Visibility in Google Images and Google Lens

Images can be found in two ways on Google: people type a query and browse Google Images, or they search with an image itself through Google Lens, Circle to Search, an uploaded photo or Chrome’s “Search this image”. The first is classic image search; the second is what Google calls multimodal search. Both depend on Google being able to find the image, understand what it shows and connect it to a useful page. There is no published ranking formula for Lens. What Google does publish is how it discovers and understands images, and, since September 24, 2026, Search Console reports that separate image-led searches from text searches.

How Google finds and understands images

Google’s image documentation sets out the basics (Google’s image SEO best practices):

• Discovery: Google finds images in the src attribute of <img> elements, including inside <picture>. It states that it “doesn’t index CSS images”. An image sitemap can help Google discover images it might otherwise miss, including images hosted on a CDN.
• Formats: Google supports BMP, GIF, JPEG, PNG, WebP, SVG and AVIF, and suggests the filename extension match the file type.
• Alt text: Google calls alt text the most important attribute for providing metadata about an image, says it uses alt text “along with computer vision algorithms and the contents of the page to understand the subject matter of the image”, and notes alt text acts as anchor text when an image is a link. Alt text is also where accessibility and search overlap: the same description serves screen-reader users.
• Context: the content and metadata of the page where an image appears shape how Google understands it, including the page title and description.
• Filenames and titles: Google recommends descriptive filenames, titles and alt text.
• Quality: Google notes that high-quality, sharp images appeal more to users than blurry ones.

Google’s lazy-loading guidance also matters: images that load only after user interaction may never be seen, because “Google Search does not interact with your page” (Google on lazy-loading).

Lens and multimodal search

When someone searches with an image, Google matches what is in the image to web content. A page benefits when:

• the product, place, plant, artwork or object in its images is clearly shown;
• the page around the image identifies it in text (product name, model, species, location);
• the page is indexable and has useful information about the thing in the image.

These are inferences from how Google says it understands images; Google has not published a Lens-specific ranking method, and no guide can promise Lens visibility.

Measuring visual search

On September 24, 2026, Google introduced Search Console reporting for “web multimodal search” in both the Search results performance report and the Generative AI features report. Google says the data includes “searches with Lens, Circle to Search on Android, image uploads to Google Search”, and Chrome’s right-click Search this image feature, and is available through a new multimodal search type filter (Google’s September 2026 announcement).

Google notes that the multimodal data appears only if a site receives traffic from these searches. The Generative AI report documents impressions; it does not document clicks.

A visual search measurement workflow

1. Establish the baseline. Export the last three months for the Image search type and for Web: multimodal and Web: text-based. Google’s announcement does not say how far back the multimodal data goes, so note the earliest date the multimodal filter returns before treating the period as a baseline.
2. Find the pages that matter. In the Pages dimension under Web: multimodal, list pages receiving image-led traffic. They are the pages Lens and similar features already connect to.
3. Check those pages. Are the images high quality, in <img> elements, described in alt text and in the surrounding copy? Is the object identified by name?
4. Check AI feature exposure. In the Generative AI performance report, filter to Web: multimodal to see which pages appear in AI features for image-led searches.
5. Improve and annotate. Make changes to images and context on a set of pages; add a Search Console annotation with the date.
6. Compare. After recrawling, compare the multimodal metrics for the changed pages with similar unchanged pages over the same period.

Image delivery versus discoverability

Fast image delivery (formats, compression, sizing, lazy loading) helps users and Core Web Vitals. It is a separate task from discoverability, though the two meet in a few places: Google cannot see images loaded only through CSS, through interaction-dependent lazy loading or behind blocked URLs. Optimize delivery without breaking those conditions.

Moving product photos out of CSS backgrounds

Take a hypothetical furniture retailer whose product photos are set as CSS backgrounds in a carousel, so they carry no alt text at all, and whose only visible label near each photo is the product SKU.

• Finding: Google Images traffic is near zero; the Web: multimodal filter shows almost no traffic to product pages.
• Changes: render the main product image in an <img> element; write alt text naming the product and its key visible features (“Oak dining table with tapered legs, seats six”); add the product name, material and dimensions near the image; include images in an image sitemap.
• Measurement: after recrawl, Image and Web: multimodal search types are compared with the baseline for the changed product pages.
• Limitation: improved discoverability makes the images eligible and understandable; it does not guarantee they will be chosen for a given visual search.

Visual search checklist

• [ ] Important images in <img> elements, not CSS backgrounds.
• [ ] Supported formats; extensions match types.
• [ ] Descriptive alt text, filenames and nearby text naming what the image shows.
• [ ] Lazy loading that works without interaction.
• [ ] Image sitemap for images Google may not discover.
• [ ] Multimodal and Image search types reported separately from text search.
• [ ] Generative AI report checked with the multimodal filter.