• by Steve Smith , Staff Writer @popeyesm, January 11, 2011

When Google Goggles launched over a year ago, I was underwhelmed by a visual search app that seemed to have been released way before it was genuinely useful.

The standalone app consistently struggled with bar codes and product labels. The results were uneven at best. On Android, Goggles has its own dedicated version. On the iPhone it is embedded into the Google search app. I was alerted to an iPhone update the other day that caught my eye. The release notes suggested that Google visual searches now could solve Sudoku puzzles. I haven't a clue if this works, by the way, but only because I have carefully avoided this puzzle craze for years. I don't even know the basic rules -- let alone what a successfully solved puzzle looks like. "It is very relaxing," my fiancée, an unapologetic math geek, assures me. I, on the other hand, am math-averse. I get regular notices from my bank that I miscalculated the basic addition on my last deposit slip... by a couple of thousand dollars. 

So, I can neither confirm nor deny that Google Goggles now can solve a Sudoku puzzle. I just aimed the Android version of the app at a Web-based puzzled and it recognized that it was Sudoku and seemed to deliver something resembling a solution.

Google has an entertaining demo of the feature, pitting its technology against a champion solver.

More to the point, the Sudoku trick is really meant as a broader demo of the evolution of visual search. In fact, the latest Goggles release for the standalone Android app and iPhone Google search app is a very strong advance over what I saw a year ago. The Android app now has an instant bar code reader that automatically recognizes when a UPC code is in the camera's view and calls up the search result without even snapping the camera. Some other bar code readers have had this, of course, but the added convenience in Goggles just makes the casual user think of it as an all-in-one shopping resource.

In fact, while Google may have cornered the market in mobile Sudoku solving, it is also reaching out to cover categories of visual search that Amazon, eBay, Microsoft Tag and others have been working on. In the well-integrated iPhone Google app especially, Google clearly is trying to become a multi-mode search app that precludes the need for brand-specific solutions. Product recognition has improved noticeably. Indeed, older, failed scans that were stored in the app's history now deliver more accurate results when I run them again. Google seems to extract from a given object both image info as well as text to come up with a couple of possible solutions. In the iPhone search app, many of the results on products offered localized results, although these local retail shopping options usually were few to none. Still, the Google app is clearly headed in the direction of the recently introduced and very good Amazon scan app that delivers online pricing from multiple vendors on many goods, plus Amazon's own catalog listing.    

I was most impressed by Goggles' indexing of magazine ads. Google says it can understand the ads in most major magazines published after last August. The scanner successfully recognized almost every print ad I could throw at it from recent issues of Entertainment Weekly and American Heritage. The scans were so sensitive that the camera even captured and understood the cover text "American Heritage" that got into a shot of an EW page.  

The recognition piece of the puzzle has been solved more effectively than the results, however. In the end, Google is still kicking you into a general search result. The HTC phone ad or the Rosetta Stone language software ad just links to a manufacturer's general Web site -- still not the kind of directed brand experience a user would get (hopefully) from a mobile code program via Microsoft Tag or JagTag. Amazon and eBay still tend to keep users within their own mobile-formatted experiences, likely getting users close to a buy button faster.  

Nevertheless, Goggles is worth a new look. The art of visual search clearly has advanced significantly in just a year. The next obvious step for this app is to encroach on the nascent augmented reality search services that we already see tested in apps like Junaio and Acrossair. The most compelling prospect for the Google app is the way in which it is integrating multiple inputs -- text, proximity, voice and visual.

I have already gotten accustomed to the idea that voice search is remarkably accurate and fast. I rarely type a search entry any more. If the user learns that visual search is as reliable as these other input methods, then we have a mobile search platform that is as integrated and simple for consumers to use -- as it may be complex and challenging for marketers to master.