For a long time, image optimization was one of those things I understood only partially.
I knew images should be compressed. I knew WordPress had an Alt Text field. I knew descriptive filenames were supposed to be better than names like image-01.webp.
But I didn’t have a clear system.
Sometimes I wrote alt text that was basically the blog post title. Sometimes I described the image. Sometimes I left it empty. My filenames were inconsistent. And if an image was the featured image of an article, I assumed writing something like “Featured image for…” might somehow help Google understand it.
The more I looked into it, the clearer it became that image SEO is not about filling every available field with keywords.
It is about giving search engines, browsers, assistive technologies, and increasingly machine-based discovery systems enough clean context to understand what an image represents, why it exists on the page, and how it relates to the content around it.
That changed how I handle images now.
Alt Text Is Not an SEO Description Field
This was the most important thing for me to understand.
Alt text exists primarily as a text alternative to an image.
If someone cannot see the image, the alt text should communicate the useful information the image would otherwise provide. W3C’s accessibility guidance makes the same distinction: informative images need meaningful alternatives, while purely decorative images should generally use an empty alt attribute, alt="". (W3C)
Google also uses alt text.
Google says it combines alt text with page content and computer vision to understand an image’s subject matter. It specifically recommends useful, information-rich alt text in the context of the page and warns against keyword stuffing. (Google for Developers)
So accessibility and SEO are not competing goals here.
Good alt text normally satisfies both because it accurately describes the image.
What changed for me was the question I ask before writing it.
I no longer ask:
What keywords should I put in this alt text?
I ask:
If this image disappeared, what useful information would the reader lose?
That usually produces much better alt text.
Should a Featured Image Say “Featured Image” in the Alt Text?
Usually, no.
Suppose my article is called:
Interesting Facts About WordPress
And the featured image contains the same title along with an illustration of a website interface, analytics, settings, and idea-related icons.
I could write:
Featured image for the blog post Interesting Facts About WordPress
But that does not really describe the image.
“Featured image” is a WordPress publishing role. It isn’t useful information about what the visual communicates.
A better descriptive alt could be:
WordPress facts illustrated with a website dashboard, analytics, settings, and idea icons
That tells someone what the image actually contains while naturally establishing its context.
But there is another possibility that I didn’t properly understand before.
Sometimes the correct alt text is nothing at all.
Sometimes alt="" Is Better Than an SEO-Friendly Alt
Imagine the featured image appears immediately below this:
H1: Interesting Facts About WordPress
The image itself also contains the words:
Interesting Facts About WordPress
And the illustration doesn’t communicate anything necessary for understanding the article.
A screen reader user does not necessarily benefit from hearing:
“Interesting Facts About WordPress. Image: Interesting Facts About WordPress.”
In a case like that, the image can reasonably be considered decorative or redundant, and alt="" may provide a better accessibility experience. W3C specifically recommends empty alt text when an image adds no information beyond adjacent content. (W3C)
This is why I no longer think:
Every image needs keyword-rich alt text.
The better rule is:
Every image needs the correct alt treatment.
An informative image gets useful alt text.
A functional image gets alt text explaining its function.
A decorative or completely redundant image can get alt="".
And leaving the alt attribute out entirely is not the same as intentionally using an empty alt attribute. Some assistive technologies may announce the filename when alt is missing. (W3C)
Alt Text Depends on Context
Another mistake is assuming an image has one universally correct alt text.
It doesn’t.
Imagine I have a screenshot showing the WordPress Media Library.
In an article about WordPress image optimization, the important information might be:
WordPress Media Library showing image alt text, title, caption, and description fields
If the exact same screenshot appears in an article explaining WordPress administration generally, a shorter description might make more sense.
The pixels did not change.
The purpose did.
That is why alt text should describe the image in the context in which it appears.
This is particularly worth remembering in WordPress because it is easy to start thinking of alt text as simply another Media Library metadata field instead of part of the page’s content and accessibility.
How Long Should Alt Text Be?
I don’t follow a fixed character limit.
Neither accessibility nor SEO improves because the alt text happens to contain exactly 100 or 125 characters.
W3C recommends making it as concise as possible while still communicating the image’s purpose. If an image requires a long explanation—such as a chart, technical diagram, or infographic—the important details should generally be explained in the surrounding content rather than forcing everything into a huge alt attribute. (W3C)
So this:
WordPress dashboard
may be too vague.
This:
WordPress dashboard showing declining page-load time after image optimization
may be useful.
