Search powerhouse Google has ascended not only to the No. 1 starting place for most Web searches, but the name has become part of the popular consciousness, even spawning the use of its name as a verb. To Google is to discover the globe.

That’s all about to change and the world of search is about to bust wide open.

Not to say that the top three search engines – Google, Yahoo and MSN – will topple. Far from it. But there are search technologies that have recently launched and some just on the horizon that aim to give searchers new ways to find what they want, especially in niche areas.

The recent growing popularity of video search and search via mobile phones are just the beginning of innovations to come. Some predict that search by image, natural language search and search by speech are the next improvements to the search experience.

As sophisticated as Google is at what they do, it is still an engine that relies on text – text is matched with text and results come from weighing the quality of the potential matches with the text you submitted. And there is slowly growing dissatisfaction with the relevancy of search results. A survey by Outsell stated that “search failure due to irrelevant results” of Google users grew from 28 to 30 percent in the last two years.

Start-up companies have perceived an opportunity to look beyond text search. And some companies believe they have novel methods for using text search to get more specific results.

Beyond Text

Moving completely away from text, Riya – based in San Mateo, Calif. – recently launched, what it calls a visual shopping search engine. The engine recognizes likenesses – in faces and in clothes, shoes and jewelry – based solely on visual cues. “In some cases words work just fine for search,” says Riya CEO and co-founder Munjal Shah. “But take a tie or jewelry pattern. There are things [for which] words fail us. We introduce the photo as your search start.”

Riya isn’t just going to help you shop for clothes. Using their technology will help families with Flickr accounts organize their vast digital photo albums by allowing users to “train” the Riya system to get better at recognizing people photos and to know when similar faces do not belong with your albums.

Perhaps most importantly, Shah says they now have a business model to go along with the cool technology. “Look inside the photo search paradigm,” he says. “The way it makes money is questionable. Face recognition software isn’t going to make much money. But now we have CPA from some merchants and CPC from some merchants.”

Part of their push is to get merchants to upload high-resolution images since better-quality pictures allow Riya to better differentiate the details of a belt buckle taken from a full-body shot.

With this model, Shah says that he isn’t in the business of competing with Google, but rather any other visual search engine to come. “Our belief is that it is too hard to win [with text],” he says. “We want to think off the map. That big of a paradigm shift is needed.”

Right now Riya is sticking with soft goods – a $30 billion sector – but Shah doesn’t rule out furniture, garden or china patterns eventually. He’s decided soft goods are the “beachhead.” They also have a mobile application that is coming soon and right now people can search Flickr photos of people using Riya’s technology.

Shah is thrilled that the technology is regarded as cool, but is continuing to look for revenue streams. “I’ve changed strategies once; I’ll change it again,” he says.

No Searching in Tongues

Still in extreme beta but looking to launch a public beta in 2007, Powerset, based in Palo Alto, Calif., is a natural language search engine. Barney Pell, CEO, sets it up like this: If you put a stream of text into a search engine, the engine will only find the same characters, he says. “It forces people to do their work on the computer’s terms. You have to figure out what the right terms are for your search.” He says that you are out of luck if you use the words in the wrong way.

Pell says, “Natural search is a complete mess.” You spend a lot of time just trying not to be a spammer, he notes. “The sites that are winning are not the ones that have the best message or solution.” He says the Powerset engine reads all Web pages and matches all words for context, not just keywords. “It can tell the difference between optimized search engine babble and content,” he says. He adds that a ranking on Powerset will rise by focusing on the quality of your site.

Marketers, he says, will love it because advertisers won’t have to anticipate the words people will use. Pell says that for Hawaiian handbags, for example, a user may type in “Hawaiian bags” or “tropical bags” and still not find what they want. “There’s a whole industry around trying to guess the right words and at what price,” he says, adding that with Powerset, marketers won’t have to get up every morning and decide which words to buy that day.

Early adopters will get dramatically better results, he says, because the public beta will probably involve a tool where you can submit according to organized categories and submit to the engine in advance. This way, users can help define the context before the engine officially launches.

Making Search a Snap

While some technologies are focused on the back end, others want to change how the users view search results. Launched in the middle of 2006, wants to take advantage of the pervasiveness of broadband to bring the user a more detailed search experience. The search engine essentially gives users home page preview panes for the text results when they plug in a search term. Users can rate the relevancy of each result so that the engine gets to know how to better serve keyword results.

“Our search isn’t for everybody,” says Snap CEO Tom McGovern. “We’re not trying to out-Google Google or out-Yahoo Yahoo. Savvy users like the preview feature and it gives us recognition.”

