Advanced image SEO: A secret manual
Learn how search engines process images and key image SEO tactics to implement from this comprehensive overview.
Image SEO is an intriguing niche because of the immediate impression images make when viewed.
Read on and I will share a comprehensive overview of my image SEO “secret manual.”
In this article:
- The impact of images.
- Images in search: How they are processed and analyzed.
- Image SEO: Ranking factors and influential elements.
The impact of images
Images are important because they communicate a wealth of information quickly, just with the initial view.
Logos are one example. Effective logos convey a unique identifier specific to one company.
They can include a mix of words, letters, and graphic iconography that references the company’s products, services, industry category, or even corporate ethos.
Other examples of highly communicative images include:
- Splash banners that are the center point of some websites’ homepages.
- Product images in e-catalogs.
- Avatars that represent one in social media.
- Portrait photos.
- Graphs of statistics.
- And the once-trendy infographic.
Image search has long been one of the top-used search vertical types. Consumers commonly search for images when they:
- Are shopping for products where a visual aesthetic may be important.
- Want to see what a subject looks like.
- Are comparing products.
- Are looking to understand how to identify subjects.
- Need to comprehend geographic or statistical information.
Because of consumers’ high interest in images, search engines have long worked on sophisticated methods to associate images with searchable keywords.
They have woven images into the keyword search results in what is frequently called “federated or “blended search.” Google branded this as their “universal search.”
Get the daily newsletter search marketers rely on.
Images in search: How they are processed and analyzed
Images show up on search in various areas and SERP features, such as:
- Regular keyword search results, embedded images subsection (via Universal Search).
- Thumbnails embedded with rich snippets.
- Included with featured snippets.
- Image search vertical results.
- News search results.
- Video search results.
It is informative to know some of the methods which Google and other search engines have used in surfacing images for users' keyword searches.
Humans view images and rapidly process them to understand the contents based on years of experience in the natural world and a highly-evolved biological image processing system.
Computing systems, on the other hand, do not have all of this sophisticated image processing capability.
- A major dependency for image search algorithms is the association of words with image content. This is accomplished using various metadata associated with the image to infer the content.
Search engines use:
- File names.
- Text near the images on webpages.
- Alt text within HTML code for the image.
- Links pointing to the image.
- And more.
But this sort of metadata is often undesirably sparse or less trustworthy.
- Filenames can be gobbledygook database IDs.
- Webmasters frequently leave out captions and alt text.
- Text surrounding images within articles is a somewhat dirty information source.
Because of this, search engines have worked on algorithms to analyze shapes within images, comparing the content of images sparse in keyword metadata with images containing similar content but having richer text data.
Some of this algorithmic analysis includes identifying text within images (i.e., a sign appearing within an image) or the text added by graphic designers to images.
This optical character recognition (OCR) technology has been around for perhaps over a century but became more sophisticated when scanned documents could be converted back into text documents from the 1970s and 1980s.
It is not clear when OCR first became incorporated into image analysis algorithms used by search engines, although it is likely this was sometime around 2005 when this started.
In 2005, Google helped make Tesseract an open-source OCR project. They also hired one of Tesseract's primary inventors, Ray Smith, who developed the technology originally at Hewlett-Packard. (Smith now works within the DeepMind research division of Alphabet, which is an AI think tank.)
In 2006, Google Labs famously launched its Image Labeler program, a means of crowd-sourcing identification of images by having people submit keywords to describe images that Google showed them.
Image Labeler gamified the process by having two separate users shown the same images, and the users would get points by matching the same words for images. (Ten years later or so, Google relaunched this project in a different form as part of its Crowdsource project.)
It is believed that Image Labeler generated a huge corpus of data about image contents which was later incorporated in helping to train AI-like algorithms to recognize identifiable shapes within images, further enhancing Google image search.
Google's image analysis algorithms have progressed further using AI-like neural networks trained on data from images associated with appropriate keywords.
One of Google's systems, Inception, is described as a novel deep convolutional neural network architecture, applied to help identify shapes, including 3D shapes. (Its name was inspired by a quotation, "we need to go deeper," from the 2010 science fiction film, "Inception.")
It is unclear if Google is using a variation of Inception with its image search vertical, or if it is using something similar. That said, it is known they are using neural network analysis in combination with other algorithmic methods to identify image content.
