Image SEO covers two separate goals that get conflated. One is ranking in image search, which drives traffic for some categories and almost none for others. The other is images not damaging the page they sit on, which applies universally and matters considerably more.
The second goal is where most of the value is, because unoptimised images are the leading cause of Core Web Vitals failures and the most common source of avoidable page weight.
Alt Text Is an Accessibility Feature First
Alt text exists so screen reader users know what an image shows. Writing it for that purpose produces good SEO alt text as a byproduct, and writing it for search produces bad alt text for everyone.
Describe what the image shows, specifically, in a sentence someone could visualise from. A photo of a dashboard should say what the dashboard displays, not repeat the page's target keyword.
Decorative images that carry no information should have empty alt attributes, written as alt="" rather than omitted entirely. An omitted attribute causes screen readers to announce the filename, which is worse than silence. This distinction is routinely missed and it is the one genuine technical rule in alt text.
Length worth aiming at: enough to describe the image, usually under about 125 characters, because some screen readers truncate beyond that.
Filenames Are Free and Universally Ignored
An image uploaded as IMG_4471.jpg tells nothing to anything. Renaming it before upload costs seconds and it is one of the few remaining optimisations that is both trivial and skipped almost everywhere.
Descriptive, hyphenated, lowercase. The same conventions as URLs, and for the same reason.
This matters more in categories where image search drives genuine traffic, which is a narrower set than people assume. Recipes, products, travel, interiors, and anything where people search visually. For most B2B content, image search traffic is negligible and the filename work is worth doing anyway because it costs nothing.
File Size Is Where the Real Damage Happens
Images are the most common cause of Largest Contentful Paint failures, and LCP is the metric most sites fail after INP, per the Core Web Vitals guide.
Three problems account for most of it. Serving an image at far larger dimensions than the container displays, so a 4000 pixel wide file renders into an 800 pixel slot. Using older formats where modern ones would be substantially smaller at the same visual quality. And no compression at all, because the file went from a design tool straight to a CMS.
Modern formats handle this well and support is now broad enough that using them as a default with a fallback is straightforward. Responsive image markup lets the browser choose an appropriate size for the viewport rather than downloading the largest available.
The single highest impact change on most sites: stop serving desktop sized images to phones. That alone resolves a meaningful share of mobile LCP failures.
Explicit Dimensions Prevent Layout Shift
Width and height attributes on every image let the browser reserve space before the file arrives. Without them, content moves as images load, which is exactly what Cumulative Layout Shift measures.
This is the cheapest Core Web Vitals fix available and it is mechanical. It requires no judgement and no design change, and it is missing from a large share of templates.
The attributes should reflect the image's intrinsic dimensions, with CSS handling the display size. Setting them to the display size defeats the purpose when the image is responsive.
Lazy Loading, With One Important Exception
Deferring below the fold images until they are needed reduces initial page weight substantially, and native lazy loading makes it a single attribute.
The exception that catches people: never lazy load the LCP image. Applying lazy loading globally through a template or plugin delays exactly the element being measured, and it makes the metric worse rather than better.
The LCP element is usually the hero image or the first substantial image above the fold. Identify it, exclude it from lazy loading, and consider preloading it so the browser discovers it early.
Image Sitemaps Are Situational
A dedicated image sitemap, or image entries within your standard sitemap, helps crawlers discover images that are not easily found through normal parsing. That primarily means images loaded via JavaScript or referenced in CSS.
For a site where images sit in standard img tags within crawlable HTML, an image sitemap adds little, since discovery already works. The sitemap guide covers where sitemap effort is best spent generally.
Where it genuinely helps: image heavy sites with JavaScript galleries, and ecommerce catalogues where product images load dynamically.
Structured Data for Images
ImageObject schema describes an image's properties, and image URLs within Product, Recipe, or Article schema connect images to the entity they depict.
For most content, the practical value sits in the parent schema rather than in standalone ImageObject markup. A product page with images correctly referenced in Product schema is doing the useful work, per the schema guide.
Standalone ImageObject markup matters more for licensing and attribution use cases, where declaring a licence and creator is the point.
Surrounding Context Matters More Than the Image File
Search engines understand an image partly through what surrounds it: the caption, the nearby text, the heading of the section it sits in, and the page's overall subject.
An image placed in a section about a specific topic, with a caption describing it, carries more contextual signal than a technically perfect image floating in an unrelated part of the page.
Captions are also read more than body text, which makes them a genuinely useful place for information rather than a place to repeat the alt text. Different content in the caption than in the alt attribute serves both purposes properly. This is the same specificity discipline covered in the SEO copywriting guide, applied to a field most people fill in mechanically.
Original Images Beat Stock Where It Is Feasible
Stock photography appears identically across thousands of sites, which gives it no differentiating value and, in image search, means competing against every other site using the same file.
Original images, including screenshots, diagrams, and photographs of actual work, are unique by construction. That is the same non commodity argument covered in the non-commodity content guide, applied to visual assets.
For technical content specifically, a screenshot or diagram that explains something the text struggles to explain is doing real work rather than decorating.
Delivery Matters as Much as the File
An optimised image served slowly is still slow. Image CDNs handle format negotiation, resizing, and compression automatically at request time, which removes the dependency on whoever uploaded the file doing it correctly.
That automation is the strongest argument for using one. Manual image optimisation works until someone uploads a file without following the process, which happens continuously on any site with multiple contributors.
Most modern hosting platforms and content management systems now include some version of this. Checking whether yours does, and whether it is actually enabled, is a five minute task that frequently resolves a problem people are planning to solve manually.
Where Image Search Actually Drives Traffic
Being honest about this saves effort. Image search traffic concentrates in visual categories: products, recipes, home and interiors, travel, fashion, and anything where the decision is partly aesthetic.
For B2B software, professional services, and most informational content, image search traffic is marginal. The optimisation is still worth doing for page performance and accessibility reasons, and expecting traffic from it in those categories will disappoint.
Check your own position rather than assuming. Search Console's performance report has a search type filter that isolates image search, which tells you directly whether this channel matters for your site.
Verifying Images Are Reachable
Images loaded through JavaScript galleries or lazy loading implementations that depend on scripts may not be discoverable by crawlers that do not execute JavaScript, which is the situation covered in the rendering guide.
The NotionCue AI Crawler Audit reports what crawlers receive from a URL, which surfaces images referenced only after rendering.
Start your free NotionCue trial and check an image heavy template. Gallery components are a common place where assets exist for users and not for anything automated.
Fastest audit available: open your heaviest page, check the network tab, and sort by size. If the largest files are images being served at several times their display dimensions, you have found the single most valuable fix on that template.
Common Questions
Does alt text help rankings for the page, not just the image?
Marginally, as one contextual signal among many. Its primary functions are accessibility and image search, and treating it as a ranking lever for the page leads to keyword stuffed alt text that serves neither purpose.
Should every image have alt text?
Every informative image, yes. Purely decorative images should have an empty alt attribute rather than a description, so screen readers skip them.
Do images need to be in a specific format?
Use modern formats where support allows and provide a fallback. The specific format matters less than the file actually being compressed and sized for its container, which is where most of the weight problem originates.