The visual results are meant to make it easier and faster to get to the right information. McGovern says Snap offers a different kind of advertiser experience as well, with a “risk free” CPA pricing model. Advertisers pay only for a predetermined action such as a sale, a lead, a download or other user engagement. The advertiser dictates the desired action and agrees to pay either a fixed rate or a variable percentage. Advertisers also choose the landing page or creative so that they pay only when they get the action and not on clicks. Click fraud is reduced this way. McGovern says their CPA model will be the standard in five years.

When users enter a search word in the field, a list of popular search phrases using your term is already populated in a drop-down window. The preview home page images are not screenshots from last week, but fairly up-to-the-minute captures, which comes in handy when cruising for news sites or video sites (the images are about a quarter of your monitor’s screen – big enough to read larger newspaper-headline- style type but not much else).

Snap – owned by Idealab – has an image search function as well, which still relies on title tags and not visual similarity, but the results are presented in nice thumbnail sizes that scroll horizontally. The site has also launched its Preview Anywhere function that offers blog visitors a preview of hyperlinks.

Other interesting features include the ability to type “movies” with a colon and a search term and it finds that term on the Internet Movie Database. Also, “fact” plus colon and search term brings up Wikipedia, and “define” plus colon plus search term brings up the Dictionary. com definition. With slight variation these features are also in Google except for the Wikipedia element.

“We want to win with the end users,” McGovern says. “They come to us because they like the technology.” He says they know they are not going to displace Google. “We want to be the secondary search engine of choice.”

The Human


Helping users find exactly the right thing from their search is what motivates all these start-ups., currently in beta, is no different, but it approaches the problem from a less-technological angle. uses human guides to help users find results. Search queries are sent to “a staff of experts” who then chat with the user to help narrow the search. Users do not have to use a guide. They can use the search field just like Google, but the results are custom- culled based on the index of questions that have been asked thus far, so the more questions the guides receive and the longer the search engine exists, the better it gets.

“The vast majority of people do not go past the third page of results,” says Scott Jones, founder and CEO of That means, “people never look past the top 30 results even though the top search engines are theoretically returning thousands or millions of results. Therefore ChaCha’s model of returning the very best results on the first page is vital. Users don’t really want millions of results in a split second. They want the answer as quickly as possible.”

There is still plenty of technology at work on the back end. In its first year, filed 18 patents. Jones was also the brains behind Boston Technology, where he invented a voice mail system that is basically the telecommunications standard, and built Escient, now known as Gracenote, the music recognition software that is used by nearly every music-ripping program and MP3 player on the market. has grown its workforce to 25,000 in its first four months of operations, signaling that the company plans to play it big. “Obviously,” Jones says, “when a search company achieves a $100-plus billion market cap in less than a decade, it creates a lure for all those who think they can do it better,” referring to Google. However, he notes, competing with Google is probably not a smart move. “Actually, we consider ourselves a collaborator with Google,” he says. “We are happy to have them around. And we’re happy to have all the other thousands of deep repositories of information that Google doesn’t even know about.”

ChaCha is free and gets its advertising dollars no differently from first-generation search engines. Because of the “very targeted keyword handling,” advertisers, Jones says, can get much better bang for their buck. The site also has its ChaCha Underground, a kind of social site where users and guides can gather together. Targeted display ads go there. Also, it can show targeted video advertisements when searching with a guide.

A Slice of Google’s Pie

While every new search engine touts its originality, many seem to be trading in on the text search market that Google pretty much has cornered.

Some other interesting search variations include, which is a browser toolbar that saves a trail of where users searched and what they found along the way. Users can save the information found so that next time they plug the search term in to the toolbar, it goes right to the page set as the preferred destination – kind of like favorites. is an engine for job seekers that culls listings from all over, adding more than a million new jobs each week. Users fill in the fields for job title and location and the direct links to the jobs on the hirer’s websites are displayed. And is built kind of like a wiki, called Swickis, which are search engines generated by individuals and used by groups. The groups then learn from users how to deliver more targeted results.

On the horizon is Jimmy Wales, who started Wikipedia, and his announced venture into a wiki-inspired search engine, but details are scarce at this early stage in Wales’ project.

Everyone is coming after Google, but the search behemoth is not standing still. John Battelle, search guru who wrote the book, The Search: How Google and Its Rivals Rewrote the Rules of Business and Transformed Our Culture, thinks a voice-driven search engine is the next innovation. He says that Google is working on search based on the engine’s understanding of a user’s previous search history “in a way that makes sense.” And while not Internet search, Battelle would like to see a desktop search capability that is intelligent.

What all these new search engines are really offering is a multiple search engine world, where users choose the engine based on what kind of information they need. And observers say that finding an engine that works for each user is about to get easier and easier.