seo-images

Image optimization analysis for SEO and performance covering alt text, file sizes, formats, responsive images, lazy loading, CLS prevention, image SERP rankings via DataForSEO, and file-level WebP and AVIF conversion with IPTC metadata injection.

Audit and optimize images for rankings and Core Web Vitals

Source AgriciDaniel
License MIT
First documented

Trigger phrases

Phrases that activate this skill when typed to Claude Code:

  • image optimization
  • alt text audit
  • image SEO
  • image size
  • optimize images

What it does

seo-images is a Claude Code skill from AgriciDaniel’s claude-seo repo. It covers three distinct workflows: on-page image audit (alt text, file sizes, format analysis, responsive images, lazy loading correctness, CLS prevention via width/height attributes), image SERP analysis (which domains dominate Google Images for a keyword, what alt text patterns top-ranked images use — requires DataForSEO), and file-level image optimization (WebP/AVIF conversion using cwebp or ImageMagick, IPTC/XMP metadata injection using exiftool, responsive variant generation at 400w/800w/1200w).

A key distinction in the skill: it separates what actually matters for Google Images ranking (alt text — critical, filename — high, page context — high) from what does not (EXIF camera data — none, IPTC keywords — none, Google ignores these). It also flags the fetchpriority="high" pattern for LCP images and the loading="lazy" misuse — applying lazy loading to above-fold images directly harms LCP scores.

When to use it

Reach for it when:

  • You want to find all images on a page that are missing alt text, oversized, or in the wrong format
  • You want to know which domains dominate Google Images for a keyword and what their alt text patterns look like
  • You need to batch-convert a directory of images to WebP and inject IPTC copyright metadata for Google Images rich results

When not to reach for it:

  • Images are loaded via JavaScript or CSS background-image — the skill audits HTML <img> elements and will flag that JS-loaded images could not be fully assessed
  • The exiftool and cwebp binaries are not installed — the skill falls back to ImageMagick or FFmpeg for conversion but will note the reduced metadata capability

Install

Copy the seo-images SKILL.md into .claude/skills/seo-images/.

Trigger phrases: “image optimization”, “alt text”, “image SEO”, “image size”, “image audit”, “optimize images”, “image metadata”, “image SERP”, “convert to webp”, “image file optimize”.

Commands: /seo images <url> for on-page audit, /seo images serp <keyword> for Image SERP analysis (DataForSEO required), /seo images optimize <path> for file-level optimization.

What a session looks like

A typical session has three phases:

  1. On-page image audit. Every <img> element is evaluated: alt text presence and quality (10–125 characters, descriptive), file size against tiered thresholds (thumbnails under 50KB, content images under 100KB, heroes under 200KB), format (WebP/AVIF recommended over JPEG/PNG), srcset and sizes for responsive images, lazy loading correctness (above-fold images must NOT have loading="lazy"), and dimension attributes for CLS prevention.
  2. Prioritized optimization list. Images are sorted by potential byte savings (largest savings first), making it clear where to spend time. Each item lists current size, format, issues, and estimated savings from fixing it.
  3. File optimization pipeline (if requested). The full pipeline: audit existing IPTC/XMP metadata, inject Creator, Copyright, and Description fields, convert to WebP, generate responsive variants, verify metadata was preserved in the WebP output, and generate the <picture> HTML with AVIF > WebP > JPEG fallback chain.

Receipts

Works well: The lazy loading correctness check is the most commonly actionable finding on modern sites — it is easy for a CMS or developer to accidentally apply loading="lazy" to the hero image, which tanks LCP. This check catches it immediately.

Backfires: IPTC metadata injection via exiftool is flagged explicitly as “display only” — it affects how Google Images shows attribution information but has no ranking impact. Some users expect it to improve rankings; the skill is clear that it does not.

Pattern that works: Run the audit, fix alt text and lazy loading first (fastest, highest ranking impact), then tackle format conversion in a batch pass (biggest page weight reduction). Do not try to fix everything at once — a prioritized list is only useful if you follow the priority order.

Source and attribution

Originally written by AgriciDaniel. The canonical SKILL.md and supporting files live in the seo-images folder of the claude-seo repository.

License: MIT. Install, adapt, and redistribute with attribution preserved.

This page documents the skill from a practitioner’s perspective. For the formal spec and updates, defer to the source repo.