How to rank in Google Images

Google Images is traffic people forget exists, and it converts unusually well: someone searching for a picture of a thing is often about to buy, cook, build or book that thing. Ranking there is not a vanity metric.

Here is what Google actually uses to understand an image, in rough order of how much it matters — with the outdated advice removed.

Google cannot see your image (mostly)

Start here, because it explains everything else. Google’s understanding of an image comes overwhelmingly from the text around it, not from the pixels. Its vision models are good and getting better, but on the open web it still leans hard on:

• the alt text,
• the filename,
• the surrounding page content and the nearest heading,
• the page’s overall topic,
• structured data,
• and the caption, if there is one.

So “ranking in Google Images” is mostly “giving Google enough text to confidently understand what the image is and why it matters”. Everything below is a version of that.

The things that move rankings

1. Alt text that describes the image’s function

The single strongest on-image signal. Not keyword-stuffed — described. The same photo needs different alt text depending on the page it is on, and Google is good at spotting a page whose images are all labelled with the same commercial keyword. Write for the person who cannot see the image and you will satisfy Google as a side effect. Full detail in how to write good alt text.

2. The page has to be about the image

An image ranks because the page ranks for the query. A photo of a sourdough loaf on a page about tax law will not rank in image search for “sourdough”, no matter how good the alt text is. The image and the page topic have to agree. Put the image near the text that explains it.

3. Descriptive filenames

cobalt-running-shoe.jpg beats IMG_4032.jpg. Small signal, zero cost at export time, so do it — but do not agonise, and do not stuff (buy-cheap-shoes-online.jpg helps nobody).

4. Speed and Core Web Vitals

Image search results favour fast, mobile-friendly pages. An oversized image hurts you twice: it tanks your Largest Contentful Paint (a ranking signal) and it is the image itself that is slow to appear. Serve WebP or AVIF, at the size you actually display, with width and height set so the page does not jump.

5. Structured data

ImageObject, and more importantly the image fields inside Product, Recipe and Article schema. This is how you become eligible for rich results and badges in image search — the things that make your result the one people click.

6. An image sitemap, for image-heavy sites

For a normal site your regular sitemap is enough. For a stock library, a recipe site, a large catalogue — anywhere images are the product — image sitemap entries help Google discover images it would otherwise miss, especially ones loaded by JavaScript.

Advice to ignore

• Keyword-stuffed alt text. Dead for a decade. Actively harmful now.
• Image title attributes as a ranking signal. They are not one.
• “More images ranks better.” Only if they are useful. Padding a page with stock photos just makes it slower.
• Exact-match keyword filenames. describe the image; that is enough.

The bit that makes it hard

Every item above is per-image work. On a ten-page blog that is an afternoon. On a ten-thousand-image catalogue it is the reason none of it ever happens — and a catalogue with no alt text is invisible to image search entirely.

So the realistic sequence for a large site:

1. Find the images with no alt text. Any crawler lists them. This is your denominator, and it is usually bigger than you feared.
2. Fix the templates first. A handful of templates — the product card, the article header — generate most of the gaps. One template fix can cover thousands of images.
3. Draft the rest with AI, then edit. Run your image URLs through a pass that writes a factual description for every one. PicsTag does this in the browser and hands you a CSV to join back onto your CMS. The drafts are not perfect; an empty alt attribute is worse, and Google treats it as “no information”.
4. Then do filenames, structured data and the sitemap.

In that order, because item 1 is worth more than items 3 through 6 put together, and most people do it last — if at all.

How to know it worked

Do not measure alt-text coverage. Open Google Search Console, filter Performance to the Images search type, and watch impressions and clicks before and after. That is the only number that tells you whether Google understood your images and started showing them. Coverage tells you that you filled in fields; Search Console tells you it mattered.