In this guide, I will present you why and how to optimize your images for obtaining the most from both the Google Images Search vertical and Images Box in Universal Search, Rich Results with images, other Google SERPs’ features and from the more and more pervasive presence of Google Lens in every Google Search environment for your Ecommerce websites.
Images and Visual Search is a topic I talk about and write about since when Google started experimenting with the Images Search tags almost 4 years ago, a time when the majority of the SEOs were dismissing them saying it was just Images SEO, hence something that did not need to be prioritized.
Dismissing Images and Visual Search as a secondary SEO task would be deeply myopic now as it was before.
The worldwide status of crisis and suspended activity for practically any kind of physical shop makes it even more urgent that Ecommerce owners and businesses focus all the attention and efforts on visual shopping.
This guide is a humble attempt to guide you and offer you ideas and ways to optimize your category and product pages for it.
Why Visual is so important?
Amazon once gave the perfect answer to this same question:
Visual Search is perfect for shoppers who face two common dilemmas:
- I don’t know what I want, but I’ll know it when I see it;
- I know what I want, but I don’t know what it’s called.
An example of the first situation is a classic of marketing: the iPhone.
Before Steve Jobs showed it to the world, nobody was thinking about buying an expensive phone to barely surf the web with (it was 2g at the time).
An example of the second situation is very common: just think when you must visit a hardware store to buy a screw. What screw? With cylindrical head or flared or KAPPEN-type or flat or whatever? And for what? For fixtures? Wood? Concrete? Straps?
I don’t know you, but I usually go to the shop with the screw in my hand and tell the clerk: “I need a box of these screws”.
Market investigations remark the increasing importance of Visual Search too.
For instance, accordingly to a survey conducted by Bizrate Insights in April 2019, 15% of people between 18 and 34 y/o and 13% between 35 and 54 y/o regularly use or sometimes use Visual Search when shopping (the so-called “Showrooming”).
The percentage of potential users of Visual Search is what should make us think even more:
- 62% of people between 18 and 34 y/o;
- 56% of people between 35 and 54 y/o.
In a similar investigation conducted by Emarketer, Images/Visual Search was preferred more than another (now a little less) hype technology: chatbots (61,2% vs. 27,7%).
Ok, but why Visual Search is important for SEO?
1. A first reason why SEOs should reconsider the importance of Images and Visual Search is the percentage of Search Market Share Google Images, which represented the 21.54% in the first quarter of 2019.
2. A second reason is the pervasiveness images had acquired in the Universal Search of Google (more about this later) having an important impact in search results CTRs as Eric Enge well explained in this study on Perficient Digital.
3. A third reason is how Images Search is barely considered in any SEO strategy, hence offering the opportunity to obtain great results to the ones, who will take advantage of this lack of serious competition.
4. A fourth reason must be looked for in the same evolution of Google from a simple Search Engine to a Conversational Search ecosystem, where its Assistant will seamlessly conversate with users with a mix of vocal, written and visual answers.
The most relevant reason, though, is this one:
5. Visual search creates a new space for research based on visual associations, and it can lead to a conversion with just one click from the image to the cart, eliminating passages that instead can cause the conversion to be lost.
Images and Search now
The Images Search vertical
Images are everywhere in Google Search… and represent an opportunity for anybody who wants to take them seriously.
I will try to explain it – quite ironically – being very visual.
Let’s search for “oversized glasses” and look at the Google Images vertical search page:
Images SERPs on the desktop present:
- The Images Search tags, which are entities related to the main entity object of the search query (e.g.: oversized sunglasses <-> Kim Kardashian) or properties related to that entity (e.g.: oversized sunglasses <-> wholesale). For this reason, Images Search tags are wonderful for entity search (side note: remember that in Images Search we can also find more “related searches” than in Universal Search);
- Shopping carrousel, present in my example because the search intent of “oversized sunglasses” is not informational but transactional;
- Organic Images Search results, which if you have properly implemented schema.org/Product – present the icon/label “product” and further information like “in stock / out of stock” if we have properly implemented schema.org/Product.
