Images as content
Search engines cannot see pictures; they infer what an image shows from text and context, and increasingly from image analysis. Google's image documentation lists what it uses: the alt attribute, the file name, the page's title and headings, the text near the image, captions and structured data. Each is a place to say what the image is.
Alt text. The alt attribute describes the image for people who cannot see it and for engines. Write it
as a description a person would find useful: the subject, what is distinctive about it, the context if it
matters. "Navy wool winter coat, front view, buttoned" is alt text; "coat winter coat buy coat" is spam; an
empty alt (alt="") is correct for decorative images that carry no information. Alt text is not a caption
and not a keyword field. Google documents it as a signal for image understanding and for accessibility; the
accessibility obligation applies whether or not the search benefit does. Where alt text is drafted with AI across a large image library, the guide to AI for alt text, captions and transcripts covers the review that keeps it accurate.
File names. Descriptive, hyphenated, lower case: cast-iron-pan-seasoning-step-3.jpg. Set at upload or by
the template from the image's record. For an online store, the naming scheme is best agreed with the photographer before the shoot, as the
tutorial on briefing product photography does.
Captions and surrounding text. The paragraph the image sits in and its caption are strong context. Images placed next to the text they illustrate are understood; a gallery of unlabelled images is not.
Structured data. Product, recipe, article and other types include image properties, and Google
documents that rich results use them; an ImageObject with dimensions in the page's markup helps the engine
pick the right image for the listing. The structured data guide covers the
types.
Image sitemaps. For sites where images are a traffic source (retail, recipes, travel, stock imagery), an image sitemap or image entries in the main sitemap list the images per page so that ones loaded by script or served from a CDN are still discovered.
Google Images as a channel
Google Images is a results surface of its own, with its own intent: people searching visually for products, ideas, references and instructions. For some categories — furniture, fashion, recipes, crafts, destinations — it is a meaningful source of visits. Documented ways to be present: images that are indexable (not blocked in robots.txtA file at a site's root telling crawlers which URLs they may and may not fetch., not loaded only by scripts the crawler cannot run), high quality and large enough, described with alt text and surrounding content, and linked from pages that are themselves indexable. Image results also show the page title, so the page's title tag matters there too.
Images as a cost
Images are usually the largest bytes on a page and the usual cause of poor Core Web VitalsGoogle's user experience metrics.. The delivery rules:
| Problem | Effect | Fix |
|---|---|---|
| Images served larger than displayed | Wasted bytes; slow LCP | Resize to display size; responsive srcset and sizes so each device gets its own size |
| Old formats | Larger files for the same quality | Serve WebP or AVIF with a fallback; most build tools and CDNs do this automatically |
| No dimensions | Layout shifts as images load (CLS) | width and height attributes or CSS aspect ratio on every image |
| Everything loads at once | Slow first paint; wasted bandwidth | Native loading="lazy" on below-the-fold images; never on the first image |
| Hero image lazy-loaded or discovered late | Poor LCP | Eager-load the LCP image; a preload hint; put it in the initial HTML, not injected by script |
| Uncompressed uploads | Multi-megabyte pages | Compress on upload; set a size budget per template |
| Images blocked from crawling | Not indexed; render incomplete | Allow the image paths and CDN in robots.txt |
This site uses the framework's image component, which handles sizing, modern formats and lazy loading by default; the one decision left to authors is to mark the first image on a page as the priority one. The Core Web Vitals guide covers the measurement side.
Lazy loading done safely
Deferring images until they are near the viewport is right for everything below the fold. Three cautions
from Google's documentation and common experience: use native lazy loading or an IntersectionObserver
approach, not a scroll-event script, so crawlers that do not scroll still trigger the load; never lazy-load
the image that will be the largest contentful paint; and keep the dimensions set, because a lazy image
without a reserved box shifts the layout when it arrives.
Text in images
Text rendered inside an image — a chart's labels, a banner's headline, a diagram's annotations — is invisible to the engine and to screen readersSoftware that announces a page's text, headings, labels and messages to people who cannot see the screen, and reports what the page provides. unless repeated in alt text or the page. Charts and diagrams should carry a text equivalent: alt text summarising what the figure shows, a caption, or the data in a table. This site's diagram component renders flow diagrams as HTML text for that reason.
