Before diving into the subject and providing advanced guidance for image optimisation, it seems important to give a brief introduction to the image ranking and recognition algorithms used by search engines.
How image ranking and recognition works
Images can be displayed in several search contexts on Google’s results pages, such as:
- Regular keyword-based search results, with an embedded image section (via universal search).
- Embedded image thumbnails with rich snippets.
- Images included with featured snippets.
- Vertical search results from image search.
- News search results.
- Video search results.
It is interesting to understand how Google and other search engines find images in response to user queries.
Unlike humans, computer systems do not have the same sophisticated image processing capabilities. Image search algorithms rely primarily on metadata associated with images, such as file names, captions, surrounding text and alt text. However, this metadata is not always available or reliable, as it is often omitted by webmasters.
Consequently, search engines have developed algorithms to analyse shapes within images. They have also used optical character recognition (OCR) technology to identify text in images.
In 2005, Google helped make Tesseract an open-source OCR project, and in 2006 they launched Image Labeler to help identify images through crowdsourcing.
These data were incorporated to help train AI algorithms to improve image search.
Google’s image analysis algorithms were subsequently improved using AI-type neural networks to identify shapes in images.
A good example of AI for image search is Inception — a convolutional neural network (CNN) model used for image classification.
Convolutional neural networks are a subclass of artificial neural networks particularly suited to image analysis. They are capable of learning to detect patterns and structures in images. Inception was designed to be highly efficient and accurate for image classification, and it won several major computer vision competitions.
However, it is important to note that Inception is just one neural network model among many that can be used for image analysis.
If you wish to continue learning directly about image optimisation from an algorithmic standpoint, you can read my article on how Google ranks images in search. If you would like a quick actionable guide with lesser-known optimisations, read on.
Image SEO: Advanced ranking factors
Note that this guide does not cover the classic image SEO optimisations. It only covers additional suggestions for ranking your images. The optimisations that follow are not the first ones to consider when improving your image SEO.
Image PageRank
To optimise your images for SEO, it is important to consider the links pointing to them, in the same way as you would optimise a page. If you want your image to rank well in search results, it is therefore important to have it appear on other websites by including a link to the article or page where it is hosted. Ideally, these links should be backlinks with the image pointing back to the article or page hosting it, which will have a positive impact on the SEO of your content and on the image itself.
You can add links to your website on each pin you publish, which can help drive traffic to your site. Furthermore, if other users decide to share your pins, this can also help generate links to your website, which can have a positive impact on your SEO.
Pinterest often appears in search results for image queries, which means that your images published on Pinterest have an additional chance of appearing in search results for those queries.
Using Pinterest to optimise your image SEO can also have a positive impact on your overall online visibility, since there are users who do not limit themselves to Google when searching for images. In other words, by using Pinterest, you could reach a wider audience and improve your visibility.
Flickr
Flickr is a platform that may seem outdated, but which continues to offer many SEO advantages on Google and still has a substantial user base. It is useful for publishing images with links to pages on your website, much like Pinterest.
In addition, Flickr automatically publishes some EXIF data, allowing you to publish photos on maps and add keyword tags.
Instagram is a major social media platform for sharing images and can also appear in image search results.
However, individual image posts on Instagram are very limited in terms of passing potential link benefits, since links cannot be added to each post.
Post descriptions can be used to indicate to users how to find your media, such as your Instagram profile URL or terms to search on Google.
Instagram images are primarily used for image optimisation in their own right. It is important to use the post description alongside the image and to include several hashtags related to its content in order to increase visibility among other people interested in those topics.
Structured data for advanced image SEO.
Structured data for images is not very important depending on the context in which it is used. If you are an image bank or a professional photographer, these can however be important — notably by specifying in the structured data that these images belong to you and that they were taken or created by you. This practice can also partially protect you from copyright issues, preventing Google from ranking images you created that a competitor uses without your permission. Finally, this could potentially give you a small boost if someone searches for your name or company on Google, as images crediting you could rank more easily.
<script type="application/ld+json">
[{
"@context": "https://schema.org/",
"@type": "ImageObject",
"contentUrl": "https://example.com/photos/1x1/black-labrador-puppy.jpg",
"license": "https://example.com/license",
"acquireLicensePage": "https://example.com/how-to-use-my-images",
"creditText": "Labrador PhotoLab",
"creator": {
"@type": "Person",
"name": "Brixton Brownstone"
},
"copyrightNotice": "Clara Kent"
},
{
"@context": "https://schema.org/",
"@type": "ImageObject",
"contentUrl": "https://example.com/photos/1x1/adult-black-labrador.jpg",
"license": "https://example.com/license",
"acquireLicensePage": "https://example.com/how-to-use-my-images",
"creditText": "Labrador PhotoLab",
"creator": {
"@type": "Person",
"name": "Brixton Brownstone"
},
"copyrightNotice": "Clara Kent"
}]
</script>
If you wish to learn more about image structured data, refer to Google’s guide.
https://developers.google.com/search/docs/appearance/structured-data/image-license-metadata
Metadata in image files (EXIF and IPTC)
Metadata goes beyond these tags and can include structured data such as Facebook Open Graph tags, Twitter Card tags, schema markup, and so on.
As for images, they can carry their own metadata that is often embedded directly in the image file itself. EXIF data contains technical information such as the date/timestamp of the shot, camera make and model, aperture, lens, focal length, colour space, photo geolocation coordinates, and more.
IPTC data can include a title, description, keywords, information about the people appearing in the image, location information, copyright information, and more. IPTC data is probably the industry-leading standard, although many legacy pieces of content and older systems use EXIF.
Years ago, it seemed that Google might use as much metadata as possible to rank images, but this has not been confirmed. EXIF data can help generate keyword data for ranking on some image-sharing websites. Geolocation data embedded in EXIF can also be useful for local websites, but this has not been confirmed either. There are software packages and online services for customising EXIF data.
Google only uses certain parts of IPTC data, such as the image creator, credit line, and licence/copyright notice. You can use image metadata via the IPTC protocol or structured data, but it is best not to use both with conflicting field values. Although there may be arguments for using other IPTC and EXIF data on social media sites such as Flickr, this should only have a marginal effect on keywords and rankings. Ultimately, image metadata is unlikely to offer your website many benefits beyond conveying creator and licence information.
Originality
It is crucial to focus on originality when creating or sourcing images for your website. Google’s image search results tend to avoid duplicate content, just like keyword-based search results. If multiple copies of the same image appear in search results, Google considers this a poor user experience and tries to avoid it. If you use an image that has already been used elsewhere, you must modify it sufficiently to make it different, or find a new one. If you use product data feeds provided by a supplier, it is preferable to take your own photos. If the product is on a white background, you can ask a graphic designer to modify it to make it unique. Changing the colour, cropping, or even flipping the image can help make it different enough to avoid duplication. If your image is identical to those on other websites appearing for the same keyword, it is unlikely to be considered relevant when Google filters results.
Do not hesitate to be inventive — for example by using saturation and contrast. In various situations, your image may appear as a thumbnail, such as in image search results or as a thumbnail in web page listings. In these cases, your image is competing with others for the user’s attention.
One way to help your image stand out, thereby increasing the chances of getting clicks, is to slightly increase the contrast and colour saturation for the thumbnail version. A well-saturated, eye-catching image is far more likely to be noticed among dozens of competing results.
It is important to evaluate the image as a whole — can it be lightened and saturated slightly while still looking natural? Such modifications can have a positive long-term impact on the image’s performance.