After having add in the last few days the “product” icon/label in the desktop organic Images Search results, desktop and mobile are identical:
We can already understand the opportunity here and how having our product images ranking in the first row of images of this vertical can offer us great opportunities in terms of qualified traffic.
Sure, in this example all photos are labeled as “product”, however, things change when we start looking at the Images SERPs of industries less structured data savvy:
Lamps industry for a search like “Lamps for living rooms”: only 4 results out of 16 (first 2 rows) are labeled as products (note: this is when we are writing and the results may be different when you read this guide).
Miniatures industry for a search like “Lord of the Rings miniatures”: only 1 result out of 14 (first 2 rows) is labeled as a product.
The most interesting part, though, comes when we click on an image search result:
The image search result above is from a product page of the official website of Games Workshop, which has a line of miniatures devoted to The Lord of the Rings, and it is a good example of how not to take advantage of the opportunity Google offers us with Images Search:
- The title tag of the image search result is… well, judge yourself. The fact is that the title tag is not even optimized (or something is not working on their backend side when it comes to automation of title tags generation).
- The image does not have any description because:
- no meta description is present in the code of the page;
- the image is not surrounded by any written content, that Google could have used as a source (in universal search, Google creates it from the first available paragraph of text).
- the alt text element of the image is not even present (alt=”” style=”opacity: 1;”>).
- Albeit being a product page, no schema.org/Product has been implemented, as the Rich Result testing tool advises us of. Not being eligible for rich results, it is not showing information like the price and the availability of the product.
This below, instead, is a good example of image search rich result for product pages:
The War and Peace Wargames website is not an SEO masterpiece, but it does its job well enough to stand out in Images Search with a rich result that can bring qualified traffic from Images Search, especially on mobile.
Images Search is especially thought for Mobile Search as we can see here below:
When we click on an image, the result occupies all the screen, making of the result itself a sort of a real product page, making of it a better result that the corresponding rich results in Universal Search, which still is just 1 of many search snippets visible at the same time.
Moreover, did you see the highlighted icons in the screenshot above?
Those icons hide another potential set of opportunities that Images, in its “visual search” variant, offers and that almost no SEO think about.
Google Lens and the opportunities of Visual Search
The first icon is the Google Lens one.
In the screenshot, we can also see a “1” on that icon. That number indicates the objects that Google lens has identified in the image. If we click on the Lens icon, after a fancy effect the individuated object will be shown to us like this:
Below, then, you can see that “Google Lens” has substituted “Related Images”.
Why? Because Lens relies on a different kind of concept than Images Search: object indexing based on image recognition
In Google, as well in Bing Images or Pinterest that came before Google Lens, this is how a lens works:
Here above we can see how Pinterest Lens works when we do a visual search
In the specific case of Google, in February it was granted a patent to help return image queries from searches identifying objects in photographs and videos. Bill Slawski wrote about this patent in this post on SEO by the Sea.
In the case of the image of the Three Hunters I used as an example before, the alternative results offered by Google Lens are not meaningful. I clicked on the Gloin miniature, and Lens presented me any kind of results but others Gloin miniatures or, at least, LOTR dwarfs ones.
Is it that Google Lens fault, though?
Well, not necessarily.
Image recognition algorithms are not perfect; however, the main reason of those misaligned suggestions is also due to the low quality of the image itself, and the worst quality of the image the worst image recognition algorithms work.
No wonder that Google insists on recommending the use of “high-quality images” in its Search Gallery when referring to the use of images in structured data.
If the image has better quality, and the object is evident and presented in a frontal view, then Lens can offer useful suggestions to its users:
On mobile, then, we can also frame a specific object even if Google does not recognize it at first (something we can do in Pinterest and Bing too).
Lens, then, will perform a visual search of the framed object and returns us results.
“Ok, that’s cool… but what does it has to do with SEO?”, you are maybe wondering.
A lot, in my opinion.
A big percentage of ecommerce websites see in their images the most relevant type of content.
Hence, knowing the principle of object indexing and how Lens works better and when it doesn’t, is important for recommending the kind of photos to use in the product pages:
- Classic product’s photos showing the product alone, as in the first example of “oversized sunglasses” I presented in this guide;
- Photos where the product is presented in a “real” setting but is always shown in a clear evident and recognizable way.
