Image Search: How Google Finds and Shows Images

Image search is not only “Google Images”: images also appear in the regular results and in Discover. Google’s recommendations for all of these places are shared, and they start not with alt text but with how a crawler finds an image at all and how it is presented. Below, per Google’s image SEO best practices: what makes an image discoverable, how to influence the choice of a page’s main image, which metadata brings the Licensable badge and how to read Search Console data. The basics of alt text, file names and context are covered in the lesson on image SEO; here is what belongs to search specifically.

How Google finds images

The technical requirements for content apply to images too, but they have their own particulars: per Google, how an image is placed and presented influences whether it is indexed at all, and for which queries.

• Standard HTML elements. Google finds images in the src attribute of an <img> element, including when it is a child of other elements such as <picture>. Google does not index CSS images (background-image).
• An image sitemap. URLs Google might not have discovered can be submitted in an image sitemap. Unlike regular sitemaps, <image:loc> may contain URLs from other domains, which suits CDNs; for a CDN Google advises verifying ownership of its domain in Search Console so you receive crawl error notices (the general logic is in the lesson on the XML sitemap).
• Responsive images. Not all browsers and crawlers understand srcset and <picture>, so always specify a fallback URL in src.
• Formats. Google supports BMP, GIF, JPEG, PNG, WebP, SVG and AVIF; the file extension should ideally match the type.

If images are loaded by script, check that they are in the rendered HTML (JavaScript SEO) and that lazy loading works without user actions. Speed matters indirectly too: images are often the largest part of a page’s weight (Core Web Vitals).

The page the image sits on

The content and metadata of the page where an image is embedded can have a great influence on where and how it appears. Google writes that choosing a preview image for a result or Discover is fully automated and draws on several sources. You can influence the choice by specifying a preferred image in one of these ways:

1. the schema.org primaryImageOfPage property with a URL or ImageObject;
2. an image property of the page’s main entity (mainEntity or mainEntityOfPage);
3. the og:image meta tag.

For the chosen image the help page advises picking one that is relevant and representative of the page, avoiding a generic image such as a logo or an image with text, and avoiding extreme aspect ratios. This is a recommendation, not a guarantee: the choice stays with Google. Markup in general is covered in the lesson on structured data.

Metadata and the Licensable badge

If you specify image metadata, Google Images can show more details: the creator, how the image may be used, credit information. For example, licensing information makes an image eligible for the Licensable badge, which links to the license and details of use (image metadata). There are two ways, and one is enough:

• Structured data. It associates the image with the page; add it for every instance of the image, even if the picture is the same.
• IPTC photo metadata inside the file: once per image is enough, and it travels with the picture.

If you use both and they conflict, Google uses the structured data. Pages with images must be viewable without an account and must not be blocked by robots.txt or a robots meta tag. Metadata makes an image eligible for the badge, it does not guarantee that it shows.

How to read Search Console data

The help page on the Performance report says: in search analytics data an image is simply a link to its host page, and Search Console does not distinguish between different images on the same page: clicks, impressions and position are counted identically. The same image can appear in both web and image results; data for search types is recorded separately and not combined. So you cannot judge “which image brought the click” from this report; the effect of work on images is seen in how the “Image” search type changes for a page.

What is not in the documentation

• The alt limit “up to 125 characters” — Google asks for useful, information-rich text without keyword stuffing.
• Shares like “alt, file name and WebP give 80% of the result” and traffic growth “by 340%” are stories, not Google data.
• Mandatory WebP or AVIF: they are in the list of supported formats but not named as required.

Checklist

1. Images are embedded with <img> tags with src, not CSS backgrounds.
2. A fallback src is specified for srcset and <picture>.
3. Important images have an image sitemap; with a CDN, domain ownership is verified.
4. A page’s preferred image is specified via primaryImageOfPage, image or og:image and is not a logo.
5. Licensable images have structured data or IPTC.
6. Results are viewed in Search Console by the “Image” search type and by page, not by individual image.

Mistakes

• Displaying pictures only via a CSS background and waiting for indexing.
• Blocking pages with images in robots.txt or requiring a login.
• Using a logo as the preferred image of every page.
• Adding web and image search data together: they are counted separately.
• Treating figures from success stories as mandatory benchmarks.

What next

Next comes video in search: video has its own discovery methods and markup, but the logic is the same — discoverability, a clear page and checking against data.