(One obvious application for image identification would be facial recognition technology. Google has not seemed to introduce this, perhaps because of the obvious privacy implications that can multiply through automated facial recognition. However, colleague Bill Hartzer has described how a client of his had apparently been outed through facial recognition in Bing search, strongly indicating this functionality is now in use in Microsoft's Bing search engine. Microsoft has played with this in the past with a novelty app, and Jeff Sauer has previously theorized that the functionality is also used in Bing, conjecturing it was introduced through its partnership with Facebook. But, Facebook announced they discontinued facial recognition in 2021.)
The point of mentioning some of these technologies is to make you aware of how search engines can use multiple methods to understand the content of images to associate the images with keyword search queries.
Search engines have done all this sophisticated work largely because webmasters have so poorly optimized image content that a huge proportion of it would be completely hidden from searchers if search engines did not make an effort to expose it better.
And, while they have done an amazing job at this, there remains a much higher confidence level about keywords associated with images if developers will optimize their images, generating much higher confidence for search engines in determining what the content of the images may be.
Even though Google has developed methods for associating keywords with images, you cannot easily tell if they associated the right or ideal keywords with your images. Thus, you still need to optimize images on your website if they are important to you.
Below, I will describe the various ranking factors for images.
Image SEO: Ranking factors and influential elements
Optimizing the image itself
The first category of image optimization factors is not intuitive – it begins with the creation or formatting of the image itself.
When creating an image, shooting a photograph, or sourcing stock art for your image, you should first focus on originality.
Google image search results, like Google keyword search results, abhors duplicate content to some degree.
You will rarely to never see five copies of the identical image appearing at the top of the search results because Google has determined that this is a bad user experience.
Thus, if you are using an image on your webpage that has already been used elsewhere, you need to use a different one or change the image to make it significantly different enough such that Google may not suppress it by some degree as being a form of duplicate content.
If you have a product data feed from a vendor, consider shooting your own photos. Or, if the product is on white background, have a graphic artist compose them onto a different one.
Recropping the image and altering its color hues may also help – or, even flipping the photo to use a mirror-image version.
If your photo is identical to other high-ranking photos appearing for the same keyword, chances are you may not when Google filters for duplicates.
It has become increasingly easy for image license holders to detect unauthorized use of their images. This goes without saying, but avoid publishing images on your websites without the proper license to do so.
Google has also begun reducing rankings for websites with large quantities of DMCA copyright removal issues, so infringing exposes you to penalties and infringement lawsuits.
While AI-generated art does not yet appear to be discriminated against in image search, I suspect it may do so rapidly if websites begin over-using AI tools to create images for articles, especially if the generated image was constructed to deceive in some way.
There is some likelihood that search engines may already be working to detect AI-created art when displaying photos of people for individuals' name searches.
Beware of depending upon AI-generated images as this could become something suppressed or flagged in the near future.
Similar to copyright-infringing content being a risk to image rankings, AI-generated is also under the presumption that such images may be composed of others' content.
Size and resolution
Google image search allows one to filter the search results by image size (any size, large, medium and icon), but there is also reason to believe that if images are too small to have good resolution, Google would be less likely to present them prominently.
Users desire to see images that are clear and focused as opposed to blurry. While your image should ideally be sized perfectly for how it will appear on a webpage, make sure it will be large enough to contain enough pixels to depict the subject.
Images may also be provided responsively according to the viewing device screen size, so providing alternate sizes may increasingly be preferred by Google. (See information on this in the upcoming section on webpage optimization for images.)
Saturation and contrast
In various contexts, your image may appear as a small thumbnail, such as in search results along with other images, or as a thumbnail in the SERP beside the webpage's listing. In such instances, you are often competing with a number of other images.
One way for your image to stand out better, increasing the likelihood of attracting clicks, is if the image is a little higher contrast and higher color saturation when presented in a thumbnail version.
Assess the overall image – can it be brightened and slightly increased in saturation or contrast, while still appearing natural? Tweaks like this can have a long-term beneficial effect on performance.
Filesize and file type
Because page speed is a ranking factor, if your image's file size is too unnecessarily large, it might ding the ranking of the page where it is located.
As some degree of an image's ranking ability must be derived from the webpage's ranking ability, then harming the page's search performance is also contraindicated for image search rankings.
Use a file type that helps compress the image without losing too much resolution. Google encourages people to use WebP and AVIF, which would be great options.
But, if your images are used by your site visitors for some purpose, you may not want to use them because plenty of other software and systems cannot process them yet. (For example, WebP cannot be used in a PowerPoint slideshow.)