For instance, imagine the classic still life photo of a model wearing a pant our ecommerce sells.
She could be wearing also other products we sell, that we will present as “related product/style” on that same page.
Google Lens image recognition will be able to
- Recognize both the pants and the other products,
- Index them as objects,
- Classify them also thanks to the other information we provide with the classic Image SEO best practices (e.g.: alt=”Our green olive pants matches wonderfully with our mimetic silk blouse”).
Recognizing them, and having us interlinked the two products, hence created a relationship between them, there is a big possibility that Lens will suggest these products of ours to the users… they will click and land in our product pages and, hopefully, convert.
Google Lens, though, doesn’t live only as a feature in Google Images.
It is present in Google Photo, as we will see later, as an independent mobile application or in Google Assistant or directly from the mobile search bar in the Google App:
What opportunity does offer us visual search?
Many; for instance, be there in front of the people’s eyes when they do showrooming:
Let say that I want to buy a toaster. I go to the local shop and I see one that I like but I want to know more.
I open the Google app and, instead of writing the name of the product, I would click on the Lens icon in the search bar, point to the toaster, shot a photo and do a visual search:
As you can see, Google Lens recognizes the product and translates it into a query (the product name) and allows us to directly look at similar products (in this case they are all the same product I searched information for) or perform a deeper search using the object to written query as our search.
Are our product images optimized to be presented as “Similar products” or is our product page ranking, so that users can find it from a visual search?
The potential of Collections
If you ever used Pinterest, you know that one of its most common uses is searching images for inspiration and to save them (“pin” them) in dashboards, so to return viewing the saved photos more quietly and, eventually, starting a further phase of deeper search using the functionalities of Pinterest Lens.
Google took good notice of this and created Collections, which works the same way.
Collections are available both on Desktop and Mobile Images Search:
Here is where you can access your collections in desktop search
The collection link is presented more outstandingly on mobile search
How do Collections work?
Let say that I am looking for ideas for my next miniature to paint.
I go to Google Images, search for “Warhammer 4k miniatures” and, instead of spending too much time (important behavioral detail) looking at each photo, I start saving the ones attracting my attention (second important behavioral detail) in a Collection I name after my search following these steps:
“This Ultramarine is amazing! Let me save it in my collection.” (tap on the Collection icon)
“Ok, add it to the Warhammer 4K Collection… or should I create a new one named “Ultramarines?”
Then, when I have more time, I return to my mobile Chrome App (Collections is not yet available in the Google Search App, at least not in the iOS version), and access to my collections from Images Search:
Collections present 3 areas:
- In the first one, we can see the list of the collections we have created. Note that Google creates by default 3 collections:
- Favorite images;
- Favorite places;
- Want to go.
It is important to notice that Favorite places and Want to go are created by Google with connecting our private Google Maps dashboard with our private Collections (and vice versa)
- The “Recently Added Items”
- Suggested Collection based on our recent search activity (aka search history), which is extremely interesting too.
This is extremely interesting because connecting Images Collections (see above my “Warhammer 4K” one) with Google Maps, Google could eventually serve the need to buy a product from online to offline, which is a strong probable need (remember, about a 46% of searches – last data we have – have a local intent).
Then, I open my collections and finally I can see only the images I saved and can start a further step in my search journey of a miniature to paint next:
As you can see, here we have the same rich image search result we saw in Google Images… but we are not in Google Images.
Our work as SEOs (structured data implementation, Images SEO et al) has an effect out of the classic search environment and can generate traffic and conversions (in fact, I visited the Empire Game Store website and bought the Warhammer 40,000: Kill team).
If you do a lazy Images SEO job or do not pay the same attention to images as you are paying to written content, then you are not going to take advantage of these kinds of opportunities Google offers us.
From searching inspiration to buying in a few taps
Visual Search, as we have seen, really offers great opportunities for ecommerce websites to earn high-quality traffic to their product pages.
What if I tell you that product images are not the only ones that can lead conversions to your website?
