This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
The Internet’s potential impact on suicide is of major public health interest as easy online access to pro-suicide information or specific suicide methods may increase suicide risk among vulnerable Internet users. Little is known, however, about users’ actual searching and browsing behaviors of online suicide-related information.
To investigate what webpages people actually clicked on after searching with suicide-related queries on a search engine and to examine what queries people used to get access to pro-suicide websites.
A retrospective observational study was done. We used a web search dataset released by America Online (AOL). The dataset was randomly sampled from all AOL subscribers’ web queries between March and May 2006 and generated by 657,000 service subscribers.
We found 5526 search queries (0.026%, 5526/21,000,000) that included the keyword "suicide". The 5526 search queries included 1586 different search terms and were generated by 1625 unique subscribers (0.25%, 1625/657,000). Of these queries, 61.38% (3392/5526) were followed by users clicking on a search result. Of these 3392 queries, 1344 (39.62%) webpages were clicked on by 930 unique users but only 1314 of those webpages were accessible during the study period. Each clicked-through webpage was classified into 11 categories. The categories of the most visited webpages were: entertainment (30.13%; 396/1314), scientific information (18.31%; 240/1314), and community resources (14.53%; 191/1314). Among the 1314 accessed webpages, we could identify only two pro-suicide websites. We found that the search terms used to access these sites included “commiting suicide with a gas oven”, “hairless goat”, “pictures of murder by strangulation”, and “photo of a severe burn”. A limitation of our study is that the database may be dated and confined to mainly English webpages.
Searching or browsing suicide-related or pro-suicide webpages was uncommon, although a small group of users did access websites that contain detailed suicide method information.
Internet usage is growing exponentially with one third of the world population having Internet access as of 2011, of which 45% were users below the age of 25. The number of Internet users from developing countries has increased from 44% in 2006 to 62% in 2011 [
Infodemiology is a new area of research described as “the science of distribution and determinants of information in an electronic medium,” [
In the area of suicide prevention, Dobson [
Search engines on the Internet are often assumed to be an efficient, if not the preferred, gateway for access to suicide-related information [
Although prior research has shown that pro-suicide information on methods of suicide are easily accessible through search engines, the findings invited the criticism that the keywords and phrases used were pre-selected by researchers and thus had poor generalizability since the studies were not grounded in the naturalistic search behavior of non-researchers, web-savvy, or suicidal people [
A web search dataset released by the AOL is publicly accessible for data analysis (see similar research in [
Each entry contained an anonymous service subscriber identity code (ID), time of submission, the rank position, and domain of the clicked-through webpage. A single search query represents one individual click on the “search” button as initiated by a user, regardless of whether the user clicked on any of the webpages in the results page(s). A unique set of one or more keywords used in a search query constitutes the search term(s). A click-through is where a webpage in the returned search results was accessed by an individual click following a single instance of a search query and consequently that webpage is referred to as a clicked-through webpage. The number of click-throughs can also be referred to as traffic.
In response to concerns about privacy, additional measures were taken to replace all possible personal identifiable information from the dataset before undertaking analysis. Each unique subscriber was assigned a random identifier number. The Human Research Ethics Committee for Nonclinical Faculties of the University of Hong Kong approved the ethical aspect of the study protocol.
To address our first research question, any search query containing the keyword
To address our second research question, web domains from the webpages in the above 11 categories that appeared to encourage suicide or promote the individual right of dying by suicide regardless of medical grounds [
We found 5526 search queries (0.026%, 5526/21,000,000) that included the keyword suicide. The 5526 search queries were generated by 1586 unique search terms and 1625 unique users (0.25%, 1625/657,000). The three most frequently used search terms were “suicide” (9.79%, 541/5526), “suicide girls” (5.34%, 295/5526), and “how to commit suicide” (2.17%, 120/5526). Of the 5526 search queries, 3392 (61.38% of 5526) clicked through to a resulting webpage.
Of 3392 search queries with a click-through, 1344 unique webpages were clicked through by 930 unique users using 908 unique search terms. Of the 1344 webpages, 30 webpages were found to be inaccessible during the study period (October to November 2011) and returned error messages such as “page could not be found”, “forbidden access”, or “unable to display page” resulting in 1314 accessible clicked-through webpages. The three most frequently used search terms were also “suicide” (7.61%, 257/3392), “suicide girls” (4.95%, 168/3392), and “how to commit suicide” (2.74%, 93/3392).
