/projects//audit/ ).
1. Image Delivery Bottlenecks on Modern Web Stores
High-resolution product photography is essential for consumer conversions, but unoptimized 3MB PNG files destroy mobile load speeds. Modern technical SEO requires serving next-gen AVIF or WebP files with automatic fallback to JPEG for legacy browsers.
LIVE PLATFORM SCREEN & TELEMETRY
SEO Runway Site Audit Performance & Core Web Vitals — Active Workspace
Below is the exact production screen with live parameters and dummy data illustrating the diagnostic findings and remediation buttons.
2. Auditing Image Assets in SEO Runway
In SEO Runway → Site Audit, review the Asset Breakdown. The audit highlights images that exceed 200KB, images lacking dimension attributes (which cause CLS), and broken image links returning 404 errors.
Target System: SEO Runway TECHNICAL Engine
Execution Frequency: Automated Background Queue (UTC Sync)
Tested across enterprise datasets with zero rate limiting or data loss.
3. Why Images Are the Biggest Performance Opportunity
Images usually make up the largest share of a web page's bytes, and the largest visible element on many pages, the Largest Contentful Paint (LCP) element, is very often an image. Optimising images is therefore one of the most direct ways to improve LCP, reduce data usage on mobile and cut hosting and CDN costs. On sites with thousands of product photos or article images, manual optimisation is impossible, which is why the goal is an automated pipeline.
4. Choosing the Right Formats
- AVIF typically offers the best compression for photographs, often noticeably smaller than WebP at similar visual quality. Encoding is slower and more CPU-intensive, which matters for on-the-fly conversion.
- WebP is supported by all modern browsers and gives substantial savings over JPEG and PNG, with fast encoding. It is a safe default and a good fallback for AVIF.
- JPEG remains the universal fallback for photographs, best served with modern encoders such as MozJPEG.
- PNG is appropriate for graphics that need lossless quality, but WebP and AVIF both support transparency at much smaller sizes.
- SVG is ideal for logos, icons and simple illustrations, since it is resolution-independent and usually tiny.
A typical strategy is to serve AVIF where supported, WebP as the fallback, and JPEG or PNG only for very old browsers.
5. Serving the Right Format and Size Automatically
Format negotiation
There are two common ways to deliver modern formats:
- The
<picture>element lists sources in order of preference (image/avif, thenimage/webp) with an<img>fallback. The browser picks the first format it supports. This works with any static hosting. - Content negotiation at the server or CDN inspects the browser's
Acceptheader and returns the best supported format from a single URL. Image CDNs and many CDN image services do this automatically. The response must includeVary: Acceptso caches store separate versions.
Responsive sizes
Serving a 2,400-pixel image to a phone that displays it at 400 pixels wastes most of the bytes. Use srcset with several widths (for example 400, 800, 1200 and 1600 pixels) and a sizes attribute describing how wide the image is displayed at different breakpoints, so the browser downloads the smallest adequate file.
Quality settings
Choose compression quality per format and image type rather than a single global value. Product photos may need higher quality than background images. Many teams find that perceptually tuned settings, checked against a sample of real images, give large savings with no visible difference.
6. Building the Pipeline
- Store one high-quality original per image in your media library or object storage.
- Generate derivatives automatically, either at upload time (a build step or upload hook using a library such as Sharp or libvips) or on demand through an image CDN that transforms and caches images by URL parameters.
- Produce each size in each format needed by your templates, named or parameterised predictably.
- Cache aggressively at the CDN with long cache lifetimes and content-hashed or versioned URLs, so updated images get new URLs.
- Emit the correct markup from templates or components:
srcset,sizes,width,heightand the right loading attributes.
On-demand transformation is simplest for large existing libraries, because you don't need to batch-convert everything upfront. Upload-time generation gives more control and predictable costs.
7. Loading Priorities
- Lazy-load below-the-fold images with
loading="lazy", so they are only fetched as users scroll. - Never lazy-load the LCP image. Load it eagerly and mark it with
fetchpriority="high"so the browser fetches it early. Lazy-loading the hero image is one of the most common causes of poor LCP. - Preload critical background images that CSS references, since the browser can't discover them until stylesheets load.
- Always set width and height (or an aspect ratio) so the browser reserves space and avoids layout shift.
8. Image SEO Beyond Performance
Faster images help rankings indirectly through page experience, but images can also earn traffic directly through Google Images and visual search results. Make sure your pipeline doesn't undermine that:
- Keep descriptive file names such as
blue-linen-shirt-front.avifrather than random hashes alone. A content hash can be added to the path for caching without replacing the descriptive part. - Write useful alt text that describes the image for people using screen readers, including relevant product or topic details naturally.
- Use crawlable image URLs. Images loaded only through JavaScript or CSS backgrounds are harder for search engines to discover. Use
<img>elements for content images. - Keep image URLs stable when you change formats. If a migration changes every image URL, redirect the old ones or expect a temporary drop in image search traffic.