One of the principles of visual shopping is, as I wrote at the beginning of this guide, the opportunity to discover something I want to buy that I didn’t know I was desiring.
Rarely, when we walk in a shop, we do it with a buying list in our Notes mobile application. We enter and see if there is something that attracts our attention.
The same happens when we read a magazine about home decoration. We pass through the pages and look at the photos of all those beautiful living rooms and bedrooms and kitchens looking for inspiration.
We have the same behavior when searching online.
Remember, an inspiration not buying.
So, we can start a search textually like “Decoración salón” (Living room décor in English):
and start binge-watching a photo after the other (and maybe saving the ones that we like the most in a Collection).
In some cases, Google Lens will tell me that there are a few objects I can investigate, like in this case:
However, in many other cases Google Lens is not able to individuate any object, as in this example below:
In these cases, though, as explained before, we can manually “focus” the object that interests us the most:
Note: this way of starting an object search was presented first by Bing, which offers a great visual searching experience (and opportunities) and about which it would be worth writing a specific guide.
You can see in the screenshot above how Google Lens was able to recognize the object – the Strandmon chair by Ikea – quite easily.
The Ikea images search result is perfect:
- The Title is a descriptive name of the product;
- Product structured data is correctly implemented and not limited only to the few required fields (name, photo, and offer or aggregate rating or review).
- The meta description is the original meta description of the page, and its microcopy is excellent because it can present the best value of this wing chair: its comfort.
I cannot but click and visit the Ikea Strandmon wing chair product page:
A conversion from a search looking for inspiration and after only two taps from the start.
Images in Universal Search
Images are not present only in the Images Search vertical.
Whenever Google individuates a potential “visual search” intent for a query, it will present an Image Box.
The bigger the implied “visual search” intent is, the more prominent in the SERP the Images box will be
For instance, for “Oversized sunglasses” on Desktop one of the first SERP features we see is the Images Search box:
Images Box is organic; hence it should be counted like this in our analyses, and it is a mini version of the corresponding Images Search page for the same query.
I call it a mini-version because it does not show all the images’ tags (#1 in the screenshot) and all the images (#2) we can see in the Images vertical search, that users can reach clicking on the “More images…” link (#3).
In this example, the tags here visible are 20 and not 30, and the images presented in the carousel are only 10.
It is interesting to notice that the images presented here not necessarily are the first 10 images we see in the corresponding Images Search page.
Very similar is the Images Box in Mobile Search:
The elements of the mobile Images Box are the same we saw in the desktop one but for an extremely important exception.
In fact, on mobile, Google is labeling the images with “Product” (or “Recipe” or “Video”).
These labels are present also in the Mobile Images Search vertical and, for a few weeks, in Desktop Images Search too.
Another interesting element based on images is the Popular product’s box:
This feature is organic, available through Google Merchant Center (and only for limited types of products) and it presents only product-related images (and categorizes them), it has a substantially infinite scrolling behavior when clicking on more results and it has prominent visibility in the SERPs (for instance, for “Oversized sunglasses”, it is the first thing organic feature we see today when I am writing this guide, end of March 2020).
When we click on any of these images, we see this:
Regarding images, Google recommends to submit high-quality images that are at least 650×850 pixels, and other additional 10 photos per each product we submit through feeds.
What does this Popular Product search result teach us? That entity search influences strongly images search, and structured data is its mean for creating this kind of search feature.
Entity Search importance is evident also in another feature that, even if it is not strictly “images search”, it is strongly biased toward a visual search behavior.
I am talking of the Refine carousels:
The refinements are all entity-based; in fact, they substantially are the same entities we see proposed in the Images Search tags carousels.
However, these results do not link to Images Searches, but Universal Search SERPs.
If we tap, for instance, on “Man” under the category “department”, we will land here:
Images Search, then, is present in 2 other Google Features:
- Knowledge Graph
- Featured Snippets
Rich Results, Images, and Structured data
Images in Universal Search are not just “Images Search”.
Images are everywhere!
