The image SEO checklist that still works

Image SEO advice has a half-life problem. Most articles on the subject are recycling tips that mattered in 2014, which is how you end up with people carefully hyphenating filenames while their largest image is a 4 MB PNG that tanks their Core Web Vitals.

Here is what still matters, roughly in order of impact.

The things that move the needle

1. Serve the right format, at the right size

This is the highest-impact item on the list and it is not about keywords at all. An oversized image is the most common cause of a poor Largest Contentful Paint, LCP is a ranking signal, and it is also the reason people leave before your page has drawn.

• Use WebP or AVIF. Both are supported everywhere that matters now.
• Serve the size you actually display. A 3000px-wide file rendered in a 600px column is 25× the bytes you needed.
• Use srcset so phones get phone-sized images.

2. Alt text that describes the image’s function

Not its contents — its function. The same photo needs different alt text depending on why it is on the page. This is a genuine ranking signal for image search, and it is the accessible thing to do, which is a rare and pleasant alignment. The details are in how to write good alt text.

The single most common mistake is keyword stuffing, which has not worked in over a decade and actively harms the people alt text exists for.

3. Lazy-load below the fold, and only below the fold

loading="lazy" on offscreen images. Never on your LCP image — lazy-loading your hero image delays the very thing the metric measures, and people do this constantly.

4. Always set width and height

Or an aspect ratio. Without them the browser does not know how much space to reserve, the page jumps as images land, and your Cumulative Layout Shift score suffers. Two attributes. Free win.

5. Descriptive filenames

cobalt-running-shoe-side.jpg beats IMG_4032.jpg. The effect is small but real, and it costs nothing at export time. Do not agonise over it; do not skip it.

6. Put the image near the text that explains it

Search engines use the surrounding content to understand an image. An image dropped into a page with no related text nearby is an image with no context. Captions help here, and they are read far more than body copy.

7. Structured data where it applies

ImageObject, and — much more importantly — the image fields inside Product, Recipe and Article schema. This is how you become eligible for rich results, and rich results are where the clicks are.

8. An image sitemap, if images are your business

For a normal site, your regular sitemap is fine. For a stock library, a recipe site or a large catalogue, image sitemap entries help Google discover images it would otherwise miss — particularly images loaded by JavaScript.

9. Do not block your images in robots.txt

Sounds obvious. It happens all the time, usually because someone blocked /assets/ to save crawl budget and took every product image down with it. Check yours.

10. Use a CDN

Latency is a ranking factor by way of LCP, and images are the heaviest thing on most pages.

11. Compress, but look at the result

Quality 80 is usually indistinguishable from quality 100 at a third of the bytes. Quality 40 is visibly ugly, and a visibly ugly product photo costs you more than a slow one.

12. Keep the metadata that matters, strip what does not

Strip EXIF on publish — it carries GPS coordinates and camera serial numbers you probably did not mean to publish. Keep copyright and creator. Most “optimise for web” pipelines strip everything, including your rights statement.

Advice you can stop following

“Exact-match keyword filenames.” buy-cheap-running-shoes-online.jpg does nothing except make you look desperate. Describe the image; that is enough.

“Put your keyword in the alt text of every image.” This is how you end up with the same keyword read aloud eleven times to a screen reader user. It does not rank you; it just makes your page hostile.

“Image title attributes are an SEO signal.” They are not. The title attribute produces a tooltip that most users never see and most screen readers ignore. Fill it in if you have a reason. It is not an SEO tactic.

“More images ranks better.” Only if they are useful. A page padded with stock photos is a slower page.

The bit that gets skipped

Everything above is per-image work, and per-image work is exactly what does not happen when a site has ten thousand images. Which is why most image SEO advice is technically correct and practically ignored: it assumes the images are a manageable number, and they never are.

If you are auditing an existing site, the realistic sequence is:

1. Find the images with no alt text at all. Any crawler will list them. This is your denominator.
2. Fix the templates first. Nine times out of ten a handful of templates generate most of the missing alt text, and one fix covers thousands of pages.
3. Generate drafts for the rest, then edit. An AI pass over a CSV of your image URLs will produce a factual description for every one of them in an afternoon — PicsTag does this in the browser and exports a CSV you can join back onto your CMS. The drafts are not perfect. Empty alt attributes are worse.
4. Then do the filenames, the structured data and the sitemap.

In that order, because item 1 is worth more than items 5 through 12 combined, and most people do it last.