Each clicked-through webpage was classified into 11 categories:
Entertainment (eg, writing, pictures, music about suicide): Webpages of original writing, pictures, music, or other content that may contain the keyword “suicide” for the purpose of amusement but not necessarily related to the conscious act of killing oneself.
Scientific Information: Webpages containing general educational materials, suicide statistics, and research studies.
Community Resources: Peer support or non-governmental organization webpages related to suicide prevention.
Specific Case/Incident: Webpages related to specific cases of highly publicized suicides.
Specific Means: Webpages that explicitly provide details on the methods of suicide, except drug overdose (see below).
Overdose and Suicide: There was a noticeable number of webpages that specifically addressed drug overdose or drug-related suicide information, and hence, a separate category from the Specific Means category was created.
Religion and Suicide: Webpages with information on suicide from a religious perspective.
Physician-Assisted Suicide: Webpages that contained information on euthanasia and physician-assisted suicide.
Discussion Forum: Any webpages with an interface primarily for discussion between users.
Families and Friends of Suicide/Attempt: Webpages with information about funeral arrangements and resources for family/friends of suicides.
Afterlife: Webpages with information and opinions related to “life after death” from suicide as well as life after surviving a suicide attempt. This category was singled out from the “religion and suicide” category because users used search terms that specifically included “afterlife”, and these searches may or may not involve religiosity issues.
Among the 1314 webpages, the proportions of the webpage categories are as follows: entertainment (writings, pictures, music about suicide) (30.14%, 396/1314), scientific information (18.26%, 240/1314), community resources (14.54%, 191/1314), specific case/incident (11.04%, 145/1314), specific means (10.65%, 140/1314), overdose and suicide (4.72%, 62/1314), religion and suicide (4.11%, 54/1314), physician-assisted suicide (2.51%, 33/1314), discussion forum (2.28%, 30/1314), families and friends of suicide/attempt (1.60%, 21/1314), and afterlife (0.91%, 12/1314).
From the search queries, 78.01% of the traffic (2646/3392) went to a webpage link belonging to first page of the search results, 9.17% (311/3392) of the traffic went to the second page of search results, and 10.23% (347/3392) went to the third to fifth page of search results, while only 2.59% (88/3392) trickled to the sixth page or onwards in the search results. These results are generally consistent with current search engine data where links that are higher in the search engine results page, especially on the first page of the search results, tend to get majority of the traffic [
Top three search terms and their search frequencies in each category.
Category | Search term | Frequency | No. of unique users |
1) Entertainment (eg, writings, pictures, and music about suicide) | suicide girls | 247 | 23 |
|
suicidegirls | 90 | 11 |
|
suicidegirls.com | 46 | 8 |
|
virgin suicides | 46 | 2 |
2) Scientific information | suicide | 69 | 49 |
|
free essays on adolescent depression and suicide risks | 28 | 1 |
|
adolescent depression and suicide risks | 25 | 1 |
3) Community resources | suicide | 130 | 89 |
|
suicide help | 29 | 12 |
|
suicide prevention | 16 | 12 |
|
teen suicide | 16 | 11 |
4) Specific case/incident | famous suicide 1970 richard cunningham | 28 | 1 |
|
vachel lindsay suicide | 16 | 1 |
|
christopher lee anderson suicide | 15 | 1 |
5) Specific means | how to commit suicide | 70 | 21 |
|
ways to commit suicide | 36 | 8 |
|
suicide | 24 | 18 |
6) Overdose and suicide | how to commit suicide | 17 | 13 |
|
cymbalta and suicide | 14 | 1 |
|
suicide by overdose of kadian | 8 | 1 |
|
suicide by overdosing | 8 | 1 |
7) Religion and suicide | how to commit suicide | 13 | 10 |
|
sermons preach for person who committed suicide | 13 | 1 |
|
suicide in the bible | 11 | 3 |
8) Physician-assisted suicide | assisted suicide | 46 | 19 |
|
physician assisted suicide | 13 | 8 |
|
should america allow suicide euthanasia | 8 | 1 |
9) Discussion forum | how to commit suicide | 11 | 10 |
|
Xxsuicide’s xanga site | 10 | 1 |
|
suicideroadmap - myspace blog | 9 | 8 |
10) Families and friends of suicide/attempt | suicide memorials | 17 | 1 |
|
survivors of suicide | 8 | 3 |
|
suicidememorialwall.com | 4 | 1 |
11) Afterlife | suicide and the after life | 9 | 1 |
|
survivors of suicide | 5 | 1 |
|
after life suicide | 4 | 1 |
When suicide was used as a single keyword search term (number of search queries = 256; 19.48%; 256/1314), users mostly clicked through to webpages that belong to the “community resources” (130 search queries), “scientific information” (69 search queries), and “specific means” categories (24 search queries). In contrast, when the search term “how to commit suicide” was used, the users most commonly clicked through to webpages in the “specific means” (70 queries), “overdose and suicide” (17 queries), and “religion and suicide” categories (13 queries).
