Wednesday, May 11, 2005

GOOGLE AND SEARCH SPAM Last Monday's Wall Street Journal carried a great column about the proliferation of search engine spam. Some highlights:
This is not the search spam of the early Internet...Rather, it's new and improved spam: pseudo-useful pages that are usually just shells for ads.

In many cases, a page might at first glance seem like a guide to your topic. But after a minute or two, it becomes evident that the information is virtually useless but is surrounded by an ocean of ads.


All of this is occurring because of a number of recent Web developments. The most important is that first Google and then Yahoo and everyone else have introduced advertising programs that make it easy and lucrative to sell ads on a Web site...

Thus, a kind of schizophrenia exists at search-engine companies. Half their engineering staff is busy trying to keep useless pages out of search results; the other half is busy coming up with tools that make it easier for people to create and profit from the useless pages in the first place.

The second development is that Web sites no longer use human beings much to help rank their search results. It's now largely done by software (even though some insiders say that Google and the rest use human editors a lot more then they let on).

The search companies have reduced their reliance on humans in part because they are expensive, and in part because the Web lately has been enthralled by the success of Google's PageRank algorithm, which ranks Web sites depending on who else links to them.

The problem, of course, is that spammers know about these algorithms and are constantly trying to trick them. Search engines respond by fiddling with the algorithm; spammers make their own adjustments; the beat goes on.
A related point.

Google is about to unveil a new advertising program that has at least the potential to exacerbate the search spam problem described above.

The background: Google ads currently operate under a cost-per-click fee structure. The advertiser only pays--and the site serving ads only profits--if users click through an ad to the advertiser's website.

But the new program will introduce a "cost-per-impression" fee model to work in parallel with the cost-per-click program: It'll allow advertisers more interested in exposure than clickthrough rates to pay based on how many people see their ad--whether the ad gets clicks or not.

And that means it may no longer be necessary for search spammers to get users to click on ads to get paid; getting web surfers to navigate over to their sites, alone, may become all it takes make money.

In that kind of environment, a high search-rank becomes an even more marketable asset. Because once the user reaches a site, the site's owner has already made money--even if the high-ranking page contains literally nothing of value to users. And so the incentive to game the system becomes that much more powerful.

Now, hopefully Google's engineers have thought long and hard about this. And depending on how the program is structured, an even more widespread search spam problem may not be the result. But the potential is certainly there.

CONTRAPOSITIVE is edited by Dan Aibel. Dan's a playwright. He lives in New York City.