We see them in classic search results like in this case for “Villaggi Sardegna” (somehow “Sardinia resorts”) in Google.it:
Pay attention to the quality of the images you are publishing on your website, especially when these are published in category pages like it is in the case of the two search snippets presented in the screenshot above.
I am talking about their technical quality (compress the images but try to not exaggerate with image compression).
However, I am also talking about their quality in terms of content. Try to choose images that can be easily understood, are clear and attractive even if reduced to the thumbnail size of 104*104 pixels.
We, then, can see images in a big number of Rich Results.
If we check the Google Search Gallery, we can see how the majority of types pretend or strongly recommend the use of images:
RICH RESULT TYPE | REQUIRED | RECOMMENDED |
---|---|---|
ARTICLE (AMP) | ||
ARTICLE (NON-AMP) | ||
EVENT | ||
HOW TO | ||
LIVE STREAMING | ||
LOCAL BUSINESS (GENERAL) | ||
LOCAL BUSINESS (RESTAURANT) | ||
LOGO | ||
MOVIE | ||
PRODUCT | ||
RECIPE | ||
VIDEO |
Pay attention to the Google Search Gallery technical specifications, especially to the details “hidden” in their pages.
The Devil tends to hide itself in those details, which also tend to be shared with other features and channels Google offers,
For instance, in the case of Articles, Google explicitly tells us that the images should be at least 1,200 px wide.
Google, then, presents us that same recommendation for obtaining the best results in Google Discover:
Everything is connected on Google Search, and we should take advantage of it.
What must we consider when doing Images SEO specifically for structured data and Rich Result?
Images XML Sitemap
It may seem a 101 task, but the real truth is that a big portion of all the websites has not an Images XML Sitemap at all.
Why is it important?
It is important because Google considers the Images XML Sitemap the most relevant source from where discovering the images published on a website.
This is even truer if the website is serving images via JavaScript.
You can check the specifications in this help guide by Google: https://support.google.com/webmasters/answer/178636.
However, creating an Images XML Sitemap by hand is literally crazy, but you can nest the images information in the general XML Sitemap (something that is possible to do) thanks to crawlers like Sitebulb or Screaming Frog.
In Sitebulb you can do this:
- Once opened a project, you click on XML Sitemap Generator;
- Click on Images;
- Tick “Include Images”;
- Tick “Include Image Alt Text” (this is not obligatory, but very strongly recommendable);
- Set up the maximum number of referencing URL to avoid inserting in the Sitemap not relevant images like the boilerplate ones.
If you have the images hosted in a CDN (or fake CDN subdomain), you can easily add the CDN’s address in the “Image CDN URLS’ box.
On Screaming Frog, the procedure is substantially the same:
- Click on Sitemap in the main menu;
- Click on Image in the pop-up box;
- Do not include “Noindex Images” (Google clearly states that images must be crawlable and indexable);
- Set up the maximum number of inlinks per image’s URL (as in Sitebulb);
- If you have the images hosted in a CDN, you can indicate it in the box “Regex list of CDNs hosting images to be included”.
If your website is a WordPress, plugins like Yoast SEO or Udinra All Image Sitemap are excellent options.
Images specification for Rich Results
When people implement structured data, they simply tag one image of choice.
This is not incorrect by itself, but better outputs can be obtained with just a little more dedication.
The problem is that how images must be tagged with structured data for obtaining the best Rich Results is not consistent amongst types, hence harder to be respected.