An old rule of thumb is that GIF is used for line drawings more and JPG for photo images. There is no hard-and-fast rule for this, however.
GIFs also can be photo images (or even animated) and JPGs can be line drawings. PNGs are a great format for any type of content.
Both GIFs and PNGs can have transparent areas. But for some types of images, such as image thumbnails appearing in keyword SERP listings, Google probably does not want transparent images or animated GIFs.
Metadata within image files (EXIF and IPTC)
In SEO circles, we are used to talking about meta tags in HTML.
The meta description tag is a brief statement describing what a page is about, and can appear as the text snippet below your pages' links in SERPs.
The meta keyword tag was used years ago to convey keywords about a page to search engines, but major search engines do not use that anymore.
But broadly speaking, metadata is data that describes data – and goes well beyond the description and keyword tags. (For that matter, Facebook's Open Graph tags and Twitter Card tags are also metadata about a page, as well as schema and some other structured data.)
But, images can frequently have their own metadata tacked onto the image file itself, and many are unaware of this because the data is largely invisible to people.
EXIF data contains primarily technical info about the image – ranging from date/time-stamps for when the pic was taken, camera make and model, aperture, lens, focal length, color space, geolocation coordinates of the pic, description, and more.
IPTC data can include a headline, description, category/genre, keywords, fields about persons (models) appearing in the image, location info (city, country, province/state, location of creation of the image, location shown in the image), copyright info, credit line, info about the image creator, the copyright owner, image license terms, and more.
IPTC data is likely the industry-leading standard, although many legacy content and older systems use EXIF.
What image metadata is used by Google? Years ago, it seemed likely Google might use as much as possible, but a lot of technical image specs like aperture would only be of interest to professional photographers and image creators, making this of very limited usefulness to most users.
Years back, I urged people to make use of EXIF, because even if Google did not directly make use of it, an argument was to be had that some image-sharing websites will publish the EXIF contents on the webpages, generating keyword data that could help with rankings.
Further, I expected geolocation data embedded in EXIFs might optimize more for local websites.
But, there is no indication that this has been used. (Only some cameras with geolocation capability can embed coordinates into EXIF.) There is software or web services out there that will allow you to custom-engineer your EXIF data.
Google is only using portions of IPTC data, such as:
- The image's creator.
- Credit line.
- Licensing/copyright notice.
You can use image metadata via IPTC protocol or structured data, which Google allows. Note that it would be nonoptimal to use both IPTC and structured data with conflicting field values.
There is still an argument that the other EXIF and IPTC data could be indirectly beneficial to a degree if you are sharing the images on social media sites similar to Flickr where the fields will get published on webpages in conjunction with the image itself.
But, this will be at best a marginal effect where keywords and rankings would be concerned. Besides conveying creator and licensing options, image metadata is unlikely to provide your website with all that much benefit.
Remember that you could reveal some private or sensitive data with your hidden metadata in some cases, too, so be aware of its existence.
Image filenames can be highly influential when an appropriate keyword is used to compose them, and if they are composed well for search engines. This is similar to how one may use keyworded URLs for webpages for SEO benefit.
Obviously, a single-word keyword can easily make a file name, such as "pangolin.jpg." If the desired keyword you are optimizing for is a phrase with multiple words, the words need to be delimited by a whitespace character, such as a dash ("-"), a period ("."), a tilde ("~"), or an HTML escape code for a space ("%20").
Despite it being a reasonable-seeming choice, do not use underscores ("_") for your spaces, as this is treated differently from spaces and will reduce exact-match evaluation when compared with search queries.
If your content management system demands an ID number for images, it is possible to design the CMS to allow you to put the ID at the end of the file name after your custom keyword, and before the filetype designator (e.g., "green-pangolin-ABC123.jpg").
Image URL path
Google states that they also use the URL path to understand images.
So if you have large amounts of image content, storing them in categorized and subcategorized directories on your server may provide some marginal keyword relevancy.
This is in addition to the much stronger relevancy of a nicely keyworded filename (e.g., "example.com/asia/exotic-animals/pangolin-in-nature.jpg").
Avoid restricted content
This may seem obvious, but images that could get confused as pornographic or hate symbols could get suppressed.
If you host user-generated content, you should already have some policing in place to block or remove restricted content. Unless you are a website already categorized as being involved in a sensitive topic, that is.
Incorporating images into HTML pages
This second category of image optimization factors comprises tactics that have been most focused upon historically by search marketers. However, there are some modern twists worth paying close attention to.
