Image SEO Best Practices: How to Rank Images on Google in 2026

What is image SEO?

Image SEO is the process of optimizing the images on your site so that search engines can find them, understand what they show, and surface them in image search, web search, and AI-generated answers. It touches three layers:

• Technical: file format, file size, dimensions, responsive markup, lazy loading, CDN delivery.
• On-page: file name, alt text, captions, the text that surrounds the image, the heading hierarchy of the page.
• Discovery: structured data, image sitemaps, internal linking, and making sure crawlers are not blocked from your image folders.

Get all three right and Google can render a fast page, understand what each image depicts, and confidently rank it. Get any one wrong and the others stop compounding.

Why image SEO matters more in 2026

Search has changed shape. The same query that returned ten blue links five years ago now returns an AI Overview, a product carousel with images, a "People also ask" panel, and a row of Google Images thumbnails. Visual results compete with text for the same screen real estate, and often win.

• Visual discovery is mainstream. Google Lens handles billions of visual searches every month, and Google Images remains the second-largest search property after Google Web Search.
• AI Overviews pull images. Google's AI Overviews regularly include thumbnails sourced from indexed images. Pages with strong image signals get cited; pages with broken or missing alt text do not.
• Core Web Vitals gate rankings. Largest Contentful Paint (LCP) is almost always an image. A poorly optimized hero image can tank your scores across the entire site.
• Ecommerce relies on it. Shopping results, product rich snippets, and Merchant Center listings all need clean, structured, accessible image data.
• Accessibility overlaps with SEO. Every accessibility-driven change you make to images (alt text, contrast, semantic markup) also improves SEO, and vice versa.

File names and URL structure

Google explicitly recommends descriptive file names. The URL path is one of the few signals available before the page is even rendered, and it shows up in image search snippets.

Rules of thumb

• Use lowercase letters and hyphens. Avoid spaces, underscores, and capital letters.
• Describe the content, not the workflow. "hero-banner-final-v3.jpg" is internal trivia; "sourdough-loaf-on-cooling-rack.jpg" is signal.
• Keep file names short but specific. Three to seven hyphenated words is a good range.
• Reuse the same canonical URL for the same image. Reuploading under a new filename throws away every link and indexing signal you had.
• Group images into folders that mirror your site structure (/images/blog/, /images/products/) so crawlers can map the topical hierarchy.

Pick the right image format

The format you choose determines file size, render quality, and browser compatibility. In 2026 the answer is almost always WebP for photography and UI imagery, with AVIF as a higher-compression option where toolchain support is there, and SVG for logos and icons.

Use the <picture> element when you want to serve AVIF to browsers that support it and fall back to WebP for everything else:

<picture>
  <source srcset="/images/hero.avif" type="image/avif">
  <source srcset="/images/hero.webp" type="image/webp">
  <img
    src="/images/hero.jpg"
    alt="Sourdough loaf on a wire cooling rack with steam rising"
    width="1200"
    height="630"
    loading="lazy"
  />
</picture>

Compression and file size

Page weight directly affects Largest Contentful Paint, and LCP is part of Google's Core Web Vitals. A bloated hero image drags down your ranking on every keyword the page targets, not just image search.

• Target: aim for hero images under 200 KB and inline content images under 100 KB. Above-the-fold imagery should ideally render in under 2.5 seconds on a mid-tier mobile device.
• Resize before compression. Most "image SEO is hard" complaints come from sites serving 4000-pixel-wide originals into a 600-pixel slot. Resize first, compress second.
• Use the right encoder settings. WebP at quality 75–82 is visually indistinguishable from the original for almost all photography. AVIF at quality 50–60 is similar.
• Strip unnecessary metadata. Camera EXIF data and ICC profiles can add 50+ KB to a single image. Tools like imagemin, squoosh, or your CDN's image API will strip them automatically.
• Audit with Lighthouse. Chrome's Lighthouse panel flags every oversized image with a specific savings number. Fix the biggest offenders first.

Responsive images and lazy loading

Serving the same image to a phone and a 4K monitor wastes bandwidth in both directions. The native HTML solution is srcset and sizes:

<img
  src="/images/product-800.webp"
  srcset="
    /images/product-400.webp   400w,
    /images/product-800.webp   800w,
    /images/product-1600.webp 1600w
  "
  sizes="(max-width: 600px) 100vw, 800px"
  alt="Hand-thrown ceramic mug in matte sage green, 12oz, three-quarter view"
  width="800"
  height="800"
  loading="lazy"
  decoding="async"
/>

• Always declare width and height. Without them the browser cannot reserve space, which causes Cumulative Layout Shift (another Core Web Vital).
• Use loading="lazy" for below-the-fold images. Native lazy loading is supported in every major browser and requires zero JavaScript.
• Do not lazy-load your LCP image. The hero image at the top of the page should load eagerly. Lazy-loading it tells the browser to deprioritize it, which is the opposite of what you want.
• Add decoding="async" on non-critical images to keep image decoding off the main thread.
• Use a CDN with image transformations. Cloudflare Images, Cloudinary, imgix, and Bunny.net can generate the WebP/AVIF variants and the right sizes on demand from a single source file.

