Hannah Thorpe, Head of SEO at Found, SEO Strategist Allegra Rosenbaum, and Rebecca Todd, Head of Search & SEO at Ryman, share their insights on whether visual search is, or should be, a necessary component of an SEO strategy.
Do you agree with Pinterest CEO Ben Silbermann that “the future of search will be about pictures instead of keywords”?
HT: It’s hardly surprising that the CEO of an image sharing website touts pictures as the future of search – but there is some truth in his comments. Search is seen increasing diversification both in terms of input and the platforms used.
The move away from keywords is happening, and has been for many years. The rise of new search methods such as voice search and, yes, visual search is leading to new ways of searching. Search engines are now much better at establishing a consumer’s real search intent, and matching these against the related entities in its index. Pictures are one new form of input, but the future of search is all about entities and intent.
AR: I think images are and will play more of a role in the future, but I think keywords will always be necessary. How else will users make searches? I think emojis might play a role in shortening long-tail phrases in the future, but I think we will always need keywords.
RT: Consumers respond better to images, for sure. They carry more meaning in less space, and it’s a fact that our brain’s can process an image much quicker than they can an image. But I’m not sure whether we’ll start using visual searches soon. I think the technology needs a little more work.
Sure, we have technology to recognise what’s in an image – my phone is grouping my photos by who’s in it every time I take a photo. But assigning meaning to those images is the technology that needs refining, in my opinion.
Search is all about deciphering meaning and intent, so until there are algorithms that can do that as well as our brains, I think search will be dominated by text.
Has your company been using visual search as part of your clients’ campaigns? If so, do you have any success stories to share?
HT: We’ve worked with a multinational FMCG brand keen to grow their presence in non-traditional search platforms for a particular product category.
We analysed the results for over 15,000 searches, bench marked performance and compared the images our brand owned with those available elsewhere. This enabled us to identify improvement opportunities to boost exposure in the space. In areas where the brand performed strongly in traditional web search but poorly for visual search, we often found imagery was lacking entirely or of poor quality, and were able to make suggestions creating new assets.
RT: No, we aren’t investing in visual search yet but when the time comes, I think it will be an easy transition. There isn’t much meaning needed to be deciphered for a picture of a pen, for example. But how refined the results can be will be the challenge, ie can we identify and serve the consumer the exact pen in their image?
Do you think the technology behind visual search is accurate enough for companies to adopt it into their search campaigns at this time? If not, what more is needed to make it more effective?
HT: We’ve come a long way since Google could not understand text embedded within images. Advances in image recognition, and particularly object detection, have been incredible in recent years. You only have to look at the ways in which sites such as Facebook are now auto-tagging images uploaded to their platform with related descriptions in “alt” attributes or the Google Vision API to get an idea of how smart the systems have become, and how much data is really out there. Whilst there are always going to be hilarities in some misclassifications, the ability of the tech giants to extract entities from an image is here now.
The introduction of shoppable ads on Google Image search shows the direction in which search engines are moving. If you’re a brand operating in the consumer goods space, you need to be considering image search now.
AR: Google still can’t crawl images as effectively as text, so I think there should still be a more heavy focus on text search. However, proper image optimisation, such as using srcset, optimising images to load quickly, descriptive alt text and image file names – all of these elements should be taken advantage of as well.
RT: The technology seems to identify the subject matter of a photo with some human programming. But the accuracy of returning the right search results with the right intent, in my opinion, is limited.
For example, I can take a search with a photo of a pen, but am looking for information on that pen like reviews, a replacement pen, or a replacement ink cartridge for the pen. Although these searches would all be related, the consumer is looking for different types of information that the search needs to serve.
With all this in mind, should visual search be a necessary component of an SEO campaign?
HT: The importance of visual search today very much depends on the sector in which you operate. B2C consumer product brands, and particularly those retailing fashion, absolutely need to be focusing on visual search now. Ensuring they have clear and accessible images from which search engines can build product entities is now a must to ensure a presence in Pinterest Lens and alternatives.
AR: SEO campaigns should be holistic in my opinion, and strive to improve all facets of search. That includes proper image optimisation for visual search.
RT: It’s important to future proof your strategies, but at this point in time, for me anyway, it’s not a necessary part of our strategy. It’s a development to monitor and bear in mind.
HT: Traditional image search through text-based queries in platforms such as Google Images has a much larger share of the search landscape than most people expect (four times that of YouTube). So even if you don’t think photo-based objection recognition will work in your space, don’t neglect image optimisation entirely!
RT: For me, regardless of medium (image or text) the key to SEO strategies centres around consumer intent. No matter HOW they search, we have to consider WHAT they search. What are they expecting to find, what is their end goal?