Does alt text actually matter for ChatGPT and Gemini, or only for classic Google Image Search?

It matters for both, and arguably matters more now, because alt text is one of the few reliably indexed, machine-readable signals a retrieval system has to match an image to a query without re-analyzing raw pixels at scale. Google's own documentation still treats alt text as core image SEO guidance, and that same text layer is what AI crawlers rely on when deciding whether an image is relevant to a user's question.

Is HowTo or FAQ schema still worth using if the rich results are gone from Google?

Yes. The visible rich result in Google's search results page for both HowTo and FAQPage schema has been removed, but both remain valid schema.org types, and the underlying markup still helps AI systems parse your page structure cleanly, including matching steps to their images. Keep the markup for extraction quality even though it will not produce a visual SERP feature anymore.

What is the difference between ImageObject and Product structured data?

ImageObject schema describes the image itself, including its subject, caption, creator, and license, and is best used when the image is the actual subject of the page. Product schema describes an item for sale, with the image as one property alongside price and availability, and is the right choice for the vast majority of ecommerce pages. Both can appear on the same page without conflict.

How does ChatGPT's shopping feature actually find and display product images?

ChatGPT's shopping experience runs on the Agentic Commerce Protocol, which lets merchants, including Shopify sellers, feed structured product catalog data directly into ChatGPT. When a user uploads a photo or describes a product, ChatGPT identifies it and matches it against connected retailer catalogs, then displays images, prices, and specs in a visual, side-by-side format.

Do I need a separate image sitemap if my images are already in my regular XML sitemap?

You need the dedicated xmlns:image namespace and tags either within your existing URL entries or in a standalone image sitemap; a regular sitemap without that namespace will have any image tags ignored entirely. It does not need to be a separate file, but the namespace and tag structure are non-negotiable for images to be properly indexed.

Can AI engines use my images without linking back to my site?

Yes, and this happens often. Being retrieved and being attributed are two different outcomes. Google Lens and Circle to Search tend to link back to source or listing pages because the interaction functions like a results page, while ChatGPT and Gemini can synthesize new visual layouts or describe an image without a direct citation, especially when no structured product feed connects your catalog to the platform.

How do I know if my images are actually showing up in AI answers?

Combine manual testing, running your own product and content photos through ChatGPT, Gemini, and Google Lens with realistic buyer prompts, with Google Search Console's Image Search performance report for classic indexing signals, and a recurring monitoring cadence rather than a one-time check, since platform behavior shifts frequently enough that quarterly spot checks miss real changes.

Should I prioritize alt text or structured data first if I only have time for one?

Start with alt text, captions, and filenames, since they are the fastest to fix, require no development work, and are read by every major crawler without exception. Structured data compounds that foundation and becomes especially important once you are selling products or publishing instructional content with step-by-step images.

How to Optimize for AI Image and Multimodal Search in ChatGPT and Gemini