The complete guide to image SEO in WordPress (2026 edition)

More than half of the web’s home pages ship images that search engines do not fully understand and that screen readers skip entirely — 53.1% of them, according to the 2026 WebAIM Million survey. The fix is rarely a plugin you install. It is a set of decisions you make once, then apply to every image you upload from then on.

This guide is the 2026 version of those decisions. Image SEO has changed more in the past eighteen months than in the five years before it: Google published a new rule in March 2026 for choosing which image represents a page, AI search engines started citing images in answers, and AVIF finally crossed into mainstream browser support. Most image-SEO advice still online predates all three.

What changed in 2026

Three shifts matter for anyone running an image-heavy WordPress site this year.

Google now documents how it picks your preferred image. In March 2026, Google Search Central added explicit guidance on specifying the image used for thumbnails in Search and Discover. You can signal your choice through structured data, the og:image meta tag, or the page’s main image. Before this, the choice felt like a black box. Now there is a documented lever, and most sites are not pulling it.

AI search reads images differently from how Google Images does. Tools like ChatGPT, Perplexity, Gemini, and Claude increasingly surface and reference images in their answers. Most of them have limited ability to run JavaScript, which means an image loaded by a script after the page renders may be invisible to them. Server-rendered markup and honest alt text matter more than they did a year ago.

AVIF is no longer the early-adopter format. Browser support sits around 93% in 2026, and AVIF files run roughly 20 to 30% smaller than the equivalent WebP. The trade-off is real (AVIF encodes far more slowly), but the format is now a legitimate default rather than an experiment.

The five layers of image SEO

Image SEO is easier to get right when you stop treating it as one task and start treating it as five layers, each building on the one below. Get the bottom layers wrong and the top ones cannot save you.

1. File — format, size, compression, and the filename itself.
2. Markup — the <img> tag: alt text, width and height, srcset, sizes, loading, and fetchpriority.
3. Schema — structured data that tells search engines what the image is and whether it represents the page.
4. Sitemap — listing your images so crawlers can find them, which matters most at gallery scale.
5. Page context — the text around the image, the caption, and the internal links that give the image meaning.

Layer one: the file

Format: AVIF, WebP, or JPEG

Serve modern formats with old-format fallbacks. The safe 2026 pattern is a <picture> element offering AVIF first, WebP next, and JPEG as the final fallback. WebP support is near-universal (around 97% of browsers); AVIF is close behind at around 93% and gives the smallest files, especially on photographs with skin tones, foliage, and gradients.

The catch worth knowing: AVIF encodes much more slowly than WebP. For a site uploading thousands of images, that encode cost is a real consideration. We cover the full decision in AVIF in 2026: should you switch from WebP?

Size and compression

An 800-pixel-wide WebP at 150KB beats a 2000-pixel JPEG at 2MB on every metric a search engine cares about, and your visitors will never see the difference. Resize to the largest size you actually display, then compress. Uploading a 4000-pixel camera original to show at 600 pixels is the single most common image mistake on the web.

The filename

Google reads filenames, and Google Images reads them more closely. IMG_4719.jpg tells a crawler nothing. navy-linen-wedding-suit.jpg tells it everything. Use hyphens (Google reads them as spaces), keep it lowercase, and use two to five descriptive words. We go deep on this in Image file naming for SEO — the rules nobody teaches.

Layer two: the markup

Alt text

Alt text does two jobs at once: it describes the image to anyone using a screen reader, and it tells search engines what the image shows. Good alt text serves both. The 2026 standard is 80 to 125 characters, descriptive, contextual, and natural — never stuffed with keywords, which Google treats as spam.

The scale of the problem is striking: the 2026 WebAIM Million survey found missing alt text on 53.1% of the web’s home pages. That is a low bar, and clearing it puts you ahead of most of your competition. Our full treatment, with twelve before-and-after examples, lives in How to write alt text that ranks.

Width, height, and layout stability

Always set width and height attributes. Without them, the browser does not know how much space to reserve, so the page jumps as images load. That jump is measured as Cumulative Layout Shift, one of Google’s Core Web Vitals, and it quietly hurts both rankings and the reading experience.

srcset and sizes

The srcset attribute offers the browser several image sizes; the sizes attribute tells it how wide the image will display. Provide srcset without sizes and the browser assumes the image fills the full viewport width, so it downloads an oversized file on mobile. The two belong together. WordPress generates both automatically when you use its image functions properly — which is one reason a well-built gallery plugin handles this for you.

loading and fetchpriority

Here is the rule that has reversed since 2021. Do not lazy-load every image. Your most important image — the largest one visible when the page first loads, known as the Largest Contentful Paint element — should load immediately. Lazy-loading it makes your page measurably slower.

Use loading="lazy" on below-the-fold images, and fetchpriority="high" on the one above-the-fold image that matters most. WordPress 6.3 and later does much of this automatically: it skips lazy-loading on the first images and adds fetchpriority="high" to the likely main image. The full explanation is in Lazy loading explained for WordPress site owners.

This is a place where the plugin you use does the work for you or quietly undoes it. FotoGrids handles the markup layer natively: it generates multiple sizes with proper srcset and sizes attributes on every upload, lazy-loads below-the-fold gallery images, and leaves the first row eager so your gallery is not the thing dragging down your Largest Contentful Paint. Responsive images and lazy-loading are in the free tier — there is no need to bolt on a second performance plugin to get the markup right.

Layer three: schema

Structured data tells a search engine what an image is in machine-readable terms. The core type is ImageObject, written as JSON-LD, which is Google’s preferred format. Useful properties include contentUrl, caption, creator, and the licensing fields that earn the “Licensable” badge in Google Images.