The search terms most commonly used to access webpages in the “entertainment” category included the keywords in the names of a specific adult website and a novel cum movie. The “specific case/incident” category top search terms involved the keywords in the names of two famous individuals and a teenager who died from suicide; of note, each of these three search queries were contributed by one single user. Webpages in the “specific means” category were most commonly clicked through when users used the search terms “how to commit suicide”, “ways to commit suicide”, and “suicide”.
Suicide page on Wikipedia.
Top three most clicked-through webpages of each category, their click-through frequencies, and the number of unique users.
Category | Webpage | Frequency | No. of unique users |
1) Entertainment (writings, pictures, music, about suicide) | http://www.suicidegirls.com | 153 | 76 |
|
http://suicidegirls.com | 78 | 59 |
|
http://en.wikipedia.org | 21 | 18 |
|
http://www.suicidedoors.com | 21 | 16 |
2) Scientific information | http://www.cdc.gov | 43 | 35 |
|
http://www.psycom.net | 25 | 23 |
|
http://www.ncbi.nlm.nih.gov | 15 | 13 |
3) Community resources | http://www.metanoia.org | 86 | 81 |
|
http://kidshealth.org | 20 | 17 |
|
http://www.afsp.org | 16 | 14 |
|
http://www.focusas.com | 16 | 10 |
4) Specific case/incident | http://en.wikipedia.org | 18 | 13 |
|
http://news.bbc.co.uk | 7 | 5 |
|
http://profile.myspace.com | 7 | 6 |
5) Specific means | http://www.satanservice.org | 49 | 39 |
|
http://www.mouchette.org | 21 | 18 |
|
http://extremeriver.org | 14 | 13 |
6) Overdose and suicide | http://www.a1b2c3.com | 31 | 23 |
|
http://www.ncbi.nlm.nih.gov | 7 | 5 |
|
http://www.everything2.com | 5 | 5 |
7) Religion and suicide | http://www.religioustolerance.org | 16 | 13 |
|
http://www.believers.org | 13 | 11 |
|
http://en.wikipedia.org | 7 | 5 |
8) Physician-assisted suicide | http://www.assistedsuicide.org | 13 | 11 |
|
http://www.religioustolerance.org | 10 | 9 |
|
http://www.amsa.org | 5 | 5 |
|
http://www.euthanasia.com | 5 | 5 |
9) Discussion forum | http://www.xanga.com | 14 | 2 |
|
http://www.zenhex.com | 13 | 12 |
|
http://lifeflame.6.forumer.com | 5 | 5 |
10) Families and friends of suicide/attempt | http://www.suicidememorialwall.com | 9 | 4 |
|
http://www.survivorsofsuicide.com | 7 | 6 |
|
http://www.parentsofsuicide.com | 5 | 4 |
11) Afterlife | http://www.near-death.com | 5 | 3 |
|
http://samvak.tripod.com | 3 | 1 |
|
http://www.survivorsofsuicide.com | 3 | 2 |
Among the 1314 accessed webpages, we could identify only two clicked-through websites that were considered pro-suicide. They are “http://www.suicidemethods.net” and “http://www.churchofeuthanasia.org”. The two websites not only provide information on methods of suicide but also portray a positive attitude towards the choice of using suicide as a way to ease pain.