Following these recommendations will often pay off in terms of superior keyword rankings because so many sites are very lackadaisical about optimizing the images. Most websites merely focus on how webpages and websites look, without making certain they construct the site technically well.
Avoid CSS image delivery
Your page's main content images should not be delivered as a background image for
<span> or other elements. Google largely will ignore CSS images.
<img> alt text
Image alt text, short for "alternative text," is included within the
<img> tag like the following:
<img src="somewhere/green-pangolin.jpg" alt="Green Pangolins sauntering around their enclosure in the city zoo.">
The alt text should be succinct and communicate the image's content clearly.
The alt text is a fundamental element used by search engines for understanding images. It is used by visually impaired persons as their audio browsers may read the alt text out loud to enable them to understand the page's contents.
Optimize the <img> tag
<img> tag functions great for SEO benefit.
Building off the previous advice, construct your
<img> tag with a good src file name and alt parameter text.
Set the height and width. Ideally, it should match the actual image size when displayed on the page. Resizing purely through the height/width parameters can cause it to display less smoothly.
In the earlier evolution of HTML, one could use the lowsrc attribute to provide a URL for a small, lower-resolution image file that could load rapidly, followed by a slower-loading, higher-resolution image. But, this parameter was deprecated in HTML 5.
<img> file works beautifully for optimizing images for search. But if your marketing tactics involve providing larger or full-screen images to attract visitors, you may need to link the image to a higher-resolution or larger version of itself.
However, Google recommends using an image srcset attribute to pass multiple size files for the same image to serve responsive pages for different screen sizes.
Their recent image SEO update clarifies that Google indexes the
<img> elements, including when they are enclosed in other elements, such as
Use the srcset attribute on <img> tags
Remember how page speed and mobile-friendliness are now important for SEO? Because of this, providing responsive images is now optimal.
The srcset is optimal for defining a group of different size versions of the same image and conveying the sizes of each.
The inclusion of the sizes specification attribute instructs browsers as to which image to display according to browser window screen size.
The viewing device browser can then determine which size will be displayed to the end user.
The src attribute specifies the URL of the default image file, while the srcset provides the URL/width pairs for each responsive image. For example:
srcset="green-pangolin-250.jpg 250w, green-pangolin-500.jpg 500w,
sizes="(max-width: 1082px) 250px, (max-width: 1282px) 500px, 750px"
alt="Green Pangolins sauntering around their enclosure in the city zoo.">
Avoid redirecting images
Similar to webpages, Google can and does handle the redirection of images. But, ideally, you should not redirect your images – use the final URL of an image when including it on your webpage.
Don't block Googlebot from your images directory with robots.txt
This should be obvious, but it happens.
Use image sitemaps to disclose your images to Googlebot
If you have a straightforward site construction and it is easy for Google to index all your webpages where your images are located, using image sitemaps may only be a backup for indexing your images.
But, if you have a large collection of images or the pages where they are found only display them via lazy loading and infinite scrolling, then image sitemaps may be a vital necessity.
Use structured data
Google's image SEO best practices cite product, video (for video thumbnail images) and recipe types of structured data. But I would also add to that list logo (for organizations, local business, how-to, and article (Article, NewsArticle, and BlogPosting) types.
Note that logo optimization is very useful in helping to optimize an entire website. Thus, it is worthwhile even if you do not anticipate users specifically searching for your logo.
But, for well-known brands, there will be many reasonable, fair-use applications for people searching for the logo. You may want to provide multiple sizes, including hi-resolution vector art, on a dedicated press kit page.
As an added criterion, Google states that your logo used with the structured markup should be designed to look good on a white background, so consider carefully if your logo is a transparent PNG or GIF.
As we mention logos, also be aware that Google increasingly uses your site's Favicon and listings in search results, particularly on mobile.
Follow Google's Favicon documentation to ensure optimal inclusion. (Note that Google uses a special Favicon crawler just for these little images! So, do not block it.)
Add Facebook's Open Graph (OG) and Twitter Cards metadata to the page
In the past, Google has used data as a fallback from these tags when other metadata and structured data are not present, which can impact search optimization.
But, the main advantage of using this is that they help to provide preview snippets when the page is shared on social media.
The images used in OG and Twitter Cards can appear in those snippets, increasing user clickthroughs from both Facebook and Twitter.
Promoting and distributing images
Image SEO is often focused primarily on optimizing for on-page and on-site factors, assuming one will be hosting the images on one's native website.
