity shows a user three products, the image is often what triggers the click. A blurry 400px image next to two crisp lifestyle shots doesn't win. The AI didn't penalize your text data. Your image just made your product look worse in context.

What's the Right Hypothesis to Test?

Before you change a single image, write down exactly what you're testing. Vague tests produce vague results. The two most useful hypotheses for most Shopify stores right now:

Hypothesis A: White-background product images (product centered, no props, pure white or light grey background) get more AI recommendation appearances and higher conversion rates than lifestyle images.

Hypothesis B: Lifestyle images (product in real-world context, natural light, human interaction) drive higher click-through from AI recommendations than white-background images, even if appearance rate is similar.

Pick one. Don't test both simultaneously across the same products. That turns your data into noise.

For most apparel, electronics, and accessories brands, I'd start with Hypothesis A. White-background images match the format Google Merchant Center is built for, and they're the format AI interfaces pull into comparison layouts cleanest. For home goods and beauty, Hypothesis B is worth testing first since context sells the product.

How Do You Set Up the Test on Shopify?

Pick 10-20 products from a single category. Don't mix categories. You want to compare apples to apples, and category-level patterns are what you're looking for.

Split your selected products into two groups:

• Group A (Control): Products that stay on their current primary image
• Group B (Variant): Products that get the new image format

Prepare your variant images before you start the clock. Google's Content API for Shopping specifies image requirements that also apply to what ChatGPT and Perplexity consume: minimum 100x100px (recommended 800x800px or larger), JPEG or PNG, no promotional text overlaid on the image.

For white-background images: shoot at 1200x1200px minimum, product fills 75-85% of frame, background is pure white (hex #FFFFFF) or very close to it. For lifestyle images: 1200x1200px minimum, natural light, product clearly identifiable as the focal point, no text overlays.

Upload to Shopify, set as the primary image for Group B products, then verify your Google Merchant Center feed has picked up the new image URL. Don't assume it has. Cached versions are common. Check the actual feed file or use the Merchant Center diagnostics to confirm.

What Metrics Do You Track Over 7 Days?

The data doesn't lie, but only if you're capturing it consistently. Here's what to track:

For manual query tracking: run 5-10 specific product queries per day across ChatGPT, Perplexity, and Gemini. Queries should mirror how a real customer would ask, something like "best [product type] for [use case] under $[price]." Log the date, query, platform, whether your product appeared, its position, and which image showed.

Do this consistently. Same queries, same platforms, same time of day if possible. Variation in your methodology creates variation in your results that has nothing to do with your images.

What Does the 7-Day Schedule Actually Look Like?

Here's the plan, day by day:

Day 1: Baseline and setup. Run your query set and log results for all test products before any changes. This is your before-state. Upload Group B variant images to Shopify. Verify feed sync in Merchant Center.

Day 2: Confirm image propagation. Check Merchant Center diagnostics. If images haven't updated, troubleshoot now. Common causes: Shopify CDN cache, feed fetch schedule, image URL format mismatch. Fix it before the test clock starts counting.

Days 3-6: Daily query logging. Run your query set once per day. Log every result. Don't change anything during this window. Consistency is the whole game here.

Day 7: Analyze and decide. Tally appearances by group. Compare conversion rates in Shopify Analytics for test products during the window. Look for patterns, not just totals. If Group B appears more often but converts at the same rate, that's still a signal. If Group A appears less but converts better when it does, that tells you something different.

According to research on visual search and AI product discovery, image quality and relevance consistently rank among the top factors in AI-driven product matching systems. Your 7-day test is a way to validate that finding against your specific inventory and customer base.

What Should You Do With the Results?

Winning image format: roll it out store-wide within that category. Don't wait. If white-background outperformed lifestyle across your Group B products, the opportunity cost of waiting is real, compounding AI recommendation appearances you're not getting.

Inconclusive results: check your sample size first. If you had fewer than 50 sessions per product variant, the conversion rate data is too noisy to trust. Run the test for another 7 days before drawing conclusions.

Tie: dig into the breakdown by platform. ChatGPT Shopping and Perplexity may favor different image types. I've seen stores where white-background wins on Perplexity and lifestyle images win on Gemini. In that case, you need to make a judgment call: pick the format that serves the platform your customers actually use most.

One thing to track going forward. This test is a snapshot. AI shopping tools are updating their interfaces and ranking factors regularly. Running this test quarterly gives you a real signal on what's working as the platforms evolve. Set a calendar reminder now.

Sources: Google Merchant Center image requirements | Google Content API for Shopping image specs | Image quality factors in AI product matching (arXiv, 2023)