Search terms used to access the two pro-suicide websites.
Website | Search term | Frequency | No. of unique users |
http://www.churchofeuthanasia.org | committing suicide with a gas oven | 27 | 1 |
|
hairless goat | 22 | 1 |
|
how to kill yourself | 18 | 3 |
|
methods to commit suicide | 11 | 1 |
|
euthanasia | 10 | 2 |
|
best way to commit suicide | 5 | 1 |
|
butcher pig | 4 | 1 |
|
committing suicide | 4 | 1 |
|
how to die painlessly | 3 | 1 |
|
the church of euthanasia | 1 | 1 |
|
butchering the human carcass for human consumption | 1 | 1 |
|
dog sex kkh | 1 | 1 |
http://www.suicidemethods.net | pictures of murder by strangulation | 26 | 1 |
|
photo of a severe burns | 22 | 1 |
|
pro choice suicide | 7 | 1 |
|
suicidal websites | 6 | 1 |
|
suicide murder pics | 6 | 1 |
|
suicide pics | 6 | 1 |
|
suicide methods | 5 | 2 |
|
asphyxia pics | 4 | 1 |
|
gory pictures | 3 | 3 |
|
gory photos | 2 | 2 |
|
homicide pictures | 2 | 2 |
|
pictures of suicides | 2 | 2 |
|
suicide | 2 | 1 |
|
bloody suicide jump pictures | 1 | 1 |
|
cut wrist pictures | 1 | 1 |
|
gory autopsy pictures | 1 | 1 |
|
hanging suicide | 1 | 1 |
|
shotgun wounds | 1 | 1 |
|
suicide attempt | 1 | 1 |
|
suicide and hell | 1 | 1 |
|
suicide photos | 1 | 1 |
This is a retrospective observational study aimed to investigate the naturalistic web searching behavior on a search engine by online users looking for content using suicide-related keywords. Specifically, we examined users’ search terms with the keyword “suicide
Despite being a older dataset, especially in the context of the Internet where content and usage behavior is constantly changing, the findings still have relevance because, unlike prior studies that took a more computational perspective to suicide-related Internet searches, we used a naturalistic approach by looking at the users’ actual search behaviors. Rather than merely looking at what is available for our perusal on the Internet, which is subject to constant change, we examined the subsequent behavior with returned results, which seems relatively consistent even over time [
Utilizing this type of dataset, however, means that although users’ search result click-throughs were logged, we do not have a record of the search results webpages they could see and potentially access. Also, users’ behavior after accessing the webpages cannot be examined. Nevertheless, it is possible to identify search trends, patterns, and popular click paths that allow us to gain insight into users’ searching goals and intent as well as pages that had high visibility based on the keywords used in the search queries.
Presumably there are many existing websites that contain information about suicide, making the Internet’s role in suicides a point of concern. Our findings highlight three important findings on Internet users’ naturalistic web search behaviors that may challenge the underlying assumptions behind past studies suggesting that the Internet is intrinsically harmful because pro-suicide information can be found easily using suicide-related search terms. First, despite previous studies finding that pro-suicide and how-to-commit-suicide websites are easily accessible through search engines [
In traditional media, newspapers, radio, and television were the major sources for information and their consumers were perceived as passive receivers of information. Information portrayed in traditional media was mainly contributed by professional journalists, and hence, regulation of that information was relatively straightforward. In terms of suicide information in traditional media, since information about suicide cases are presented and received passively by the masses, media reporting guidelines on suicide information developed for professionals has been considered as a core prevention strategy [
In the Web 2.0 era, the new media is interactive. People become proactive consumers and producers of information. Not only can people search for information and contribute, messages and information can also be widely and easily shared among friends via social media and networks. The ease and low cost of creating and copying content on the Internet contributed to its exponential growth and now, the amount of information that can be found on the Internet can be considered as almost infinite. In 2004, Google’s whole data storage was approximately five petabytes, which is the quantity of data a thousand times larger than the Library of Congress’s print collection [
Internet users use the Internet for various reasons. Segev and Ahituv conducted a cross-national analysis of popular search queries used in Google and Yahoo! over a 24-month period from January 2004 to December 2005 and found that there are cultural differences related to Internet usage [
Our findings have provided a glimpse into the information consumption preferences of online users and corroborates prior studies on people’s Internet usage. Even for a domain-specific topic, suicide, our study generated consistent findings that most search engine users in the United States used suicide-related queries to search for websites that belong to the entertainment category.
