For years, the advice on product feeds focused almost entirely on text: write better titles, add more attributes, fill in every field in your data specification. The images mattered for click-through rate, but they weren't doing any of the heavy lifting when it came to how Google understood and classified your products.
That's changing. Google's Gemini AI models are genuinely multimodal — they process images and text together, not text alone. In the context of Google Shopping and Performance Max, this means Google is increasingly using your product photos as an input when deciding what your products are, what searches they're relevant to, and how to classify them within its systems.
This is significant. And depending on the quality of your current product catalogue, it's either a competitive advantage or a problem you need to fix.
What Multimodal AI Actually Does With Your Product Images
The practical implication is easier to understand with an example. Take a product listed with the title "Premium Gift Item" — deliberately vague, perhaps to fit a product that could be given for several occasions. Previously, Google's understanding of that product was almost entirely dependent on what the title and description said. A vague title produced vague relevance.
Now, if the product image shows a dark brown bifold leather wallet, Google's AI can see that. It can identify the object, infer the material from visual texture, understand the style, and cross-reference that with what it knows about how people search for items like this. The product may now appear for searches like "leather wallet gift" or "men's bifold wallet" even if those exact phrases aren't in the title.
On the surface, that sounds helpful. For products where titles have historically been underdescriptive, the image can partially compensate.
But — and this matters — the AI is making inferences, not reading ground truth. An image that's ambiguous gives the AI less to work with. An image that's misleading can actively harm relevance.
Poor Images Are Now More Costly
This is the part that should concern a lot of e-commerce businesses. Product photography standards vary widely. What looks acceptable to a human shopper scanning a page may not give Google's AI the clear signal it needs.
Consider a few common image types that create problems:
Lifestyle images as the primary photo. A leather wallet resting on a wooden desk next to a coffee cup and a notebook makes for an attractive marketing image. It also shows Google a desk, a cup, a notebook, and somewhere in the composition, a wallet. The AI sees the whole frame — not just the product you want it to classify.
Multi-product images. A photo showing three products together (perhaps a gift set) makes it harder for the AI to determine which item is the primary product being listed. Confidence scores drop when the subject is ambiguous.
Blurry or low-resolution images. The AI needs sufficient detail to identify materials, styles, and features. A photo that's sharp enough for a thumbnail but lacks detail at full size gives the system less to work with.
Dark or high-contrast images where product details are obscured. A black product against a dark background may look dramatic, but the AI struggles to extract meaningful information about texture, construction, or features.
The Combination That Now Matters Most
The shift to multimodal AI doesn't make product titles and descriptions less important — it makes them more important in combination with strong images.
Here's why: Google's AI compares what it reads with what it sees. When a precise, well-written product title says "Dark brown bifold leather wallet, 8-card slots, full grain leather" and the image shows exactly that, the two signals reinforce each other. Google's confidence in how to classify and match the product increases. Relevance improves. You appear for more of the right searches.
When there's a mismatch — a detailed title but a lifestyle image that doesn't clearly show the product, or a clean product photo but a vague title — the signals don't reinforce each other as strongly. You're leaving relevance on the table.
The businesses that will gain most from this shift are those that treat product titles and product images as two halves of the same quality standard, not separate concerns managed by different teams.
Practical Actions Worth Taking Now
The most useful thing you can do is run a review of your best-selling products — start with the top 50 — and check both the image and the title for each one.
For images: Is the main product image showing the product clearly against a clean background? Is the product the unambiguous focus of the frame? Is it sharp enough to show material and construction detail? If the answer to any of these is no, those images are candidates for reshoot or replacement.
For titles: Does the title include the key descriptors that a shopper would use to search for this product? Material, colour, size, key features, use case — these should appear in the title, not buried in an attribute field or left out entirely. Think about what a well-informed person would type into Google if they were specifically looking for this item.
The gap between your title and your image should be as small as possible. If your title says "navy blue men's chino trousers, slim fit, 32 waist" and your image shows navy slim-fit trousers clearly, you're in good shape. If your title says "men's trousers" and your image shows a model wearing them from the waist down in a busy lifestyle setting, you're not getting the full benefit of either.
The Direction of Travel
Google is investing heavily in its ability to understand products at a semantic and visual level. The long-term direction is a system that can classify and match products accurately even from incomplete or imperfect data.
But that doesn't mean the quality of your product data becomes less important. If anything, the inverse is true. The businesses that describe and photograph their products with precision are getting better signals into the system — better classification, better relevance, and better match rates against the searches their customers are actually using.
The AI rewards clarity. Give it something clear to work with.
For help auditing and improving your product feed quality, take a look at the product feed management service — covering title optimisation, image standards, and feed structure.