Alt text: the highest-ROI image signal

If you only fix one thing on this list, fix alt text. Google has been explicit for over a decade that the alt attribute is a primary ranking signal for image search, and the WebAIM Million study finds that over a third of all images on the web still lack it. That is unusual leverage: a fixable signal most of your competitors are getting wrong.

Good image SEO alt text is descriptive, specific, keyword-aware without being keyword-stuffed, and matches the surrounding context. We have a complete walkthrough in What is Alt Text? The Complete Guide, but the short version:

• Describe what the image actually shows. "A golden retriever puppy on a wooden park bench", not "dog".
• Include the relevant keyword only if it accurately describes the image.
• For products, include the brand, product name, and distinguishing attributes — these match how shoppers actually search.
• Skip the prefix "image of" or "photo of"; screen readers already announce the element type.
• Use empty alt="" on decorative images so they are skipped, not announced as "graphic."
• Write unique alt text for every meaningful image. Duplicating descriptions across a gallery wastes the signal.

Captions and surrounding context

Google's image documentation calls out captions, page title, and the text that immediately surrounds an image as ranking signals. Crawlers use the page context to validate alt text and to understand images that have none. A photo of a sneaker placed next to a paragraph about hiking boots is going to confuse the model.

• Use real captions. A caption beneath the image is read by users and indexed by search engines. The HTML5 <figure> and <figcaption> elements are the semantic way to do it.
• Keep alt text and captions different. A caption usually adds context ("Mt. Hood from the Timberline trail, August 2025"); alt text describes what is in the frame ("A snow-capped Mt. Hood under a clear blue sky").
• Place images near relevant text. An image embedded in the section that talks about it ranks better than one shoved into a sidebar.
• Use the title attribute sparingly. It is mostly a tooltip; do not rely on it as a substitute for alt text.

Structured data for images

Schema.org markup is how you tell Google explicitly what an image represents and what rich-result types it can appear in. The three highest-ROI types are:

A minimal Article example with image:

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Sourdough Crust Troubleshooting",
  "image": [
    "https://example.com/images/loaf-1x1.webp",
    "https://example.com/images/loaf-4x3.webp",
    "https://example.com/images/loaf-16x9.webp"
  ],
  "datePublished": "2026-04-22"
}

Image sitemaps

An image sitemap helps Google discover images it might otherwise miss — those loaded by JavaScript, rendered from a CDN with an unusual hostname, or buried deep in paginated archives. You can either extend your existing XML sitemap with the image:image namespace, or publish a separate sitemap.

<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9"
        xmlns:image="http://www.google.com/schemas/sitemap-image/1.1">
  <url>
    <loc>https://example.com/blog/sourdough</loc>
    <image:image>
      <image:loc>https://example.com/images/loaf.webp</image:loc>
    </image:image>
  </url>
</urlset>

• Make sure your robots.txt does not block the image folder. A surprising number of "missing from Google Images" cases come down to a stale Disallow: /images/ rule.
• Submit the sitemap in Google Search Console and watch the Image indexing report.
• If you serve images from a separate hostname (a CDN), verify that hostname in Search Console too.

Optimizing for AI search and Google Lens

AI Overviews, ChatGPT search, Perplexity, and Google Lens are all multimodal — they read both the image and the text describing it. They depend even more on alt text and surrounding context than traditional crawlers do, because they are trying to synthesize an answer, not just match a query.

• Treat alt text as a semantic anchor. Even multimodal models cross-reference alt text against the visual data to resolve ambiguity.
• Lead with direct answers. AI engines pull the first 40–60 words after a heading. If your image illustrates a how-to step, put the step description directly above or below the image.
• Use FAQ schema where it fits. Pages with FAQ markup are cited at higher rates by AI Overviews.
• Do not block reputable AI crawlers if you want them to cite you. OAI-SearchBot, PerplexityBot, and ClaudeBot are the cite-and-link kind, not the train-only kind.

We go deeper on this in our guide to Generative Engine Optimization (GEO).

Common image SEO mistakes

The image SEO checklist

A copy-paste checklist for every image you publish.

Frequently asked questions