But this:
WordPress dashboard WordPress image optimization SEO WordPress speed optimization best WordPress images fast WordPress website
is not optimized.
It is keyword stuffing.
Google explicitly advises against that. (Google for Developers)
I Also Stopped Writing “Image of…”
Most alt text doesn’t need to begin with:
“Image of…”
“Picture of…”
“Graphic showing…”
A screen reader generally already communicates that the element is an image.
W3C similarly recommends avoiding generic words such as “image,” “icon,” or “picture” unless the distinction itself is meaningful. (W3C)
Instead of:
Image of a WordPress dashboard showing image optimization settings
I would normally use:
WordPress dashboard showing image optimization settings
There are exceptions. Saying “photograph,” “illustration,” “diagram,” or “screenshot” can be relevant when the medium matters.
Context decides.
Filenames Matter, but Much Less Than I Used to Think
This was another useful correction.
Google says the filename can provide very light clues about an image’s subject matter. It recommends short, descriptive filenames instead of generic names such as IMG00023.JPG. (Google for Developers)
So I now rename files before uploading them.
Instead of:
Screenshot-2026-09-14-at-3.21.18-PM.webp
I would use:
wordpress-image-seo-alt-text.webp
And instead of:
Interesting Facts About WordPress.webp
I prefer:
interesting-facts-about-wordpress.webp
Google’s broader URL guidance recommends hyphens rather than underscores to separate words because they make concepts easier for both people and search engines to identify. (Google for Developers)
That means I prefer:
image-seo-guide.webp
over:
image_seo_guide.webp
and definitely over:
imageseoguide.webp
But I don’t treat filenames like some hidden ranking hack.
Google itself characterizes their contribution as a relatively light signal. (Google for Developers)
The filename is one piece of context, not the strategy.
What Matters More Than the Filename
The surrounding page matters significantly more.
Google says it extracts information about an image from the page around it, including nearby text, captions, page content, titles, and image metadata. It recommends placing images near text relevant to their subject. (Google for Developers)
This made image SEO feel much more logical to me.
If I publish an image about WordPress image optimization inside a detailed article about WordPress image optimization, under a relevant heading and beside explanatory content, Google already has strong contextual information.
The filename and alt text reinforce that understanding.
They should not be trying to manufacture relevance that the page itself does not have.
Image SEO Is Also Performance SEO
There is another side of image optimization that is easier to overlook when discussing alt text: the actual file.
A perfectly written alt attribute does not compensate for a 4 MB hero image slowing down the article.
Google’s image guidance recommends balancing high-quality imagery with image optimization and responsive delivery because images are often among the largest contributors to page weight. Google currently supports formats including JPEG, PNG, WebP, SVG, and AVIF in Search. (Google for Developers)
For most blog graphics, I now think about WebP or AVIF before uploading large PNGs.
I also care about dimensions.
There is little point serving a 4000-pixel-wide graphic if the content container displays it at around 1200 pixels.
Responsive images matter too. WordPress normally generates multiple image sizes and can output srcset, allowing browsers to choose an appropriate version for the device.
That saves bandwidth without requiring me to create every variation manually.
Featured Images Have Another Job: Search and Discover Previews
This is where things became particularly relevant to the featured images I create.
Google doesn’t only use images inside Google Images.
Images can appear in normal search results, article results, Google Discover, and other search surfaces.
Google recently clarified that it can use sources such as schema.org image properties and og:image when selecting a preferred image for Search and Discover. (Google for Developers)
For articles, structured data can specify the representative image through the image property of BlogPosting or Article. Google also supports primaryImageOfPage for identifying a page’s preferred image. (Google for Developers)
That means my featured image workflow now extends beyond the WordPress Featured Image box.
I also want to know that my SEO or social plugin is producing sensible:
og:image
Article structured data
and an appropriate representative image URL.
My 16:9 Featured Images Already Have One Advantage
Google’s current Discover documentation recommends compelling, high-quality images that are at least 1200 pixels wide and contain more than 300,000 pixels in total. It specifically discusses 16:9 imagery as suitable for landscape presentation and recommends allowing large previews through max-image-preview:large. (Google for Developers)
So a featured image at:
1536 × 864
already clears those size requirements comfortably.
But there is an important catch.
Text-Heavy Featured Images Are Not Always the Best Search Preview
This is something I am paying more attention to now.
I like editorial featured images with strong titles. They work well on blog archive pages, social posts, and branded content layouts.
But Google recommends avoiding text-heavy images when selecting preferred images for Discover and similar previews. (Google for Developers)
That creates an interesting distinction.
