(CNN) -- According to Facebook's new search feature, only two of my 526 friends like cats. Judging by the number of cat photos filling my feed every day, this is obviously not accurate.
It also demonstrates one of the big problems with Facebook's approach to search.
The company's new tool, awkwardly named "Graph Search," was announced with much fanfare at Facebook's Menlo Park campus Tuesday. The new search feature lets you draw connections between people, their profile information and their interests on Facebook. In theory, it's a good recipe for finding recommendations for doctors, businesses, products, TV shows or bands.
It can also be used to find people that fit a specific profile, such as "men over 30 who live in Cleveland."
A powerful search function is a logical and useful addition to the site, but the beta version is far from being the Yelp, LinkedIn or Match.com killer that Facebook may be hoping for. For that to happen, there would have to be some major shifts in how people use the social network.
Much of Graph Search's power and problems start with the Like button. People just don't wield the Like as often and as discerningly as is needed to turn Facebook into a useful recommendation tool. It's also too easy for those deep-pocketed companies who can afford to maintain a social media presence to buy more likes and come out on top.
A search for Mexican restaurants that my friends like here in San Francisco shows 12 options, most with just one friend's thumbs up. By comparison, Yelp has 543 reviews with star ratings for the El Castillito taqueria in San Francisco, while Facebook has 97 likes (none from my friends), and a few scattered wall posts on a sparse unofficial landing page.
Alternately, you can also search for restaurants your friends have visited, which turns up locations they have checked into using Facebook. But that information isn't helpful without knowing if they enjoyed their meal, if they hated it, or if they were just there as part of their job as city food inspector.
It's not that the social network doesn't have the data to turn it into a powerful recommendation tool. But for it to be more effective, Facebook would have to greatly improve how it collects information from people going forward, or expand its search powers to comb through status updates and comments.
Every post you make reveals slivers about who you are and what you like (not just "Like"). Technically, Facebook should be able to detect if you like cats, even if you didn't take the time to hit the Like button for a page called "cats." But it can't do this without rightfully alarming its already privacy-sensitive 1 billion users.
Searching for people presents its own set of issues. Graph Search scours your profile information so people can find you based on what school you went to, where you work, your religion or who your friends are. Searches can be refined using filters for every available profile field, including likes, work info, family connections and the Facebook apps people use.
Next time you need to find single male models in Omaha who are Buddhist and speak Spanish, go straight to the Facebook Graph Search.
A more practical use would be to find job candidates, which Facebook demonstrated at its press event. The possibility of using Facebook for job-recruiting searches in intriguing, but beyond listing current employers, people don't regularly update their Facebook profiles like their resumes. Facebook profiles are crafted with friends and family in mind, not potential bosses, and for many that's a welcome separation of worlds.
There is value in being able to search for friends of friends who work at a company where you are gunning for a job, but while blindly contacting someone you've never met on LinkedIn is expected, it may be uncomfortable on a network of friends.
Speaking of uncomfortable, Graph Search can be used to find new "connections," like single friends of friends who share your interests in canning and 18th-century costume dramas. Are we ready to turn Facebook into a dating site? Friendster and MySpace doubled as online pick-up spots, but these days people set up separate profiles on Match.com or OK Cupid if they want to find dates.
As with LinkedIn, dating-site profiles are constructed specifically to show one side of a personality. The things you'd like a potential date to know about you are not the same things you'd share with old high school friends, your aunt or job contacts.
Facebook is also not a place people want to receive pick-up messages from strangers (even if they are friends of friends).
Facebook recently announced you could pay $1 to send a message to a stranger outside of your network ($100 if that stranger is Mark Zuckerberg) and go straight to their inbox, a feature that makes a lot more sense with Graph Search. Suddenly there's a whole world of people you might want to contact, for jobs or articles or long walks on the beach.
Graph Search introduces new ways to search Facebook that are great in theory. The tool works amazingly well in the idealized Zuckerbergian world where all Facebook members are real people who complete their profiles honestly and update them frequently.
But in reality, the data people share on Facebook is flawed and incomplete. And so is Graph Search, at least for now.