To make the life of SEOs better, here I present you a table with all the specifications presented at the same time:
TYPE | QUANTITY | QUALITY | SIZE/PIXELS | RATIO | IMAGE FORMAT |
---|---|---|---|---|---|
ARTICLE (AMP) | More than 1 ("provide multiple..."), and representative of the article | High resolution | 800K px minimum as width*height result. 1,200 px wide minimum | 16:9 - 4:3 - 1:1 | .jpg - .png - .gif |
ARTICLE (NON-AMP) | More than 1 ("provide multiple...") and representative of the article | High resolution | 300K px minimum as width*height result. 696 px wide minimum | 16:9 - 4:3 - 1:1 | .jpg - .png - .gif |
EVENT | At least 1 | High Resolution | 50K px minimum as width*height result. Recommended 1,920 px wide. Minimum 720 px. | 16:9 - 4:3 - 1:1 | .jpg - .png - .gif |
HOW TO | 1 image for each How To step | Not indicated | Not indicated | 16:9 - 4:3 - 1:1 The 1:1 ratio is the recommended one | .jpg - .png - .gif |
LOCAL BUSINESS & LOCAL BUSINESS RESTAURANT | At least 1 | High resolution | 50K as width*height result minimum | 16:9 - 4:3 - 1:1 | .jpg - .png - .gif |
LOGO | 1 image | Not indicated | 112*112 px minimum | 1:1 | .jpg - .png - .gif |
MOVIE | At least 1 | High resolution | Not indicated | 6:9 | .jpg - .png - .gif |
PRODUCT | At least 1 | High resolution | 50K as width*height result minimum | 16:9 - 4:3 - 1:1 | .jpg - .png - .gif |
RECIPE | At least 1 | High resolution | 50K as width*height result minimum | 16:9 - 4:3 - 1:1 | .jpg - .png - .gif |
VIDEO | At least 1 (thumbnail) | Not indicated | 600*30 px minimum | 16:9 - 4:3 - 1:1 | .jpg - .png - .gif |
The Image License Metadata
Recently Google presented a new set of structured data for websites to indicate in Google Images Search if one or more images hosted are available for syndication via license, and to offer a link to whoever may be interested in using the image(s) discovered in Images Search.
The guide offered by Google in its Search Gallery is complete and easy to follow.
These are the main things to consider:
- Images must be accessible without needing a log in or having an account;
- Googlebot must access the pages containing the images
- Licensable images should be presented in an XML Images Sitemap
- Licensable images must be tagged with these structured data properties:
- license (required);
- acquireLicensePage (recommended).
- Each licensable image must have IPTC photo metadata (they are embedded into the image itself). The IPTC photo metadata are needed just once per image:
- Web Statement of Rights (required);
- Licensor URL (recommended).
How to do SEO for Google Images Search?
In this guide, I have presented already some of the of things that we must consider when optimizing our images for Images Search (and their use in Universal Search too):
- XML Images Sitemap;
- Not blocking the images via robots.txt;
- Not impeding the indexation of the images via meta robots or x-robots “noindex”.
Here below, I present and explain other classic but so often overlooked Image SEO factors.
Alt text
Google indicates still the alt text (known also as alt tag or alt description) as the most important factor for optimizing an image for Images Search.
The alt text is the short alternative textual description of the image and Googlebot uses it as a mean for understanding what is represented in the image itself.
Pay attention to now fill this tag with exact matching keywords, as well to not use the same exact keywords in the alt text of all the images present in a page (for instance, your product page).
On the contrary, the alt text over-optimization can contribute to the so-called keyword stuffing penalization.
For instance, an image like the one here above, should have this alt text:
<img … alt=”Warhammer 40,000 Primaris Intercessors squad kit by Games Workshop” …>
Descriptive file names
Albeit not being a primary factor for ranking well in Images Search, naming the images files in a descriptive way is a way to offer Google a better context to understand the meaning of the images themselves.
For instance, the Primaris Intercessor photo should be named so: primaris-intercessor-squad.jpg.
As you can see, also the image file names should be both concise and descriptive.
There is, then, a question that sometimes pops up in International SEO and related to images file names: “Should we need to localize the images file names?”
The question is legit. However, the answer is: “No, you don’t need to localize them”.
The reason is less obvious than we may think and relies on Entity Search.
An apple is always an apple, even if I call it apple, mela (Italian) or manzana (Spanish). The entity “apple” is permanent, the word describing it changes.
That means that Google knows that the file name of the image of the entity “apple” is the image of an apple and manzana and mela, also thanks to the semantic context that image is used.
If you want to investigate more about this fascinating topic, I recommend you to start doing it from this excellent deck by Cindy Krum.
Captions
A caption is a title or brief explanation accompanying an image.
You have seen that I used captions in many of the images I published in this guide.
