Rosie Hopes, our Head of Insight and Strategy, has taken a deep dive into the search landscape in 2023 in our latest content series. Read on to find out how visual search has developed over the past decade, how it is disrupting SEO strategies and what we as marketers can do to take advantage of it.
When we hear “search marketing”, normally we think “SEO and PPC” – but the search landscape has expanded far beyond a simple text search in Google. Users are seeking information, inspiration and social proof on more platforms than ever, from social media to chatbots. This shift is driven both by changes in user behaviour, especially among younger generations, and by technological leaps forward in machine learning and AI.
One of the important developments in search that is sometimes overlooked in all the furore over ChatGPT is the growth of visual search as opposed to traditional text search. The usefulness of this is best described by Pinterest when they first released their “crazy-fun new visual search tool” in 2015:
… we’ve got a new tool that lets you find all those things you don’t have the words to describe.
Rather than using text to describe what you’re looking for, visual search enables you to use an image as your search query, generating (mostly) visual results.
How Does Visual Search Work?
This is not new – Google’s Reverse Image Search has been available since 2011 – but it is growing ever more sophisticated thanks to developments in AI and machine learning. The process works as follows:
- A smartphone user uploads or takes a photo with their phone
- The software uses computer vision to interpret the image and pull relevant results
- Search results, which could be similar images, product listings, websites or translations (the list goes on!), appear on the screen
Visual Search Tools: The Ones to Watch
Pinterest Lens
The most established player in the visual search field is Pinterest, which has been innovating in the space for over a decade. The hard work of their software engineers came to fruition in 2017 when they released three visual search tools:
- Pinterest Lens – allowing users to ‘point and shoot’ their phone camera at something in the real world and generate relevant pins
- Instant Ideas – allowing users to refine their feed by self-selecting which pins they like the look of, which generates a series of more relevant results
Shop the Look – allowing users to pinpoint different items inside a pin and browse through product listings for them
What is particularly interesting about Pinterest is the sophistication with which their AI can generate results that go beyond just finding a visually similar image. For example, if you point Pinterest Lens at an avocado sitting on your kitchen counter, the results won’t just be other pictures of avocados. You could get pins that feature avocado recipes, growing tips, DIY beauty treatments… It’s about idea generation rather than just product matching.
Google Lens
While Pinterest may have been the trailblazer when it comes to visual search, Google is responsible for making it commonplace with Google Lens. What started as a standalone app in 2017 has been slowly integrated into other Google products (including Google Photos and Google Assistant) and, as of 2022, has replaced Reverse Image Search and is fully available on the web.
There are so many uses for the technology:
- Shopping – like Pinterest, you can take a picture of an item and pull results of visually similar items online
- Barcode and QR code recognition – you can follow a link directly from a QR code to a website without having to download a specific app
- Item identification – it can help you to identify a species of bird or a flower that you snap with your phone
- Location identification – it can identify your surroundings and provide information like business reviews or opening hours
- Smart text selection – it can detect written text and allows you to copy and paste it; useful for things like WiFi passwords, addresses and dates and even to create calendar invites based on the information
- Translation – taking the smart text function a step further, it allows you to translate menus or street signs by simply holding your camera up to it
The big jump forward here is the combination of visual search with augmented reality. When you hold your camera up to a menu and click ‘translation’, the results aren’t pulled into a SERP – they are superimposed on the image in real-time.
Google Multisearch
In April 2022, Google announced the next iteration of Google Lens: Google Multisearch, which allows you to augment your visual search using text prompts.
For example, you could use Google Lens to search based on a black jacket that you like and then add the query ‘blue’ to change the colour of the results.
Google Multisearch addresses the rising appetite for visual answers and a resistance to waiting, scrolling and sifting through too much information. Ultimately it makes finding answers quicker and easier than ever.
Visual Search and User Intent
What is really interesting about visual search is the way in which it subverts the traditional idea of keyword research and search intent.
To understand this we need to wind right back to the fundamentals of search marketing. At the dawn of the search engine, marketers had a revelation: they didn’t have to identify the right message, find the right audience and engage them while in a buying mindset like they did offline. Instead, the buyer would show up when they wanted something, type in exactly what they needed and all marketers had to do was be found by them. The key to knowing what their audience wanted and making sure that their brand was there to answer them was keyword research.
A huge part of SEO has always been understanding what people are searching for and categorising their search queries into different intents depending on where they are in the funnel. Search marketers are used to serving these intents with different tactics per channel.
