by Jakob Nielsen

The following table shows the number of visits that have been recorded in the Useit server logs as coming from search engines and directory services (so-called "portals") in a one-month period (March) in each of the years from 1998 through 2004. The true number of visits is slightly larger since some browsers don't report accurate referrer URLs. Also note that the number of users is likely to be smaller than the number of visits since people sometimes use multiple search engines or repeat a search.

Portals and search engines accounted for about half of the "unique visitors" to Useit. In comparison, the HypertextNow site gets 30-40% of its visitors from portals and search engines.

Overall search-engine-driven traffic increased by 9% during the last year. This was much slower growth than the 95% recorded from 2001 to 2002. The extreme growth in 2001 was likely due to the emergence of Google as the dominant search engine, since Google has a tendency to drive traffic to high-quality websites.

The slower growth since 2002 is realistic for the more stable Web we are seeing now that Internet use is close to saturation in English-speaking countries. Even though Useit has an international outlook, it is written in English and thus attracts the majority of its traffic from the U.S., Canada, the U.K., and Australia. Other countries still have rapid growth in Internet use, but people are unlikely to find an English-language site through search engines if they enter query terms in other languages.

Traffic to a site from a search engine depends on three factors:

• the engine's raw traffic (so almost everybody gets a lot of visitors from Google and Yahoo)
• the interests of people using a certain kind of engine (so Useit got uncommonly many visitors from Google in its early years since early adopters tend to be the kind of advanced users who care about my topic)
• whether the engine emphasizes quality or quantity (so, since Useit aims at high quality content, it will get more visitors from Google, Ask Jeeves, and Mining Co. because of their emphasis on guiding users to the best sites - in contrast Useit gets relatively little traffic from weblog-dominated search engines like DayPop because Useit only posts a new article every two weeks).

The most interesting observations from the table are:

• Google continues to dominate Internet search . It accounted for 76% of the referrals in 2004 and 72% in 2003. In a study of how journalists use the Web in early 2003, we found that half of the journalists turned to Google as their first action when asked to perform research for a story assignment. Still, Google is not the only game in town. Yahoo and MSN are important as well.
• Looking back, Google went from number six in 1999 to a clear winner in 2001 and a virtual "takes-all" position in 2002, showing that the Web is nowhere close to being locked down : there is still plenty of opportunity for new sites . Google didn't even exist the first year I collected these statistics.
• Still, things seem to have quieted down somewhat lately: the top-10 list was the same in 2004 as in 2003, and the only change in the rank ordering of the top-10 search engines was that Netscape and About swapped places (#8 and #9). A few new search engines appeared and bear watching even though they haven't yet cracked the top-10: myway.com, MyWebSearch, and WebSearch.com.
• Microsoft Network is getting to be big as a source of visitors. This is probably partly due to an increased customer base, but it must also indicate that MSN is better than AOL at letting users out on the open Internet and helping them find higher-quality content than simply the service's own walled garden. AOL is much bigger than MSN and yet lets a much smaller part of its users find Useit.
• There are only about 7 search engines worth paying attention to : Google, Yahoo, MSN, Ask Jeeves, Lycos, AltaVista, and AOL. Together they account for 97% of traffic (same as in 2003).
• Services can stagnate as was the case for AltaVista , Lycos , and About (originally known as The Mining Company).
• Excite virtually died in late 2001. In fact, since Excite@Home closed shortly after suffering this big drop in traffic, it's remarkable that there are any visitors from Excite.
• Disney's Go (formerly Infoseek) went out of business in 2001; in fact, it is surprising to find any referrals from a dead search engine. More interesting, I predicted Go's troubles based on my analysis of the search referrals in March 2000.

The type of statistics shown in this table may be seen as a leading indicator for the future of portal companies : their ability to guide users to high-quality content is likely to influence their future use since people will tend to stay with services that give them good results and will tend to recommend the good services to their friends. Services that do not succeed in guiding users to good pages may suffer a drop in usage over the long term since people will change service when they discover a better one.

Number of visits referred by navigation services during one-month periods.

The next table compares the referral statistics to Useit in 2004 with general referral statistics for a large sample of the Web at StatMarket during the same period.

Proportion of referrals from various navigation services to Useit and on the Web in general in early 2004.

The last table compares the referral statistics to Useit in 1999 with general referral statistics for a large sample of the Web at StatMarket during the same period.

Proportion of referrals from various navigation services to Useit and on the Web in general in early 1999. A * indicates missing data.