Although pro-suicide websites and websites that contain detailed information on suicide methods can be accessed easily, the majority of search engine users, at least in the United States, did not access them using suicide-related queries. A number of keywords used to access pro-suicide webpages were instead related to violent or bloody pictures including “gory pictures”, “gory photos”, and “homicide pictures”. With these queries, we suspect that those who accessed the websites might not have an intent to die. Westerlund [
In-depth interviews with individuals with nearly lethal suicidal behavior may help to understand this phenomenon. In Biddle et al’s study [
In our study, we found that a small group of individuals intentionally searched for information about ways to complete suicide and an even smaller group of individuals were eager to find discussion forums using suicide-related keywords. A few individuals used queries like “suicide by overdose of XX” or “suicide by XX” to get access to webpages that provide information on prescribed drugs and suicide. In contrast, websites that include information on suicide means were accessed by larger numbers of unique users. One individual used a specific discussion forum address as a query (“xxsuicide’s xanga site” as shown in the table) to get access to that particular webpage 10 times.
There is a lack of research on suicide pacts arranged online, and so it is difficult to predict the major consequence of a large number of discussants getting access to this sort of discussion forum. On one hand, there have been some documented instances of completed suicide pacts in Japan [
Current efforts to reduce the potential negative impact of the Internet on suicide include remedial actions like removing and blocking entries of harmful websites by using Internet filters, constant monitoring of potential harmful websites by “web-based police officers” as implemented in Japan and Korea, and displaying helpline resources information while users key in suicide-related queries. These strategies are in place in an attempt to either shield suicidal and vulnerable individuals from suggestive material or abort their suicidal wishes by diverting their attention to help resources. There are several challenges that these strategies face. First, these strategies do not observe the way that information is created, found, and accessed online. Accordingly then, restricting or removing potentially harmful websites is an arduous if not impossible task. Furthermore, cyber regulations are extremely difficult to enforce since the Internet is not clearly under any jurisdiction, and identification of a physical user can be elusive.
Instead, it is important to further investigate the natural online behavior especially of vulnerable individuals and integrate that information with the way that search engines provide information. With the Internet inundated with information, the major search engines, ie, Google, Bing, and Yahoo!, have complex algorithms to return relevant search results to users. Consistent with the uses and gratification theory, relevance in a search engine is defined by variables such as user popularity, credible web domains, and high-quality content. As such, well-made and informative websites will tend to rank higher in search results and also be visited more often. Consequently, instead of the labor-intensive strategy of finding, sorting, and removing potentially harmful websites, more resources and effort should be spent on developing high-quality, informative, interactive, and user-friendly websites that maximize the likelihood of ranking highly in search results and therefore, being found and accessed.
Also, more recent studies find that the Internet provides more positive than negative effects on the vulnerable, especially in enhancing the social support of isolated individuals though social networking sites [
Our research presented the naturalistic search behavior of search engine subscribers using any search terms with the keyword “suicide”. Currently we know that cybersuicide or Internet suicide pacts represent a small fraction of overall suicides. While those attempting suicide have searched for suicide-related information on the Internet before completing the act, we also know that the Internet can provide a unique platform to reach those individuals previously inaccessible, and web-based psychotherapies are producing promising results in helping depressed individuals, with the acknowledgment of the potential digital divide phenomenon. As such, it can be concluded that we should neither underestimate nor overestimate the potential negative impact of the Internet on suicide.
However, the threshold between offline and online activities is quickly receding and in the near future will soon disappear. It is uncertain how much of today’s searching, and even learning, behavior will remain unchanged. There are plenty of daily examples that show how much of our learning behavior has transformed. “
We thank the anonymous reviewers for their helpful suggestions. The publication of this article was partly supported by the General Research Fund (HKU 756211H granted to PW).
None declared.