However, for those sites where images are a major marketing component, hosting and promoting in multiple ways can be more advantageous than only hosting on the native website itself.
If you believe you may get website referral traffic by people searching for images and then clicking through to your website, then images may also be advantageously hosted on other channels in addition to your native site.
This is an area where the lines can blur a bit between organic search and social media marketing.
The obvious question when promoting on social media is – will links from social posts benefit my native website's content?
This is a fairly complex topic at this point, but you should understand that while most of the links from social media platforms will be using the rel="nofollow" link parameter, they still may be conveying some link benefit that can translate into organic ranking benefits.
Nofollow generally halts the transfer of link benefit, but Google stated in 2019 that they may treat it as a "hint," signaling that under some circumstances they will indeed count the link weight conveyed.
Now, Google's system analyzes social media platforms' link graphs and popularity/engagement signals on social platforms also coordinate closely with links in those systems.
So, it seems pretty likely that an account's popularity measured through its links and the popularity of a particular post (including its images) would influence whether it might pass any link signal.
And, accounts and posts with low popularity/engagement likely do not pass any link signal. Thus, you will need to achieve popularity or significant engagement for a post to influence search rankings.
Should I also mention that just in December 2022, Google's John Mueller delurked to add a comment on Reddit when a user posted the question, "Should we rel="nofollow" social media?"
Mueller answered, "Links to your own profile? Absolutely not. [omg, this will certainly be a fun ride]".
Now, this was likely just focused upon how nofollowing links to your own accounts flies against the web standard because one does not disallow links to one's own accounts.
But, it also does suggest that interlinking among one's own social media accounts while using nofollow can challenge Google's ability to use the links to determine relationships between the accounts, something that could seriously affect the algorithmic analysis for their Knowledge Graph.
Automatically nofollowed social media accounts sometimes happen, such as through ignorantly implemented default settings in WordPress themes.
This is something to be aware of and avoid. But, the obvious implication is also that this could have an effect on organic search rankings.
Here are just a few ideas for promoting your images.
Post images on Facebook and Twitter
You can use the page URL where the images are located and have the images appear as previews on Facebook and Twitter.
But if your primary content is the images themselves, posting them directly on Facebook and Twitter and providing good description text with the posts (and alt text when it is an option) will give additional opportunities for the images to appear in image search results.
The posts could then include links back to the original content pages on your website. Obviously, if your main feature is an article page, then you will want to initially promote with just a link to the page and rely upon the preview image.
But, it is often the case that we may optimize on social media by posting about the same item a number of times, so the second or third time you post you could just upload an image and add the link to the page.
Further, if a page has multiple images, it may be advantageous to post directly on Facebook or Twitter so that you can add the full gallery of images, increasing potential engagement on the social media platform while linking back to your content.
Option: Make slightly different versions of images when posting simultaneously across social media platforms and the native website
Earlier in this article, I covered how Google can and often does suppress duplicate content so that the same image does not appear multiple times for a search.
This is because it is considered a poorer quality experience to see the same picture for a search. See the above recommendations regarding making alternate versions instead of exact duplicates.
Pinterest can be great for distributing images, in part because each Pinterest "Pin" page can be linked back to the page where the image is found on your website.
This website seems increasingly like a vintage social media platform, but it continues to enjoy some great benefits in Google search and has a decent user base.
It is also great for posting images with associated links to coordinating pages on your website, like Pinterest.
In addition to this, Flickr publishes some EXIF data automatically and allows you to:
- Post pics on maps to associate images with locations.
- Tag keywords.
- Host multiple image sizes.
- Have multiple image licensing label options.
- Group photos into one's own or public albums based on themes – multiple ones – which are constructed generally well.
Instagram is one of the prime social media platforms for image sharing. It can show up very well in image search results.
However, it is extremely limited in conveying any theoretical link benefit because links cannot be added to each individual image post.
You can use the post description to tell users how to find your media, such as telling them how to find the website in your Instagram profile URL, or what words to search for in Google.
But, Instagram pics are mainly used for optimizing images on their own. Be sure to use the description text with the image and include multiple hashtags related to its content to increase its visibility to others interested in the topics.
Image search optimization is worth the effort
Many social media platforms provide image-sharing services that can convey some benefits. Experiment and see what works well for your business and website.
Image search optimization can be well worth the effort involved, so pay attention to reap the benefits.
Once you have your processes and technology configured for image SEO, it will become less labor-intensive and will seem like a walk in the park.
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.