The best blog banner and the best search-preview image are not automatically the same asset.
For some articles, I may still use my branded title-based featured image on the site, but create a cleaner visual version—with less embedded text—as the preferred image for Open Graph or structured data.
That lets the editorial design do its job without assuming that every distribution channel wants exactly the same creative.
What About Image Titles and Captions?
I used to see WordPress fields such as Title, Caption, Description, and Alt Text and assume they were all SEO fields that needed to be completed.
They are not interchangeable.
Alt text is the text alternative for the image.
A caption is visible page content associated with the image. If a caption genuinely helps readers understand what they are looking at, it can be useful context, and Google can also use surrounding text and captions to better understand imagery. (Google for Developers)
A media title can help with organization, but I don’t fill image fields just because WordPress exposes them.
If a caption adds nothing, I don’t add one simply for SEO.
More metadata is not automatically better metadata.
What About AI Search and GEO?
This is where I think people can easily start inventing new rules.
There is no special “AI alt text” field.
I don’t write one version for Google, another for ChatGPT-style systems, and another for GEO.
What I can do is make the page easier for machines to interpret:
clear page structure, descriptive headings, meaningful surrounding text, relevant images, accurate alt text, understandable filenames, crawlable image URLs, structured data where appropriate, and consistent metadata.
That is good SEO already.
It is also the kind of semantic clarity that gives AI-powered discovery systems better information to work with.
Google itself now integrates AI-oriented search experiences alongside traditional Search surfaces, but that does not suddenly make keyword-stuffed image metadata a good strategy. (Google for Developers)
So when I say I am optimizing images for SEO, GEO, and AI search, what I really mean is that I am making their meaning explicit and machine-readable without compromising the human experience.
I think that is a much safer approach than trying to guess hidden prompts or future ranking factors.
AI-Generated Images Add Another Consideration
Since AI-generated and AI-assisted visuals are becoming normal in content workflows, image provenance is becoming more relevant too.
Google can read certain image metadata, including C2PA information, and may surface information about how an image was created or edited through features such as “About this image.” Google also recognizes IPTC metadata related to creators, copyright, licensing, and digital source type. (Google for Developers)
This does not mean I need to stuff every generated graphic with metadata.
It means I am becoming more careful about automatically stripping all metadata without knowing what is being removed, particularly when creator, rights, or provenance information matters.
The Workflow I Use Now
My process is much simpler now because every field has a reason behind it:
- Create an image that actually supports the article, rather than adding a generic stock-style visual.
- Export it at an appropriate resolution, usually at least 1200px wide for important article imagery and featured images.
- Use an efficient format such as WebP or AVIF where appropriate.
- Rename the file before uploading it, using a short descriptive lowercase filename such as
wordpress-image-seo-alt-text.webp. - Use hyphens between words, rather than underscores or concatenated filenames.
- Decide whether the image is informative or decorative before writing alt text.
- Write concise, contextual alt text when the image communicates useful information.
- Use
alt=""when the image is purely decorative or completely redundant with nearby content. - Avoid “image of,” “featured image for,” keyword stuffing, and repeating the entire article title unnecessarily.
- Place images near relevant content so the page itself establishes context.
- Check responsive sizing and compression rather than uploading oversized originals everywhere.
- Verify the featured image is represented correctly through
og:imageand article/schema metadata. - Use a high-resolution representative image for Search and Discover, while considering whether a less text-heavy version would work better as the preferred preview.
- Check the final page rather than only the Media Library, because context determines whether the implementation actually makes sense.
That is essentially my current image SEO system.
The Biggest Change Was Understanding the Purpose
I originally thought image SEO meant filling fields.
Filename: add keyword.
Alt text: add keyword.
Title: add keyword.
Description: maybe add the keyword again.
That approach is easy, but it misses the point.
I now think about image optimization as a combination of accessibility, context, performance, discoverability, and metadata.
A good filename gives a small clue.
Good alt text provides an appropriate text alternative and gives search systems useful context.
Nearby content establishes relevance.
Compression and responsive delivery protect page performance.
Structured data and Open Graph metadata help identify the image that represents the article.
And all of those signals work together.
I am also much more comfortable leaving a field empty when there is a valid reason to leave it empty.
That may be the biggest difference.
I’m no longer trying to make every image look “SEO optimized.”
I’m trying to make every image correctly implemented.
And that gives me much more confidence that the same image can work properly for readers, accessibility tools, traditional search, image search, Discover, and increasingly AI-driven discovery as well.