Image captions are very important to give Google (and our readers) the exact context in which an image is being used.
Furthermore, it is possible that Google uses the caption as the meta description of our image in Google Images, so we must pay attention and write a caption that is descriptive.
The text surrounding the image
Apart from the caption, Google considers also the text that surrounds the image for understanding the context in which the image is used.
As we just saw in the case of the captions, Google may use the written content surrounding an image for using it as an Image Search result meta description, and we saw a few examples of this behavior in the screenshots from the SERPs presented in this guide.
Correct image sizes, formats, and file types
As we have seen in the chapter dedicated to images and structured data, Google insists a lot on the quality of the images, ratios and file types.
Resuming, these are the characteristics of a high-quality image for Google:
- High definition;
- Ratios of 16:9, 4:3 and 1:1 (better all 3, but at least one of these ratios is needed);
- The images files must be or .jpg or .png or .gif.
How to have high-quality images and a good PageSpeed at the same time?
In this guide, we have seen how Google continuously asks us to publish high definition images.
However, at the same time we know that Google correctly suggests us how much images are guilty of the bad PageSpeed metrics many websites suffer of both in their desktop and mobile versions.
Considering how much PageSpeed has an impact both on conversion rate and organic visibility, how can we satisfy the apparently contradictory indications of Google?
We can do it if we take care of these things:
- Image compression;
- Responsive images;
- A Correct choice of image file type.
Though, a topic like images optimization for PageSpeed will need another guide to be explained in each detail, hence – if you want to know more about it – I strongly suggest you visit images.guide website.
What really matters for this Images and Visual Search guide is what to use to compress images without losing too much quality and what image type is better to use.
In the first case – what to use to compress images without losing too much quality – we have a few options:
- Using image optimization CDN like Akamai Image Manager, imgix, Image Engine, Cloudinary, Cloudflare Polish or Uploadcare;
- Using image optimization APIs like Short Pixel, Kraken.io, TinyPNG, Imagify;
- Using online solutions like Optimizilla or Compressnow or Squoosh App;
- Using a GUI tool like ImageOptim.
Secondly, regarding what image file type, Google is very clear: if you want to have a good PageSpeed, the recommended file format is WebP.
However, the same Google says us that the only file type we can use if we look for rich results (i.e.: product rich results) are .jpg, .png and .gif.
Contradictions of Google, I know!
A few weeks ago, I asked John Mueller of Google to solve the mystery, and this was his answer:
You can specify multiple formats on the page using the picture tag & srcset attribute. For rich results, I’d recommend marking up the images as documented.
— 🍌 John 🍌 (@JohnMu) February 28, 2020
How to do SEO for Google Visual Search?
You may think that SEO for Visual Search does not really exist. How can you possibly do SEO for something that is ruled by image recognition algorithm?
Legit question but answering “You cannot” would be myopic.
Let’s look again at how Pinterest Lens (hence quite probably Google Lens too) works:
Let me try to explain it in plain English.
We can see how 3 different searches work in parallel to finally blend into a search result “page” output:
- Visual Search, which is based on image recognition. In this case, each object recognized element of an image is associated to other identical or considered identical objects recognized in similar images thanks also to elements like “salient colors”;
- Object Search, where all the objects individuated by Visual Search are retrieved from the image index on a larger scale (in the example, the object “shoe” is recognized and retrieved from still life photographs);
- Images Search, which is based on the annotations (i.e.: “loafers” and “shoes”) every object has.
Finally, all these single results are blended in the search result page we see.
Knowing this process, then, we can individuate ways to do SEO for Google Lens search.
Images SEO
Images Search is part of Google Lens search, therefore everything described in the previous chapter about Images SEO must be executed.
Using structured data and correct images specifications
As every SEO should know, structured data are not a ranking factor and neither they are useful just for obtaining rich results.
The most important function of structured data is to help Google (and other search engines) to understand the meaning of the content and the relationships it has with other content published on our website and/or others.
Using structured data for images especially helps Google to understand the relation between the object image and the rest of the content of page, hence to the rest of the website (aka “this image is related to this product, which is related to these other pages and website”)
The better Google understands the meaning of a specific bit of content and the relations it has with the other bits of content on the page, the more will be the opportunities that Google will consider that bit of content answering to a wider set of searches.
