Update, 2/21/2009:


In my attempt to understand how mobile keyword research differs from keyword research for the World Wide Web, I took a look at the top mobile queries in AT&T’s Beyond MEdia Net search engine
from September 07 as displayed in the public JumpTap Mobile Keyword Tool at AT&T
. What follows are those top 100 mobile keywords by volume, followed by breakdowns by subject and query intent. Analysis of these mobile queries led to a few surprising learnings, and
some best practices for marketers doing mobile keyword research.

Top 100 Mobile Keywords


Mobile Query Intent

Subjects of Mobile Queries

Mobile Web Versus WWWeb Queries




The JumpTap Mobile Keyword Tool on AT&T’s Mobile Developer platform shows mobile
searches from AT&T’s Beyond MediaNet mobile search engine. When a character is put into the
keyword tool it returns the ten mobile queries with the highest search volume during the month of September 2007 that contain that character. Inputting the alphabet and the numbers 0-9 produces, if
not an exact representation of the most popular queries in the index, then certainly a close approximation. Duplicates were removed before the data was categorized with the Live Search ad intelligence tool, and web search volume was added for the sake of comparison. When the Ad Intelligence tool
did not return a category or returned an obviously inaccurate category, the keyword was categorized using ODP categories. Query intent was defined using the method
set forth in Determining the User Intent of Web Search Engine Queries.

According to the classification, the mobile queries were overwhelmingly navigational in nature, with almost three quarters of the queries coming from users who already knew what they were
looking for:

query intentpercent of queries

What’s remarkable about this is that the results differ greatly from most studies of computer-based queries in that navigational queries are in the minority for most computer-based searchers.
Early studies by Rose and Levinson and Broder put navigational queries at 24.5% of the sample at most.

The disparity could be the result of a number of things, including the fact that the sample size consists entirely of head terms, which could be primarily navigational in nature. However, the
abundance of navigational queries is consistent with a hypothesis put forth by Google researchers Kamvar and Baluja in Deciphering Trends in Mobile Search (2007) about the lack of diversity in mobile queries that searchers are using
queries that they know return “usable” sites. By entering a query for a site or a subject that they know exists on the World Wide Web, mobile searchers could be hoping to find usable sites.

Best practices for marketers:

  • Using navigational queries such as branded terms, competitor terms and names of known products as core keywords could lead to increased visibility in mobile search engines.
  • Since this and the Kamvar/Baluja research seem to show that users are looking for usable sites, developing a mobile web site and including the term “mobile” in the copy as a secondary keyword
    could lead to increased visibility in mobile search engines.

There have been a couple of interesting studies on subjects of mobile queries, including the aforementioned Google research and an iCrossing panel study. One of the striking differences between this categorization of mobile queries and the Google
research is the near absence of adult queries. Of the nearly 3 million searches in the AT&T mobile search index, only 25,000 could be considered adult in nature, and the query itself (“hot
girls”) is too benign to be considered adult by most definitions of the word. It’s possible that adult queries were filtered from the index, or that AT&T’s mobile subscribers are more genteel
than the average Google user. However, given the huge disparity between the more than 25% adult queries in the Google study and the less than 1% in the AT&T top 100, it’s likely something else
is at play.

The overwhelming majority of searches in the AT&T index were in the Computers & Electronics category, followed by Sports & Recreation and Arts & Entertainment:

Within the category with the highest search volume, keywords related to search engines/portals were the most popular by far, followed by keywords related to social networking. This is surprising
given the prominence of the search box on the MEdia Net and Beyond MEdia Net portal pages. These searchers could be looking for popular webmail services, or simply prefer Google and Yahoo! search
to MEdia Net’s alternative. This last point would contradict the notion of users searching for “usable” sites, however.

The Sports & Recreation category was dominated by general sports searches like “espn”, which isn’t surprising given Ad Age’s recent revelation that the ESPN mobile site sometimes gets more
traffic than the web site

In the Arts & Entertainment category, celebrity queries were the clear winner, followed by the photography category. It should be noted that “photography” as categorized by the Live Search
Excel add-in includes words like “images” and “pictures”, which could be images of just about anything, including celebrities.

The following graphs represent the remaining broad categories with more than one subject within.

What the Ad Intelligence tool classifies as Science, Social Sciences & Humanities consists of exactly two keywords: “horoscope” and “weather”. Of these, weather is the most popular, with
over 120,000 searches during the month of September.

Half of what is here classified as Education & Instruction– keywords related to maps– could also be classified as pertaining to local information, which is a category used in the
aforementioned iCrossing mobile search study. Taxonomy notwithstanding, maps dominated this category, with less than 10,000 searches related to the keyword “dictionary”.

In the shopping category the transactional query “free” dominated, besting both ebay and the popular home improvement retailer Home Depot.

Finally, in the Families & Relationships category, Kids & Teens was the most popular category, followed by Romance, represented by the query “love”.

For the full breakdown of queries by category, see the complete mobile query data.

Best practices for marketers:

  • Understanding what categories mobile users are searching for most can help when developing mobile content, as high volume categories will likely drive more organic traffic.
  • Though local and mobile are often synonymous when discussing mobile search marketing, it’s clear that users of AT&T’s mobile search engine have interests beyond their own city limits.
    Marketers should stop thinking of the two as one and the same and consider the mobile searcher as having more diverse needs.

Any keyword tool on the market is only able to give relative search volume, and the Ad Intelligence tool provides impression data for an engine with roughly 13% market share, so the following
should be taken for what it’s worth.

Nonetheless, comparing web search volume to mobile web search volume gives marketers some insight into the difference between computer-based queries and mobile search queries.


QuerySept 07 Mobile SearchesSept 07 Web Searches MSN% of MSN Web Searches
cool talk84851386148.55%
myspace mobile33232416137.54%
spider-man 343374320100.39%
Scarlet johansson2970301098.67%
ring tones118473288836.02%

Of the queries with more than 30% of the MSN monthly search volume, nearly 70% are related to mobile content.

Of those that aren’t mobile-specific, the queries “scarlet johansson” and “spider-man 3” are unconventional spellings of popular entertainment keywords with almost as much volume as MSN Live

The query “sports” has almost 65% of the total web search volume with nearly 215k impressions in the month of September on AT&T’s proprietary engine alone. With other non-mobile-specific
queries like “answer”, mobile search volume for the month of September nearly equals its total web search volume.

Best Practices for Marketers

  • Popular keyword tools like WordTracker and KeywordDiscovery should always be taken with a grain of salt when used to represent absolute search volume, but this is especially true with mobile
    keywords, which can differ in volume estimates by as much as 6149% from their web counterparts. When doing mobile keyword research and using web keyword tools, it is recommended that the
    researcher filter the keyword list by characteristics of mobile queries.
  • Not only can popular mobile subjects pull as much or more traffic in the mobile space as on the Web in general, but much of the traffic appears to be coming from mobile search. Optimizing a
    relevant site for high traffic keywords will likely lead to an increase in mobile search traffic.


Though not intended to be the last word on mobile keyword research, this study should provide marketers with a few best practices related to mobile search keywords that can lead to increased
visibility in the mobile engines.

In the near future I’ll be posting an overview of various methods for mobile keyword research, which should provide additional tactics for aligning mobile queries with mobile content.