Do Search Engine Helpline Notices Aid in Preventing Suicide? Analysis of Archival Data

Qijin Cheng, Elad Yom-TovJournal of Medical Internet Research 2019 · OpenReview

Abstract

Background: Search engines display helpline notices when people query for suicide-related information. Objective: In this study, we aimed to examine if these notices and other information displayed in response to suicide-related queries are correlated with subsequent searches for suicide prevention rather than harmful information. Methods: Anonymous suicide-related searches made on Bing and Google in the United States, the United Kingdom, Hong Kong, and Taiwan in a span of 10 months were extracted. Descriptive analyses and regression models were fit to the data to assess the correlation with observed behaviors. Results: Display of helpline notices was not associated with an observed change in the likelihood of or future suicide searches (P=.42). No statistically significant differences were observed in the likelihood of people making future suicide queries (both generally and specific types of suicide queries) when comparing search engines in locations that display helpline notices versus ones that do not. Pages with higher rank, being neutral to suicide, and those shown among more antisuicide pages were more likely to be clicked on. Having more antisuicide Web pages displayed was the only factor associated with further searches for suicide prevention information (hazard=1.18, P=.002). Conclusions: Helpline notices are not associated with harm. If they cause positive change in search behavior, it is small. This is possibly because of the variability in intent of users seeking suicide-related information. Nonetheless, helpline notice should be displayed, but more efforts should be made to improve the visibility and ranking of suicide prevention Web pages.

Introduction

For over a decade, major search engines have utilized "helpline notices"—prominent boxes containing contact information for crisis centers—as a primary intervention for users searching for suicide-related terms. While these notices are intended to divert vulnerable individuals from harmful content toward professional support, their actual impact on user behavior has remained largely unquantified. This study addresses a critical gap in public health research: Do these notices work, or are they overshadowed by the search results themselves?

Methodology

The researchers conducted a retrospective analysis of archival search data from November 2016 to August 2017 across two primary datasets:

  1. Bing Search Logs: Anonymous data from users in the US, UK, Hong Kong (HK), and Taiwan, providing detailed insights into click behavior and sequential search patterns.
  2. Browser Toolbar Logs: Anonymized data from a widely distributed add-on, allowing for cross-engine comparisons between Bing and Google.

The study employed logistic regression to predict click likelihood based on page characteristics (rank, bias, presence of notice) and hazard models to determine if exposure to specific information predicted future searches for suicide methods or prevention resources within a 7-day window. Six trained coders manually classified hundreds of frequently appearing webpages into categories ranging from "antisuicide" to "prosuicide."

Results and Analysis

The analysis revealed that the presence of a helpline notice did not significantly alter the likelihood of future suicide-related searches ($P=.42$). Furthermore, 83.7% of users who clicked the helpline notice also engaged with other links on the page, suggesting the notice serves as a supplement rather than a definitive diversion.

Distribution of page types (top) and click-through rates relative to display frequency (bottom) across the four study regions.

As shown in the figure above, the content landscape varies significantly by region. Notably, users in Taiwan were exposed to a higher proportion of prosuicide content (13.8%) compared to other regions. The bottom chart highlights that "antisuicide" and "neutral" pages generally receive positive click-to-display ratios, but "prosuicide" content in Hong Kong shows a disproportionately high click rate, indicating a high level of user engagement with harmful material in that context.

The most significant finding regarding behavior change was that the volume of antisuicide content was the primary driver of help-seeking. Having more antisuicide webpages displayed was the only factor significantly associated with an increase in subsequent searches for prevention information (Hazard Ratio = 1.18, $P=.002$).

Discussion

The study provides a "no harm" verdict: helpline notices do not cause iatrogenic harm or trigger users to seek method-related information. However, their limited statistical impact suggests that a "one-size-fits-all" notice may struggle to address the high variability in user intent—ranging from academic research to acute crisis.

The findings underscore that ranking is king. A page’s position and its surrounding context (the "social norm" effect of seeing multiple prevention resources) are the strongest predictors of user clicks.

Conclusion and Recommendations

While helpline notices should remain a standard feature, they are not a panacea. Search providers should move toward algorithmic interventions that proactively improve the visibility and ranking of high-quality, antisuicide webpages. Ensuring that prevention-oriented results dominate the first page is likely more effective than any standalone notice in guiding vulnerable users toward help. Future research should explore dynamic notices that adapt to the specific sentiment or urgency of a query.

Reference: J Med Internet Res 2019;21(3):e12235 doi:10.2196/12235

Summary updated

Abstract

Background: Search engines display helpline notices when people query for suicide-related information.Objective: In this study, we aimed to examine if these notices and other information displayed in response to suicide-related queries are correlated with subsequent searches for suicide prevention rather than harmful information.Methods: Anonymous suicide-related searches made on Bing and Google in the United States, the United Kingdom, Hong Kong, and Taiwan in a span of 10 months were extracted. Descriptive analyses and regression models were fit to the data to assess the correlation with observed behaviors.Results: Display of helpline notices was not associated with an observed change in the likelihood of or future suicide searches (P=.42). No statistically significant differences were observed in the likelihood of people making future suicide queries (both generally and specific types of suicide queries) when comparing search engines in locations that display helpline notices versus ones that do not. Pages with higher rank, being neutral to suicide, and those shown among more antisuicide pages were more likely to be clicked on. Having more antisuicide Web pages displayed was the only factor associated with further searches for suicide prevention information (hazard=1.18, P=.002).Conclusions: Helpline notices are not associated with harm. If they cause positive change in search behavior, it is small. This is possibly because of the variability in intent of users seeking suicide-related information. Nonetheless, helpline notice should be displayed, but more efforts should be made to improve the visibility and ranking of suicide prevention Web pages.J Med Internet Res 2019;21(3):e12235doi:10.2196/12235
BibTeX
@article{cheng2019do,
  author = {Qijin Cheng and Elad Yom-Tov},
  title = {{Do Search Engine Helpline Notices Aid in Preventing Suicide? Analysis of Archival Data}},
  journal = {Journal of Medical Internet Research},
  year = {2019},
  url = {https://openreview.net/forum?id=UyUt38M5dl},
}
Figure