A furniture retailer's product images
A product template serves a single large image at desktop resolution to every device, lazy-loads it along
with the rest of the gallery, and names files by upload ID. Mobile LCP fails; Google Images shows competitors'
pictures of similar products but not the retailer's. The template change: the first image eager-loaded with
responsive sizes and a preload hint; the gallery lazy-loaded with reserved dimensions; alt text generated
from the product's name, colour and material; file names from the product record; ImageObject entries in
the product markup; and the image CDN path allowed in robots.txt, where a legacy rule had blocked it. Field
LCP improves across the template over the following weeks, and the product images begin to appear in image
results.
Choosing a format
| Format | Use it for | Notes |
|---|---|---|
| SVG | Logos, icons, simple diagrams | Scales without loss; tiny; text inside is readable by engines when inline |
| AVIF | Photographs and complex images | Smallest files at a given quality; wide browser support now; slower to encode |
| WebP | Photographs and graphics where AVIF is unavailable | Good compression; near-universal support; the common default |
| JPEG | Fallback for photographs | Universal; larger than the modern formats |
| PNG | Screenshots, images needing transparency where SVG does not apply | Large for photographs; use WebP or AVIF with alpha where supported |
| GIF | Nothing new | Replace animations with video or animated WebP/AVIF |
Serving the modern format with a fallback is done by the image component, the build pipeline or the image CDN, which negotiates by what the browser accepts; authors upload one high-quality source and the pipeline derives the rest.
Image CDNs and transformation
An image CDN or a transformation service takes a single source image and returns it resized, reformatted and compressed to the requested parameters on the fly, cached at the edge. For sites with large catalogues or user-uploaded images, this replaces per-image optimisation with a template rule: the product template requests the sizes it needs and the service produces them. The SEO requirements are that the transformed URLs are stable (the same image always has the same URL, so it can be indexed and cached), that the CDN host is crawlable, that the image sitemap or markup references the canonical image URL, and that the transformation does not strip the dimensions the template relies on for layout.
Licensing metadata
Google documents support for image licensing metadata — a licence URL and an acquisition URL supplied
through ImageObject structured data or IPTC photo metadata — which can surface a licensable badge in
image results and a link to where the image can be obtained. Photographers, stock libraries and
publishers with original imagery benefit directly; the same metadata helps any site establish that its
images are its own, which matters where product photographs are copied across retailers.
Auditing images at scale
A site-wide image audit runs from the crawl and the field data, by template:
- Alt coverage: share of images with no alt, with empty alt (acceptable only when decorative), with alt that is a file name or a keyword string.
- Dimensions: share of images without width and height or an aspect-ratio rule.
- Weight: distribution of file sizes by template; the largest offenders listed.
- Format: share served in legacy formats to browsers that accept modern ones.
- Loading: whether the first image on each template is eager and preloaded, and everything else lazy.
- Access: image hosts and paths allowed in robots.txt; images returned with cacheable headers.
- Field impact: the templates whose largest contentful paint element is an image, from the field data, which prioritises the fixes.
The output is a short list of template changes, not a list of thousands of images.
Checking a page's images
The SEO page checker reports images without alt text and without dimensions on any live URL. A site crawl reports the same across the site, plus oversized files and images blocked by robots.txt. Search Console's performance report, filtered to the Image search type, shows whether Google Images is sending clicks and for which queries. The audit sequence: alt coverage by template, dimension coverage, format and size distribution, the first image's loading behaviour, and crawl access to the image hosts.
When a hero image is the largest contentful paint element and the metric fails, the page on diagnosing a slow LCP shows whether size, lazy loading or late discovery is to blame.
Common mistakes
- Keyword-stuffed or missing alt text. Both fail the reader; the first also fails the guidelines.
- Lazy-loading the hero. A self-inflicted LCP failure.
- No dimensions. The most common and cheapest layout-shift cause.
- Desktop-size images for every device. Bytes wasted on every phone.
- Images on a blocked CDN. Invisible to the index and absent from renders.
- Important text only inside images. Invisible to engines and assistive technology.
- Treating images as decoration. For many categories they are the traffic.
The headings and structure guide covers where images belong in a page's structure; the e-commerce section covers product imagery as part of product pages.
