Are your product images ready for AI shopping?

AI isn’t coming. It’s already changing how people shop

The rollout of ChatGPT shopping has disrupted the ecommerce growth game. The floodgates are now open to a new customer acquisition channel – generative engines – and GEO (generative engine optimization) is the new SEO.

And it’s not just ChatGPT that’s targeting shopping and commerce. Google, Claude, Perplexity and other LLMs are expected to support product discovery, comparison and recommendations.

Recent announcements from Shopify and Klarna, paired with the skyrocketing growth of AI search platforms, make one thing clear: ecommerce is being refactored through the lens of generative AI. These interfaces are changing how consumers shop—and how product content needs to be formatted, surfaced, and understood.

Now’s the perfect time to get ahead of your competitors and get LLM-ready to maximize your chances of being featured in AI chats.

Why image optimization for LLMs is so critical

What once functioned solely as visual merchandising assets on a brand’s website are now becoming modular, machine-readable units in AI-generated answers. But most brands are not ready. While their images may look beautiful on a curated storefront, they often break, disappear, or get ignored entirely when rendered inside AI-generated carousels or product cards.

That’s because AI platforms have specific visual requirements, and they aren’t the same as traditional ecommerce norms.

ChatGPT, for example, displays products inside visually rich carousels. These layouts require images to be square (ideally 1024x1024 pixels) and cleanly centered.

Perplexity, which has quickly become a favored source for AI-assisted product research, presents product recommendations as “visual product cards,” often favoring 800x800 pixel square images with minimal compression artifacts and no branding overlays.

Klarna’s shopping plugin, integrated directly into ChatGPT, also pulls from structured product feeds and favors 1:1 high-resolution visuals.

What happens if you ignore these requirements?

If your product images don’t meet these format and quality standards, your listings may be down-ranked, omitted, or displayed poorly in these environments. And once these channels start driving serious buying intent—as early signs suggest they already are—this gap will translate into real revenue loss and brand erosion.

This is the essence of LLM-native merchandising: ensuring your content isn’t just beautiful, but also structured and formatted for AI-driven discovery.

Metadata: the invisible backbone of AI shopping

The requirements go deeper than size alone. Image metadata plays a crucial role in how generative platforms understand and recommend products. 

This includes descriptive ALT text written in natural language, semantically named image files (think “organic-cotton-tee-sage.jpg” instead of “IMG_3982.jpg”), and properly configured structured data within your product schema. These fields provide LLMs with the context they need to connect visual content to a user’s prompt.

Consider the difference between a clean, well-lit 1024x1024 image titled “wide-leg-linen-pants-beige.jpg” with alt text like “woman wearing wide-leg beige linen pants in an urban summer setting,” versus a low-res photo with a branded sticker in the corner and no surrounding metadata. To a human browsing your site, both may look fine. To an AI engine tasked with selecting three relevant products to recommend for the query “lightweight summer pants for travel,” only one of them qualifies.

Lifestyle context is becoming a ranking signal

Another common mistake is relying solely on traditional PDP hero shots – images with white backgrounds, floating product cutouts, or tight crops. While these are still necessary, they’re increasingly insufficient. AI models also look for lifestyle context: how a product appears when worn, used, or integrated into daily life. These images help LLMs answer implicit intent within user queries, such as “boots for muddy hikes” or “dresses that photograph well at weddings.”

The lesson here is to diversify your image set. Maintain high-quality product-only images, but supplement them with contextual visuals that show real-world use. Then, make sure those lifestyle images are embedded with descriptive alt text and included in your product’s structured data so they’re discoverable outside your site.

Speed, compression, and responsive design still matter

Speed is another invisible factor. AI crawlers increasingly favor sites that load quickly, especially on mobile. This means your image files should be optimized to maintain high visual quality without exceeding 300KB. Tools like WebP conversion, Cloudflare image optimization, or Shopify’s built-in image compression can help strike this balance.

But file size isn’t enough—you should also serve images responsively using modern HTML srcset attributes. This ensures the right image size is delivered to the right device or platform, and prepares your content for future multi-device LLM interaction environments, including voice-driven search on mobile or embedded shopping assistants on smart TVs.

Use schema to help LLMs understand your images

An overlooked but powerful enhancement is to include image metadata within your JSON-LD structured product data. This means defining the image’s url, name, and description fields inside the Product or MediaObject schema. Doing so gives AI assistants an explicit description of what’s in the image, increasing the odds it’s correctly parsed and selected for inclusion in a generated answer.

In essence, what we’re seeing is the emergence of a new kind of product photography standard—one defined not just by aesthetics or brand tone, but by compatibility with machine reasoning. And for once, it’s not the most expensive or flashy assets that win. It’s the most structured, versatile, and readable ones.

What leading brands are doing differently

This is why early adopters are moving quickly. Some are building media asset management systems that can automatically generate square variants of each hero image at 1024x1024 resolution. Others are integrating AI image tagging tools to enrich alt text and schema metadata across thousands of SKUs at scale. The goal isn’t just to look better, it’s to be interpretable by an algorithm deciding whether or not to feature you.

And yes, this effort is happening before these platforms have even rolled out formal monetization models. This is key. Just like with Facebook, Instagram, and TikTok in their early days, AI platforms are currently rewarding organic content because they haven't yet layered in paid prioritization. That window will close. The visibility you're getting today — on ChatGPT, on Perplexity, on Bing AI — is underpriced attention. But only if your content qualifies to be included.

Don’t be left behind

As AI continues to influence not just traffic but conversion pathways, the old assumptions around what makes “good” product content are eroding. Being image-rich is no longer enough. You must be image-readable.

So, is your ecommerce brand ready?

Your product images are no longer just part of your branding or conversion stack. They are now your gateway into AI-generated shopping journeys. And just like with SEO in the early 2010s or mobile optimization in the late 2010s, those who optimize early will capture disproportionate growth (until the rest of the market catches up).

Now is the time to audit your image formats, metadata, and schema. Review how your products appear when queried in Perplexity or ChatGPT. Are your visuals showing up? Do they look clean and clear? Or are you missing entirely?

To help you self-audit your images and content for LLM shopping, get our LLM Readiness Checklist

And to never fall behind your competitors, get ShopVision, your ecommerce AI SuperAgent and virtual teammate to automate research, reporting and marketing/merchandising tasks.