When it comes to visual search, however, the intent-based model that SEOs are so familiar with is being disrupted. Rather than providing a key phrase with modifiers that help to return the right kind of result, searchers input an image they have taken with their camera.
Why is this important?
- As every image is user generated and reflects what the searcher sees in front of them at that time, every search query could potentially be unique – making search volume hard to accurately quantify
- An image on its own doesn’t necessarily indicate any search intent – on Google Lens, users select what they intended to find out using their image by scrolling through the ribbon at the bottom of the screen
For example, I took this photo of a picture I have hanging on my living room wall. The picture itself – aka the search query – doesn’t tell Google what I want to find out. Do I want to use Google Lens to translate the German annotations on the picture or do I want to shop online for similar pictures? I can use the ribbon (offering Translate, Text, Search, Homework, Shopping, Places and Dining options) to find the results I’m looking for.
To my knowledge, Google does not give SEOs any information about what identifiers are most used in images for visual search, or what intent the searcher went on to select. This essentially means that keyword research, that vital tool in an SEO’s arsenal, is impossible for visual search.
How to Optimise for Visual Search
So how can SEOs optimise for visual search queries if they don’t have any keyword data?
There’s no need to panic quite yet. Visual search is never going to completely replace text search, because it has its limitations. While it might be quicker to snap a photo of a plant you pass in the park and do a Google Lens search to find out what species it is, how would you perform a visual search for something like “what is hyperconverged infrastructure” or “best places to travel in April”? Keyword research – based on text searches – is still going to be a vital source of information when it comes to understanding what users want and creating content to serve it.
The big change is how those users access your website from the search engine. Anyone who has used visual search will have noticed that the SERPs are, well… more visual. Although the ‘Search’ tab can pull through traditional text results like a mobile search, often queries will generate image-led results. The ‘Shopping’ and ‘Places’ tabs are entirely visual. This really drives home the need for a proper image optimisation strategy for your website.
If you want your website to be found in visual search, image optimisation is vital.
Image optimisation guide
There are a few important points to remember when optimising the images on your website:
- Balance image quality and file size
Tech SEOs have been fighting with website managers for years to reduce the size and weight of web images, in accordance with Google’s guidelines on site speed and its importance as a ranking factor. While it’s still essential to keep images as light as possible, images shouldn’t be low-resolution or blurry. This might make it more difficult for Google to identify it as a match for a visual search query – plus, it’s likely to get a bad click-through rate if it isn’t clear to users what is being shown in the image.
- Optimise image file names, alt tags and captions
The AI that is used to interpret images does so by generating descriptors for what it finds (eg. green, shoe, lace-up, Nike, etc.). These descriptors are then compared with all of the results in its index. The more context you give to the images on your site, the easier you make it for the crawler to find and retrieve a visual result to pull into the SERPs.
For example:
File name: pink-axolotl-on-rock.jpg
Alt text: A pink axolotl perches on a rock in an aquarium
Caption: We think that axolotls rock!
Note that, regardless of image optimisation, alt text for images is also an important accessibility measure. People who use screen readers rely on it to understand what an image is about.
- Use proper structured data markup and add an image sitemap
Again, this is about making it as easy as possible for a search engine crawler to navigate your site and retrieve the information it needs to rank your images. You can find out more about schema markup options here and image sitemaps here.
Optimising your owned presence
Don’t focus entirely on your website and forget about the other ways that users might find your business while using Google Lens.
- Keep your Google My Business profile up to date
Your Google My Business profile is an important mine of information for Google Lens. Make sure that your logo, your address, your opening times, your menu and any photos of the premises are kept up to date so that it can match visual searches to your brand and serve the most relevant information to users as easily as possible. This is particularly relevant for the ‘Places’ and ‘Dining’ options.
- Keep your Knowledge Panel and Wikipedia page up to date
Again, this is important for exactly the same reason. Google will use these data sources to surface results, so make sure that you keep them updated regularly to show the most recent information about your business.
Key Takeaways on Visual Search
- Machine learning and AI have revolutionised visual search, making it increasingly sophisticated, useful and user-friendly
- While visual search is unlikely to replace text search completely, usage is likely to grow for quick, on-the-go search queries
- Visual search surfaces a problem for SEOs who may feel like they are operating blind because of a lack of keyword research to guide optimisation – but gains can be made immediately by properly optimising onsite images and ensuring that data sources like Google My Business and Wikipedia are kept up to date
We’re fascinated by the most recent developments in search, and we’re always keeping an eye on developments in the search landscape. To read more on this topic, download our Future of Search whitepaper or get in touch to have a chat with one of our digital marketing experts.