by 
Last week, Google held an AI-centric event in Paris. Among other things, it launched its Chat GPT competitor, Bard. This represents Google’s efforts to catch up to the excitement around generative and conversational AI, much of which originated within its own walls.
Buried in all the event action, there was also a notable nugget for AR: Google Lens is now used for 10 billion visual searches per month. This follows the 8 billion mark that it announced at Google I/O in May, and the 3 billion that it announced in June of the previous year.
Though this volume is miles from the 2 trillion+ annual Google searches, it validates growth and broadened use cases for visual search. Launched initially with promoted use cases around identifying pets and flowers, the endgame is monetizable searches like shoppable products.
Search What You See
Backing up, what is Google Lens? It’s Google’s visual search play, which means you can point your phone at objects to identify them. Google calls this “search what you see,” or a CTRL+F function – the keyboard shortcut for on-page keyword searches – for the physical world.
Just as it sounds, this applies a fair amount of AI and computer vision. And though Google is getting lots of flak over the past few weeks for being late to AI, it’s been all over the underlying technology for years, including its work with Tensor and transformer (the “T” in Chat GPT).
Meanwhile, Google isn’t alone in visual search. Its chief competition comes from two places. The first is Pinterest Lens, which is Pinterest’s similar tool that lets you identify real-world items and pin them. The second is Snap Scan which lets Snapchat users identify items for social fodder.
The difference between all these players traces back to their core products and company missions. For example, Google will work towards “all the world’s information” while others zero in on things like food & style (Pinterest), and fun & fashion (Snap). It’s a matter of focal range.
With all the above, anything that involves products is where the opportunity is. The endgame is monetizable visual searches which boil down to “shoppable” items, as noted. For example, one Snap Scan use case is “outfit inspiration,” while Google Lens offers local business discovery.
All of these cases tap into the fact that Millennials and Gen Z have a high affinity for the camera as a way to interface with the world. Brand advertisers are likewise drawn to immersive AR marketing. This makes visual search one way these players can future-proof their businesses.
Killer-App Potential
The various approaches outlined above raise the question of what types of products shine in visual search. Early signs point to items with visual complexity and unclear branding. This includes style items (“who makes that dress?”) and pricing transparency in retail settings.
The common thread is shopping (follow the money). Pinterest reports that 85 percent of consumers want visual info; 55 percent say visual search is instrumental in developing their style; 49 percent develop brand relationships; and 61 percent say it elevates in-store shopping.
Another fitting use case is local discovery, as noted. Visual search could be a better mousetrap for the ritual of finding out more about a new restaurant — or booking a reservation — by pointing your phone at it. Google advanced this use case further with “Multisearch Near Me.”
These are all things that Google is primed for, given its knowledge graph assembled from 20 years as the world’s search engine. This engenders a training set for image matching, including products (Google Shopping) general interest (Google Images), and storefronts (Street View).
To further grease the adoption wheels, Google continues to develop visual search “training wheels.” This includes making Google Lens front & center in well-traveled places and incubating it in web search. This could reduce some of the friction around visual search.
And if Google is able to do this and cultivate user habits, visual search has killer-app potential. By that we mean it has the makings for a frequently-used and widely-applicable utility (just like web search), with a massive addressable market. We’ll see if it lives up to that potential.
More from AR Insider…