The honest take: not every image needs schema. A decorative photo in a blog post often does just as well with a good og:image tag. Schema earns its place on product pages, recipes, large article hero images, and licensable images. The March 2026 update added primaryImageOfPage and related properties as a way to nominate the image that represents the page — a genuinely new lever. We cover when schema helps and when it is overkill in Structured data for images: when it matters, when it doesn’t.

Here is what a minimal, correct ImageObject looks like in JSON-LD — the shape Google expects:

{
  "@context": "https://schema.org",
  "@type": "ImageObject",
  "contentUrl": "https://example.com/navy-linen-wedding-suit.jpg",
  "caption": "Navy linen three-piece wedding suit, tailored fit",
  "creator": { "@type": "Person", "name": "Your Studio" }
}

If hand-writing JSON-LD for every gallery sounds like a job you will never finish, this is exactly the kind of thing worth handing to the plugin. FotoGrids generates Schema.org ImageObject and ImageGallery markup for every gallery automatically in its Pro tiers, along with an image sitemap and Open Graph tags — so the schema layer is handled at the gallery level instead of image by image.

Layer four: the sitemap

An image sitemap lists your images so crawlers can find them even when they are loaded in ways a crawler might miss. For a single blog post this matters little. For a gallery, a portfolio, or a product catalogue, it matters a great deal — a single page can list up to 1,000 images in the sitemap.

This is the layer where image-heavy sites either pull ahead or fall behind. A 200-image gallery with no sitemap entry is 200 images a search engine may never index. We walk through gallery-scale sitemaps in How to optimize gallery images for SEO.

Layer five: page context

An image does not exist in isolation. The text around it, the caption beneath it, and the page it sits on all tell a search engine what the image means. A photo of a navy suit on a page about wedding tailoring is understood differently from the same photo on a page about fabric dyeing.

Captions are worth special mention because they are read by both people and crawlers, and because surprisingly few sites use them well. A caption is a second, visible chance to describe the image in context — and unlike alt text, your visitors actually see it.

The 2026 AI search layer

AI search sits on top of all five layers rather than replacing any of them. The practical implications are still emerging, and anyone claiming certainty about how ChatGPT or Perplexity choose which images to cite is guessing. What we can say with confidence is narrower but useful.

• Most AI crawlers have limited or no JavaScript rendering, so images need to be present in server-rendered HTML, not injected by a script after load.
• Alt text and captions are the text these systems read to understand an image — the same fields that serve accessibility and traditional SEO.
• Structured data is a discoverability layer for AI as much as for Google; an ImageObject with a clear caption and creator is easier to attribute.
• The crawlers identify themselves with user-agents (GPTBot, ClaudeBot, PerplexityBot, and ImagesiftBot among them), which gives you a robots.txt decision to make about who you allow.

The full picture, including the documented differences between the major engines, is in How AI search engines crawl images.

Image SEO inside WordPress, specifically

WordPress does some of this for you and leaves the rest to your theme, your gallery plugin, and your own habits. Three places to look.

In core: WordPress generates multiple image sizes with srcset and sizes, lazy-loads below-the-fold images, and since 6.3 handles fetchpriority for the likely main image. You get a solid baseline for free.

In your theme: themes vary wildly in how they output images. A poorly built theme can undo the core defaults — hard-coding image sizes, dropping srcset, or lazy-loading the hero. It is worth checking.

In your gallery plugin: this is where image-heavy sites live or die. A gallery is image SEO at scale, so the plugin’s handling of alt text, captions, formats, sitemaps, and lazy-loading compounds across every image. FotoGrids was built with this in mind: lazy-loading, responsive srcset images, and per-image captions are included in the free tier, and structured data, AI-generated alt text, and automatic WebP/AVIF conversion are available in the Pro tiers. The point is not which plugin you choose, but that you choose one that treats image SEO as a first-class concern rather than an afterthought.

The image SEO audit

Run this on any page that matters. Each item maps back to one of the five layers.

• Are images served as AVIF or WebP, with a JPEG fallback?
• Is every image resized to its display size, not the camera original?
• Do filenames describe the image in hyphenated, lowercase words?
• Does every meaningful image have descriptive, natural alt text?
• Do all images have width and height set?
• Is srcset always paired with sizes?
• Is the main above-the-fold image loaded eagerly with fetchpriority="high", not lazy-loaded?
• Does the page have appropriate ImageObject schema where it earns its place?
• Are gallery and catalogue images listed in an image sitemap?
• Do captions and surrounding text give each image context?

Your first hour: what to actually do

The full audit above is the thorough version. If you only have an hour, do these five things in this order — they are ranked by impact for a typical image-heavy WordPress site.

1. Fix your Largest Contentful Paint image. Run PageSpeed Insights, find the LCP element, and make sure it is not lazy-loaded. This is the single highest-impact change because it affects speed, Core Web Vitals, and rankings at once.
2. Turn on image sitemaps. Enable “include images” in your SEO plugin and confirm your galleries appear. For a portfolio or shop, this exposes images a search engine may never have indexed.
3. Add alt text to your top pages. Use the Media Library list view to find what is missing, and write natural descriptions for the images on your highest-traffic pages first.
4. Switch on modern image formats. Make sure you are serving WebP or AVIF with a fallback, either at the server, the CDN, or the plugin level.
5. Make your robots.txt decision about AI crawlers. Allow or block the AI bots on purpose, not by default.

Everything else on the audit is worth doing, but those five move the needle first. If you run your images through a gallery plugin, points one, two, and four are largely a matter of choosing one that handles them for you — which is the difference between image SEO being a recurring chore and being a setting you configure once.