A picture says more than a search term

Someone on the subway sees a jacket with exactly the right cut, but can name neither the brand nor the model. In the past, that observation would usually have led nowhere. Today, a photo taken with a smartphone is enough, and an image search returns similar or identical products within seconds, often with a direct option to buy. It is exactly this shift, away from describing things in words and towards the direct visual query, that is currently changing how people search for products – faster than many e-commerce companies have prepared their own systems for it.

How big is the shift really?

The numbers behind this trend are now too large to dismiss as a fringe phenomenon. Google Lens now processes almost 20 billion visual search queries per month and is used by more than 1.5 billion people – four years ago, this figure was still around 3 billion queries a month.¹ Around one in five Lens searches is already directly related to shopping, and about 40 percent of product searches by younger users now start with an image instead of a word.¹ Germany alone accounts for an estimated 250 to 300 million of these monthly queries.²

This growth curve is remarkable because it does not stem from a single viral event, but from a steady development over several years that has accelerated with every new generation of smartphones and every improvement in the underlying image recognition models. Unlike many short-lived trends in e-commerce, visual search has thus evolved from an experimental add-on feature into a firm habit in search behaviour, which has now reached six times the volume it had four years ago.

Why people reach for the image instead of the word

The reason for this shift lies less in the technology than in a very human problem: certain visual characteristics can hardly be described precisely in words. Which shade of green is meant, which pattern, which fit, which furniture silhouette? A survey shows that 62 percent of the shoppers surveyed already prefer visual search to text-based search when it comes to finding a specific product – with younger user groups between 18 and 24 driving this trend particularly strongly.³ For them, the camera has long ceased to be an add-on feature; it is the most obvious starting point for a product search.

Platforms driving the change

This change is particularly evident at Pinterest, which put visual discovery at the centre of its platform from the very beginning. A study by Pinterest itself shows that 80 percent of active users begin their shopping journey with a visual search, compared with 58 percent of people who do not actively use the platform. In addition, when buying clothing or furniture online, 85 percent of respondents place more value on visual than on written product information.⁴ Google is following suit with Lens, as are Amazon, eBay and various fashion retailers, which have now integrated image search into their apps as a standard feature.¹ What was a niche feature for tech-savvy users just a few years ago has thus finally arrived in the mainstream.

What is interesting is that the platforms differ in their focus: while Google Lens works mainly as an extension of an existing, text-based search and is often used to quickly identify an object someone has already seen, Pinterest is built from the ground up around visual inspiration. For customers, this means two different starting points: some already know exactly what they have seen and want to identify it. Others are still in an early, inspiration-driven phase and let visually similar suggestions lead them to new products. A shop that is only prepared for the first group misses a considerable share of the demand from the second.

The underestimated side effect: fewer returns

What is often lost in the discussion about visual search is its effect far beyond mere findability. A global survey by Cloudinary of almost 2,700 consumers in Germany, the UK, the US and Australia shows that targeted use of image and video material such as 360-degree views, 3D models and user-generated videos can increase the conversion rate by up to 57 percent.⁵ Even more remarkable: companies that use visual content consistently can reduce their return rate by up to a third. Almost a third of respondents, 30 percent, said they had already sent goods back because the product did not look in reality the way it did on the website.⁵ Visual search and visual product presentation therefore have a double effect here: they not only help customers find the right product in the first place, but also help them develop more realistic expectations of what will arrive.

What this means technically

Technically, visual search works on a similar basic principle to modern semantic text search, only at the image level instead of the word level: an uploaded photo is converted into a vector, a mathematical representation of its visual features such as shape, colour, pattern and texture, and compared with the correspondingly prepared images in the product catalogue. Images that are visually similar lie close together in this vector space, even if the product texts do not share a single word. For retailers, this means: for their own products to appear in such a search at all, images must be available in high quality, with a neutral background, clear structures and recognisable attributes such as material, cut and brand logo, complemented by structured data and metadata according to common standards such as Schema.org.¹

An often underestimated point: these requirements differ significantly from what has been optimised for classic product photography in your own shop. An image that looks good on your own product page, with a decorative background or several products in the picture at once, can be difficult for automated visual recognition to evaluate, because the system cannot clearly determine which element in the image is actually the relevant product. Anyone who takes visual search seriously therefore usually needs at least one additional, cut-out reference image per product, independent of the rest of the image gallery on the product page.

How Eywora prepares companies for visual search

This very capability is part of Eywora's semantic AI search: in addition to text-based and hybrid search, the platform also supports visual AI search, where customers can upload a photo or image to find matching or similar products in the shop's own catalogue – without having to build a separate, additional system. Text and image search run on the same data basis and the same reranking model, so that results are sorted consistently by relevance, availability and individual context, regardless of the input format.

To make sure this channel works properly from the start, the Health Score module also shows for which products image quality or missing attributes could limit findability via visual search, before customers even notice. Anyone who wants to see how visual search can be applied to their own catalogue can test Eywora free of charge for 14 days without a credit card, or book a 30-minute live demo on a real customer shop.

A practical roadmap: how to prepare now

Anyone who wants to prepare for visual search should not start with the technology, but with their own image database. Product photos should be of consistently high quality, with neutral backgrounds and clearly recognisable details of material, colour, shape and brand. Structured product data according to Schema.org and an up-to-date Google Merchant feed ensure that external search systems such as Google Lens can assign your products correctly in the first place. And anyone who wants to measure the effect in the long term should set up tracking from the start that evaluates visual search traffic separately from classic text traffic – because visibility in an image search is not yet revenue if nobody tracks whether it actually leads to purchases.

Conclusion: if you only think in words, you miss half the demand

Visual search is no longer a future scenario, but already a market of tens of billions of monthly queries that is increasingly complementing, and in some cases replacing, classic text search, especially among younger target groups with strong purchasing power. Companies that continue to align their product data and image quality exclusively with text-based search engines are giving up a growing share of demand that simply can no longer be put into words. Those who invest in clean image data and visual search capability today are building an advantage that will become a basic requirement tomorrow.