“OK, Google. Write me a blog post about visual search”. Worth a try, right?
10% of the UK population have a “smart speaker” at home. Whilst voice search is creating plenty of noise in the technology sphere at the moment, it’s lesser-lauded contemporary, visual search, appears to be lurking in the slipstream.
However, products like the Echo Show and Echo Spot, which made their debuts in 2017 and 2018 respectively, demonstrate that despite the popularity of voice search, we still see value in visual.
What is “visual search”?
Twenty years ago, ‘visual search’ meant looking for things using your eyes, plain and simple. Turning the house upside down in pursuit of the car keys, for example.
Today, visual search is a booming area of technology set to transform how we seek and consume information, yet the underlying principle hasn’t really changed. Visual search still boils down to searching for information, products and more using visual means.
These days it’s just less about eyes, and more about AI (Artificial Intelligence). Instead of using text, visual search uses an image as an answer to a query (reverse image search is also a form of visual search).
Ahh, AI. the much-maligned fodder of science fiction novels and films the world over. Whilst some of us may live in fear that AI might start a humans vs robots war, the rest are less war and more wardrobe, putting AI to work finding them the latest styles.
Pinterest Lens, launched in 2017, lets users snap a photo with their phone, or use an existing image on their camera roll, to find themes and pins related to the photo. Essentially, Pinterest Lens uses computer vision to help users identify, buy or create things they see online or in the physical world.
Google have an app called ‘Style Match’ which uses image recognition technology to provide shopping results for users. The idea is simple: you see an item of clothing you like, open the app and point your phone’s camera at said piece of clothing. Google will then pull up the exact same piece, or similar pieces for purchase through Google Shopping. It’s a seismic shift to the way consumers find products, which some have dubbed the ‘See Snap Buy’ area.
Through the Google Lens
Google Lens is a technology powered by AI. Whilst the name makes it sound like some sort of intelligent digital panopticon, its uses are far more benign (at least for now).
Google Lens brings together your phone camera and deep machine learning to make intelligent decisions about objects. Not only can it detect an object, it can also understand what it’s looking at and offer actions based on what it ‘sees’. (If you’re keen to learn more about Google Lens, this recent article from Pocket Lint should do the trick.)
In the name of research, I decided to borrow a Google Pixel phone and try it out. My chosen target? The dog. I held the camera aloft, half expecting results ranging from a fox to a fruit bat…
Pretty impressive! Like 90% of the general population, Google can’t tell the difference between the two corgi breeds, but I’ll forgive them that.
Directions go (more) digital
How many times have you been following Google Maps only to find yourself wandering in circles or missing the turning? Forget chasing Pokemon, AR for Google Maps is transforming the way we navigate; whether you’re perdido in Toledo or seeking the nearest loo in Peru, with Google Maps you’ll be able to hold up your phone and see exactly where to go.
Despite its incredible possibilities, progress on visual search has been frustratingly slow. Just like with voice search and natural language processing, there’s still a lag between the capabilities of technology and the human mind. Studies have shown that human brain can correctly identify images seen for as little as 100 milliseconds. Shapes, colours, sizes, patterns, all in the fraction of a section. That’s a tall order for a machine to replicate.
Changes to visual search
Earlier this year, Google removed the ‘View Image’ button from search results, in a move to make it more difficult for users to ‘steal’ images, after Getty Images filed antitrust charges against Google for image scraping. Whilst this might be frustrating for users, it does have its benefits for website owners, who may have gained more referral traffic since users are now clicking through to the website to access the image.
Monetising image search
In 2017, Google faced a backlash after they played an advert for ‘Beauty and the Beast’ to users via Google Home devices. Unsurprisingly, users felt the ad was intrusive and it was not well received.
Although Google vehemently denied it was, in fact, an ad, their interest in monetising voice search is not altogether surprising. After all, if consumers increasingly move towards using smart assistants to seek and consume information, this could have a detrimental impact on Google’s other ad products, especially AdWords.
Given that interest in monetising voice search is apparent, it’s little surprise that images are now being monetised too.
Shopping ad carousels are now appearing above organic results on mobile for the UK, as well as the US:
According to Wordstream, Google explained that the goal behind this change is to improve the user experience for its image search functionality, allowing users to find products more easily without needing to click through to various websites to check prices and so on. Shopping ads work well here since they provide merchant and pricing information directly alongside the image, something that’s especially useful on mobile devices.
How can companies capitalise on visual search?
Although it may be tempting to dive head-first into this glistening pool of technological opportunity, as with anything, groundwork needs to be laid. Before you start dreaming of 360 home viewings or virtual safaris, start with getting your house in order.
What does that mean? It means starting with the visual content you already have, namely images and videos.
A good place to start is Google’s Image Publishing Guidelines document, which was updated in May of this year to add more details around structured data, page speed, title management and user experience details. Some of these ideas are outlined in more depth below.
It sounds like common sense, yet many sites still don’t name their image files appropriately. If your image filenames are still DC20045, time to rethink. For further reading on this, head over to this article.
Ensure your images have appropriate ALT tags. If you sell products with model names or serial numbers, be sure to include them in your ALT attributes, however, remember that your ALT tags should be descriptive and not ‘salesy’ or keyword stuffed.
For web images, aim to keep the size below 100kb (tools and plugins like TinyPNG are great for reducing image file size). Large images tend to slow down your page speed and cause user frustration (particularly slow pages can subsequently have a negative impact on rankings).
If you really need a larger, higher quality image then offer users the option to view the larger image in a pop-up or on a separate page.
Page speed may seem like a minor factor, however, it can have considerable consequences on your bottom line. It has been widely reported that Amazon would likely face a $1.6bn annual loss in sales if their site was just one second slower. Now, not every website can hope to become a behemoth like Amazon, however, the consumer expectation of speedy sites is unwavering and without exception.
Another option is to make use of image sitemaps. These can contain URLs from other domains, unlike regular sitemaps, which enforce cross-domain restrictions. This allows webmasters to use CDNs (content delivery networks) to host images. It’s worth bearing in mind that using sitemaps doesn’t necessarily guarantee that your images will be indexed by Google, but it’s worthwhile considering.
To read more about image sitemaps, take a look here.
Ensure images served from a CDN can be accessed and crawled by search engines. Also be aware that by placing your images on a CDN, you are most likely removing the image from your own domain and placing it on the domain of the CDN, which can be problematic when other sites come to link to yours. Some CDN’s allow you to add a canonical header to images to allow you to point back to the primary version on your own domain, so this is well worth exploring too.
If you include structured data, Google Images can display your images as rich results. Furthermore, image search on mobile now includes ‘badges’ to help users find exactly what they’re looking for, such as recipes. In the screenshot below you can see images with a ‘recipe’ badge, highlighting to searchers a recipe is available. Badges are available for products, videos and animated images, in addition to recipes.
This needs to be done carefully, however, and further guidance is available here.
Undoubtedly, the rise of both AI and AR means there’s great scope for transformation of the search landscape, particularly the visual aspect of search. However, it’s important to first strengthen the foundations of your image SEO strategy before embracing the more advanced aspects of this fast-moving vertical.