
By GUY DIXON
Thursday, November 21, 2002
Page B19
TORONTO -- There's a new corporate secret out there that's as closely guarded as Kentucky Fried Chicken's original recipe and Coca-Cola's soft-drink formula. It's the method Google Inc. uses for ranking Internet search results.
With Google's soaring word-of-mouth popularity as one of the fastest and easiest search engines on the Web, spammers and programmers have been poring over their sites to get a higher ranking on Google's keyword searches.
The stakes are high. Web-based businesses selling advertising and other Web services based around Google's search ranking system can find themselves shunted to one side when the search engine makes a small change in its computer codes.
A top ranking on Google counts, because being high on the list when a user types in a particular keyword search is as good as creating a massive billboard promoting a Web site.
A lower listing -- say, 25th -- although still pretty good, requires the user toscroll down three pages of Google search results, effectively turning the site into a tiny digital flyer fluttering in the wind.
As a result, Web properties are making it their business to get a higher position.
Common tactics range from inserting popular keywords and Web addresses in the text of a Web page to act as lures for Google's Web-indexing software to more dubious tricks such as attaching non-pornographic keywords to porn sites, all in a bid to manipulate Google's search technology.
It's a far cry from the innocent view of the Internet that the Silicon Valley company portrays with its clubhouse corporate culture, playschool-like logo and minimalist Web site.
"As Google gets to be bigger and bigger and becomes more of a force in everybody's life, a lot of people want to show up No. 1 for a certain search," said Matt Cutts, a software engineer at Google. "If I sell sneakers, I want to be the No. 1 result for sneakers on Google."
To keep Web sites from abusing its service, Google won't say much about how it counters others' manipulation. At issue are hyperlinks, those ubiquitous underlined words and pictures sprinkled throughout Web sites that link one site to another with the click of a mouse and that are at the core of Google's ranking system.
The basics of its technology have been widely known since company founders Larry Page and Sergey Brin created the program as PhD students at Stanford University in the mid-1990s.
The two originally called it the BackRub to indicate how hyperlinks can help a search engine determine the relevance of a Web page to a particular search. It worked so well that hyperlink analysis became the new gold standard for search engine technology. The underlying premise of what Google now calls PageRank is to assign a ranking from one to 10 to each site based on the hyperlinks linking to the site.
Thought another way, a hyperlink is a vote for a site. So, the more links leading to a particular Web site, the greater that site's importance.
Sites doing the linking are also judged by PageRank for their level of importance, creating a sort of majority rule among sites.
Google likes to see its role as hands off.
As the company's literature says: "There is no human involvement or manipulation of results, which is why users have come to trust Google as a source of objective information untainted by paid placement."
But as the basics of PageRank became widely known, the search engine became open to manipulation. Now Google uses ranking techniques well beyond its hyperlink basics. And that's where the controversy starts. Just how hands off is Google?
The company will say only that it uses more than 100 ranking methods beyond PageRank, such as analyzing the actual contents of a Web page.
"It's kind of a mix. We definitely are a lot more open than just about any other search engine. But it's also the case that we have moved beyond the basic things that you read about in the papers that we've published," Mr. Cutts said.
Much of this, he said, has to do with trying to fix core problems. "As we have time, as we have resources, we shut down various tricks, and those tricks stay pretty much shut down," he said.
One trick programmers have used is link farms, which are Web sites set up to do nothing but provide links to other sites, or link feedback, where one site agrees to link with another and the other reciprocates, and then both pages try to become more important in PageRank's eyes.
"It's sort of like an arms race. They come out with some sort of link farm that works; we fight it. They come out with something else, and we fight that," Google spokesman Tim Armstrong said.
But some have found other ways to benefit from Google's system. SearchKing Inc., an Oklahoma City-based Internet portal company, has a service that sells advertising space on Web sites that have a high search rank.
For $199 (U.S.), an advertiser can buy a five-word ad with a link on the site with a high level of importance of eight out of 10 in PageRank's eyes.
Most important, the ad can help to boost the advertiser's site by establishing a link back to it from the PageRank of eight site -- so the advertiser not only gets a widely seen ad, but also a higher Google search ranking. An ad on a site with a PageRank of five goes for only $29.
"Maybe the Net needs to be told what and how to link and maybe the people that should do the telling are Google, but that isn't how the Net was built and that isn't freedom," Bob Massa, who runs SearchKing, says in a statement on his Web site.
He is one of those who feels trampled by Google, even though he has built much of his business around the search engine. In October, his company filed a lawsuit against Google. The suit contends Google deliberately lowered SearchKing's rank of importance after it launched its ad service, and this also hurt other sites connecting to SearchKing, the company said.
A spokesman for Google would not comment on the lawsuit but said the company sometimes changes its programs to improve overall search quality "and, in some cases, the modifications do result in changes to the site rankings."
Another criticism is that PageRank has a bias toward established, popular sites to the disadvantage, some believe, of newer sites.
"Google is so important to the Web these days that it probably ought to be a public utility," said Daniel Brandt who operates a Google-suspicious site (http://www.google-watch.org).
Google counters that its search programs do take into account the age of a hyperlink and the importance of other pages doing the linking to reduce the chances of new Web sites getting lost in the shuffle.
Many would like to know much about just how Google's technology determines what's important and what's not, what's manipulative and what's legitimate. But Google isn't telling.
|