Image SEO Audit: A 30-Day Plan That Finds What Actually Matters

What an image SEO audit is (and is not)

An image SEO audit answers one question: across your pages and listings, which images are costing you traffic or conversions, and why? It is not a checklist you apply blindly to every file. A shop with 40 listings should not spend a month renaming 800 photos; it should fix the 12 images that appear in search results and product grids.

The output of a good audit is a ranked list of fixes with an expected effect and an effort estimate. If your audit produces a 60-item checklist with no order, it is not an audit — it is procrastination with a spreadsheet.

Scope it by surface: pages you own (a store, a blog) and listings you rent (Etsy, Amazon, eBay) behave differently. On your own site you control filenames, alt text, formats, caching and structured data. On a marketplace you mostly control the image itself and the order of the gallery.

Week 1: build the inventory

Start from what exists, not from what you remember uploading. Export a list of every URL or listing that contains images, then record six fields per page: URL, page type, number of images, total image weight, largest image weight, and whether the main image appears in search results.

For your own site, the fastest source is a crawl: Screaming Frog, Sitebulb or a simple sitemap crawl will give you image URLs, sizes and alt attributes. For marketplaces, export your listings from the seller dashboard — most platforms let you download a CSV of active listings.

Do not skip the 'largest image weight' column. In most catalogues, 10% of images account for 70% of the page weight, and those are the ones that move Largest Contentful Paint.

Week 2: prioritise with an impact matrix

Score every page on two axes: business value (traffic, revenue, strategic importance) and fixability (how many changes are needed). Fix only the pages that are both valuable and cheap to fix. Everything else goes on a backlog with a date.

A practical rule: start with pages that already get impressions in Search Console but have a weak click-through rate. They already rank, so a better main image or a tighter crop can convert existing visibility into clicks without waiting for a new page to be indexed.

Week 3: the technical pass

Work through five checks in this order, because each one can invalidate the next: format and weight, dimensions, main-image clarity, alt text, and metadata. If a hero image is 4 MB, rewriting its alt text is irrelevant until the file is resized.

For format and weight, resize to the largest size the page actually displays (roughly 2× for retina), then compress. For dimensions, check that the intrinsic width and height are set so the layout does not shift while the image loads. For the main image, view it at 200 px wide on a phone; if the product is not identifiable, the thumbnail is the problem.

For alt text, one sentence per meaningful image, describing object, detail and context. For metadata, verify that the fields you injected actually survived — on your own site this is under your control; on a marketplace it may not be, which is why the next guide exists.

• Largest image on the page is under the display size, not the camera size.
• Every image has width and height attributes (or an aspect-ratio box).
• Hero image is not lazy-loaded; below-the-fold images are.
• Alt text is unique per image and under ~125 characters.
• Filenames are descriptive, lowercase and hyphenated.
• Metadata fields are populated and verified, not assumed.

Week 4: measure, then decide what not to do

Pick two numbers and watch them for 30 days: image-search clicks in Search Console and Largest Contentful Paint on your top three pages. If neither moves, the changes were cosmetic for your catalogue and you should stop before doing more of them.

Do not judge by file size alone. A page can lose 40% of its image weight and keep the same LCP because the bottleneck was render-blocking CSS or a slow server response. The audit exists to find the binding constraint, not to make a number smaller for its own sake.

Write down what you deliberately decided not to fix and why. That list is the most valuable part of the audit next quarter, when the same suggestions come back.

Common false positives

• “Every image must be WEBP.” Not on marketplaces that reject it; export per channel.
• “Alt text must contain the keyword.” It must describe the image; the keyword should appear naturally, not be forced.
• “Compress everything to quality 60.” Text, gradients and skin tones show artefacts early.
• “Rename every file again.” If a URL is already indexed, renaming forces a redirect for no gain.
• “More images always rank better.” Ten near-duplicates are worse than five distinct angles.

Frequently asked questions

How long should an image SEO audit take?

For a shop with under 100 listings, one focused day for the inventory and one day for fixes. The 30-day part is measurement, not work.

Do I need paid tools?

No. Search Console, PageSpeed Insights and a browser network panel cover most of it. A crawler helps above a few hundred pages.

Should I audit before or after a redesign?

Before. An audit tells the redesign what to preserve — URLs, alt text, image dimensions — instead of discovering it after launch.

What is the single highest-impact fix?

Usually resizing the largest image on your highest-traffic page. It is cheap, immediate and measurable in LCP.

How often should I repeat the audit?

Quarterly for active catalogues, or after any platform change that affects image handling.