As you can see, I am not talking about ranking but of organic search visibility.
Using high-quality images and photos
Image recognition algorithms are improving very fast, but they still are far from perfect.
This is something we experience every day when we do a visual search from our smartphone shooting a badly illuminated photo, or if we use a blurred or an extremely compressed image as a source for our visual search.
In other words, the worst is the quality of the image source, the worst will be the output of the visual search.
In practical terms, this means that the product photos (but also the “still life” photos showing the use of a product) must be sharp, clear, well-lit photos without aesthetic effects such as, for example, filters that alter the natural color of the light.
Avoiding mess
As we saw, image recognition works better when the objects composing the image (and that matter the most to us) are easily distinguishable.
This is valid also for the still life kind of photographs; when shooting them, the objects we want image recognition algorithms to recognize must be evident.
In other words, disorder leads to confusion, confusion leads to incorrect image recognition and incorrect image recognition leads to an image poorly optimized for Visual Search.
Trying not using stock photos or…
… at least modify them.
This suggestion will be more evident in a few lines, however, consider this simple logic:
if we use a stock image of a product that is used by hundreds of other websites, when someone will search for a product using Visual Search, our result will look identical to those hundreds one.
In other words, this is the same situation as having the same product description that the distributor gives us and that we republish as such.
For this reason, if you use stock photos, at least modify them so to make them unique.
Letting your images to be used by others
This suggestion seems partly contradicting the previous one.
In reality, it is not, because I am talking of “our images”, not images of others like the stock photos ones.
Why we should let others use our images, better if with the Creative Commons Attribution-Share-Alike license or the Attribution-No-Derivatives one if you want people to not modify them (more about CC licenses here)?
The answer is in the example here below:
In the image, we see how Google Cloud Vision “reads” the image of the Warhammer Eternal Castigators and it is able to know what websites have used that image.
The more websites use an image, the bigger the context Google has available to understand the meaning of the image itself.
In the example above, we can see how Google labels the Warhammer image with annotations like “figurine”, “miniature”, “knight” and “games”.
This applies also to entity recognition and search:
Surely, many of the web entities Google associates to this image are thanks to the original page where the image had been first published, but the context of all the other pages republishing it helps confirming them (i.e.: Warhammer Age of Sigmar or Games Workshop) and even adds new ones (i.e.: Citadel Easy-To-Build Castigators with Gryph).
Conclusions
Images Search has always been a sort of ugly duckling, or so the majority of SEOs consider it still.
However, since a few years and with the constant improvements not only in the same Images Search area but also in the image recognition and Visual Search technology, it would be extremely myopic to not pay attention to this facet of search.
People’s behavior is changing trained by years of visual social and because of the same human nature.
What was unnatural was typing a query, not voice or visual search, because “We begin by coveting what we see every day […] And don’t your eyes seek out the things you want?”, as well said the character of Hannibal Lecter to Clarice Starling in “The Silence of the Lambs”.
Super. Now sprinkle in some automation please 🙂
My only suggestion is to pull entities from Google Cloud Vision API and use these for tagging/structured data etc.
That’s literally the most comprehensive article about image SEO I have seen so far. Thank you for putting this together.
You can also try
https://imagecompresser.com
For compress multiple image at a time.
Do you think the effort in creating an image sitemap is worth the reward?
Great article BTW.
Nice! Very clear and comprehensive guide to optimize images. Thanks.
Ciao Gianluca,
articolo eccezionale. Grazie per aver condiviso le info.
Thanks for the article and for sharing your thoughts and data on image SEO. What are your thoughts about serving response images and using lazy loading for e-commerce sites to help with site speed. Particularly, when you cover a blog post regarding a product in more depth? This is actually something we cover in our image SEO optimisation guide (https://seosherpa.com/image-seo/) and have found lazy loading to be particularly useful in improving page load times.
Thanks Gianluca for sharing great SEO tips and insights, I highly appreciate it. Keep up the good work.