- Add images to structured data (Product, Article and Recipe schema) and consider image sitemaps if images are served from a separate CDN domain.
9. Measuring Results
- Total image bytes per page by template, before and after.
- LCP at the 75th percentile from field data, especially on mobile.
- The LCP element: check in PageSpeed Insights that it loads early and isn't lazy-loaded.
- CDN bandwidth and cost, which often fall dramatically once modern formats and responsive sizes are in place.
- Image search impressions and clicks in Search Console, filtered to the Image search type, to make sure the change didn't harm discovery.
10. Common Mistakes
- Converting everything to AVIF on the fly without caching, which overloads servers with encoding work.
- Forgetting
Vary: Acceptwith content negotiation, so caches serve AVIF to browsers that can't display it. - Over-compressing product images until texture and detail are lost.
- Serving one huge image to every device size.
- Lazy-loading the hero or LCP image.
- Missing width and height attributes, causing layout shift.
11. Image Optimisation Checklist
- One high-quality original stored per image.
- AVIF and WebP generated automatically with JPEG/PNG fallbacks.
- Responsive
srcsetandsizeson all content images. - Derivatives cached at the CDN with versioned URLs.
- LCP image eager-loaded with high fetch priority; others lazy-loaded.
- Width and height on every image.
- Descriptive file names and alt text retained.
- Field LCP and image search traffic monitored after rollout.
12. Handling User-Generated and Legacy Image Libraries
Sites with years of uploads, or with images uploaded by users and sellers, face extra challenges. Originals arrive in every size and format, often far larger than needed, sometimes with embedded colour profiles and camera metadata. A robust pipeline should:
- Normalise on upload: cap the maximum stored dimensions (for example 3,000 pixels on the longest side), convert colour profiles to sRGB and auto-rotate based on orientation metadata.
- Strip unnecessary metadata such as GPS coordinates and camera details, which add bytes and can expose private information.
- Reject or flag problem files, such as images too small to display well or files that aren't really images.
- Backfill gradually: for an existing library, let an image CDN transform images on first request, or run a background job that converts the most viewed images first.
13. Choosing Between an Image CDN and a Self-Built Pipeline
Image CDNs and CDN image services (offered by most major CDNs and by dedicated image platforms) handle format negotiation, resizing, caching and quality tuning through URL parameters, with little engineering effort. They are usually the fastest way to optimise a large existing library. The trade-offs are ongoing cost, which is often priced by transformations or bandwidth, and dependency on the provider's URL scheme.
A self-built pipeline using Sharp or libvips at upload time gives full control, predictable costs and portable URLs, but needs engineering time to build, test and maintain, including fallbacks and backfills. Many teams start with an image CDN to get quick wins, then move high-volume, stable assets (such as product photos) to pre-generated derivatives once they understand their usage patterns.
14. A Phased Rollout Plan
- Fix the LCP image on your top templates first: correct size, modern format, eager loading and high fetch priority.
- Add responsive
srcsetandsizesto product and article images. - Switch to AVIF and WebP with fallbacks through your CDN or pipeline.
- Add lazy loading for below-the-fold images and dimensions everywhere.
- Backfill the legacy library, starting with the most viewed images.
- Add automated checks that fail builds when templates output images without dimensions or exceed an image weight budget.
15. Setting an Image Performance Budget
Budgets keep image weight from creeping back after the initial clean-up. Practical budgets for most sites are: an LCP image under about 100–200 KB on mobile, total image weight on above-the-fold content under a few hundred kilobytes, and no individual content image larger than necessary for its displayed size at 2× pixel density. Enforce them in CI with Lighthouse or a custom check that inspects template output, and in the CMS by warning editors when they upload oversized originals or forget alt text.
Review the budget quarterly against field LCP data. If templates consistently pass the budget but LCP is still poor, the bottleneck has moved elsewhere (server response time, render-blocking CSS or fonts), and image work has done its job.
Image optimisation is rarely a one-off project. New templates, campaign pages and third-party widgets reintroduce oversized images regularly, so the automated pipeline and the budget together are what keep the gains in place over time.
A final practical tip: give editors and marketers a simple rule they can remember, such as "upload originals no wider than 3,000 pixels, always write alt text, and never place text inside images". Most image problems on content-heavy sites start in the CMS, not in code. A short guideline in the upload screen prevents more issues than any later optimisation can fix, and it makes the automated pipeline's job far easier. Revisit the guideline whenever you add new templates or image-heavy features, so it keeps matching how the site actually uses images.
Want the complete procedural runbook?
Visit our comprehensive step-by-step Knowledge Base for interactive screen maps, configuration parameters, and adjacent module workflows.
Open Knowledge Base Hub →

Incredible breakdown. We implemented this exact workflow last week and already saw a noticeable shift in our telemetry. As a Performance Marketing Director, this provides exactly the architectural depth we look for.