Your Customers Search With Their Camera Now. Can Google Even See Your Photos?

Google now tells a business how often people reach its website by searching with a picture instead of words. On 24 September 2026 Google added a multimodal search type to the performance reports in Search Console, covering searches made with Google Lens, Circle to Search on Android, image uploads to Google Search and the "Search this image" option in Chrome's right-click menu ¹. Search Console's documentation now splits web traffic in two: text-based, from queries typed into the search bar, and multimodal, from people who search using images, such as with a smartphone camera ². The same filter appears in the report Search Console keeps for Google's artificial intelligence (AI) answers, which counts how often links to a site are shown in AI Overviews and AI Mode ¹³. For the first time, a business can see whether customers who photograph a product, a dish, a room or a spare part ever land on its website, or whether the match goes to someone else.

The volume behind that filter is not marginal. Google said in October 2024 that Lens was used for nearly 20 billion visual searches every month ⁴, and put the figure at more than 20 billion in February 2025 ⁵. In a shopping announcement the same October, Google stated that 20 percent of all Lens searches are shopping-related ⁶. Circle to Search, which lets an Android user circle anything on screen and search it, was on more than 300 million Android devices by July 2025 ⁷. These are the latest totals in the Google announcements cited here, so they describe scale at those dates rather than today. Either way, a search that starts with a photo often starts with the object already in view, and the open question is who supplies it.

How Lens decides where to send that person is the part that lands on a website's build. Google has described Lens as having identified billions of objects for many years, such as plants, products and landmarks, "by matching against a database of images indexed from the web" ⁵. Newer AI models now let Lens go further and answer questions about more novel or unusual images ⁵, but the visual match, the result that says this product is on that page, depends on that same database of indexed images. A photo Google never indexed cannot be that match. That turns an old technical requirement into a commercial one: a website's images have to be found, indexed and understood before an image match can point back to them.

This is where many attractive websites fail without anyone noticing. Google's image documentation is plain: it finds images in the src attribute of a standard HTML image element, and "Google doesn't index CSS images" ⁸, meaning pictures applied as decoration through Cascading Style Sheets (CSS), the code that controls a page's look. Full-width hero photos, product shots set as section backgrounds and gallery tiles built as styled boxes are a familiar technique in website templates, because a background is easy to crop and layer text over. To a visitor they look identical to a normal image. To Google's image index they do not exist, and neither does the chance of a camera search landing on them.

The second silent failure is images that load only when a visitor acts. Deferring images until they are about to appear on screen, known as lazy-loading, is sound performance practice, but Google warns that a poor implementation can hide content from it ⁹. Loading has to happen when an image enters the visible area, never in response to a click or a scroll action, because Google Search does not interact with a page ⁹. A gallery that reveals its photos only after a click, or a slider whose later slides load on a swipe, can leave a business's best images outside the index. Google's own test is whether the image address appears in the src attribute of the image element in the rendered HTML ⁹.

Being indexed is not the same as being understood. Google states that it uses alt text, the short written description attached to an image, "along with computer vision algorithms and the contents of the page to understand the subject matter of the image" ⁸. It advises placing images near relevant text, on pages about the image's subject, and notes that a filename gives only very light clues ⁸. Here most websites are measurably behind. The WebAIM Million, an annual accessibility audit of the top one million home pages, found in February 2026 that 53.1% of them had images missing alternative text, and that more than one in four images on those pages had missing, questionable or repetitive alt text, such as the word "image" or a file name ¹⁰. A product photo described as IMG_0423, on a page that never names the product, gives Google's systems a picture and nothing to say about it.

Quality counts at the other end. Google notes that sharp, high-quality photos appeal to users more than blurry ones and are more likely to draw traffic from a result thumbnail, while warning that images are often the largest contributor to page size ⁸. A business needs both at once: photos good enough to be chosen, delivered light enough not to slow the page. Google Search supports modern formats such as the AV1 Image File Format (AVIF) and WebP ⁸, so a site can use them without giving up image search.

Inside most businesses, nobody owns this. Photography is commissioned for a brochure or social media, handed to whoever builds the website, and placed wherever it looks best, often as a background, with a generic description or none. Until 24 September there was no report that separated camera searches from typed ones ¹², so the loss had nowhere to show up. A customer photographed the product, found a competitor's page or a marketplace listing, and the business never knew the search happened. The new filter shows only traffic a site already receives ¹, so a low or empty count does not prove that nobody is searching. It can also mean the site's images are not being found.

So, does a business need to fix its website's images for camera search? If it sells anything a customer can photograph, yes: products, dishes, rooms, venues, finished work. Every image that shows what the business sells should be a real HTML image, loaded without waiting for a click, described in plain words, and placed on a page that names what it shows. Decorative textures and abstract backgrounds can stay as they are. The tie-break for anything in between: if a customer could plausibly point a camera at the real thing, the photo of it on the website has to be indexable.

Italian DesAIgns builds websites with semantic HTML and accessibility designed in from the first component, the same foundation an image index depends on: real image elements, meaningful descriptions and pages that say what they show, with Core Web Vitals prioritized so image quality does not cost speed. The reporting side of AI search, and what Search Console still leaves out, is covered in why Google shows AI impressions but not the clicks, the way AI agents read the same page structure in AI agents operating websites for customers, and crawlers that never run a page's scripts in why AI crawlers can't read JavaScript. The insights archive follows visual and AI search as both evolve. The free AI visibility check scores one page out of 100 on metadata and OpenGraph tags, heading structure, structured data and crawler access, then offers the full breakdown in a free consultation. It does not analyse a page's images, so whether a camera search can find them starts with a look at how the site was built.