Previous studies have shown that individuals may search for suicide-related terms on the internet prior to an attempt. A recent study showed that in the 60 days prior to a suicide attempt, individuals had searched for terms expressing suicidal ideation and suicide means but also relevant keywords not directly associated with crisis (eg, feeling empty, divorce, and alcohol use) [1]. In addition, studies have found that the increases in the volume of internet searches for suicide-related terms predicted subsequent increases in national suicide rates [2-4]. Thus, targeting internet search engines to intervene in such searches, and connect individuals to help, may be an important avenue for intervention.

Internet search ads can be used to reach out to individuals contemplating suicide, or in crisis, by identifying what keywords people are likely to search, presenting an advertisement on the search page when the keywords are searched, and linking to a landing page with appropriate resources and help-seeking information. The reach and effectiveness of these campaigns can be assessed by measuring the impressions (how many times the advertisement is shown), clicks or click rate (clicks on the advertisement), and conversions (specific behaviors performed on the website) [5]. Note that the party placing the advertisements must determine what behaviors they would like visitors to do on the website, counting these behaviors as conversions.

One reason to use advertisements over organic (or nonadvertisement) searches is that when using organic search, it is not guaranteed that a link will appear first on the search result page. A study found that on average, a link on the search page is twice as likely to be clicked on versus the link directly below it [6], suggesting that individuals are most likely to click and engage with the search result presented first. Further, organic search relies on similarities of the keyword and the page itself; thus, a person who is searching for terms pertaining to “loneliness” may not be shown a link to a page with suicide help. Finally, internet searches are a part of daily life, with 93% of browsing sessions starting on the search page [7]; thus, individuals may be more likely to engage with help if the process by which they access help is a part of their day-to-day behavior.

In some search engines, a hotline number will appear in a prominently displayed banner at the top of the search results if certain suicide-related terms are searched. Reports from the United States have shown that in certain instances, the banners have increased calls to the hotline by 10% [8]. However, there is no public information on what keywords will trigger the banner and whether past search history increases or decreases the likelihood of this banner appearing. Further, a Google spokesperson has said and past research has shown that not all relevant keywords will trigger this hotline banner [8,9]. Research has also shown that not everyone in a suicidal crisis would like to call a hotline, suggesting a need for more diverse offerings [10]. Further, a phenomenon called “banner blindness” is often observed, in which eye-tracking data suggest that individuals will tend to ignore elements on a page in the form of a banner [11].

The overall objective of this research was to investigate patterns of engagement with advertisements and landing pages—that is, the first page shown after a link is clicked—designed for those in crisis and contemplating suicide, which was examined in 2 studies. The first study aimed to assess whether the presence of a Google Ads campaign promoting a crisis line would encourage connection with the crisis service. The aim of the second study was to conduct a 2-arm trial investigating the additional benefits of age tailoring while extending the campaign to include individuals at the precrisis or contemplation phase and assess the different engagement levels of different types of search keywords. Both studies are reported in CONSORT (Consolidated Standards of Reporting Trials)–compliant formats.

Within the 16-day duration of the ad campaign, the advertisement was shown 16,505 times and was clicked 664 times, yielding a 4.02% click rate. From those who clicked the ad (n=664), there were 101 calls to the hotline, yielding a 15.21% conversion rate with an average cost of US $13.57 per helpline call. The campaign had an overall conversion rate (conversion/impressions) of 0.61% (101/16,505).

To investigate how our campaign performed against industry standards, we conducted a rate comparison test to compare our campaign data versus publicly available industry data. The industry data were drawn from approximately 985,804,416 impressions, 31,250,000 clicks, and 1,171,875 conversions, with a 3.17% click rate and 3.75% conversion rate, collected from the Google Ads agency [17]. The analysis suggested that our campaign overall performed better than the industry standard (IRD 0.004931, 95% CI 0.004405-0.0005457; P<.001; ie, 0.49 percentage points higher than the industry standard;).

In study 1, we investigated the efficacy of a Google Ads intervention in encouraging help seeking by calling a helpline by running a campaign across the United States. The results revealed high click-through rates and conversion rates when compared with the industry average across all sectors (click rate: 4.02% vs 3.17% industry average; conversion rate: 15.21% vs 3.75% industry average; overall conversion rate: 0.61% vs. 0.12% industry average), having a 5.08 times greater total conversion rate than the industry average.

As information about when the banner was or was not triggered is limited, it is not possible to provide a further breakdown of the ad engagement based on whether the banner was triggered. Nevertheless, as good engagement was observed with an ad that promoted the national hotline, 2 general conclusions can be drawn. First, if the hotline banner was triggered, this suggests that some individuals skipped over the banner but did click on the ad to call the hotline. Second, if the banner was not triggered, then the ad led to an individual seeking help that otherwise would not have. Overall, the data provide evidence for the utility of an intervention using Google Ads.

In study 2, we developed a Google Ads campaign designed to be helpful for individuals who may be contemplating suicide but are not in immediate crisis [18]. In this context, we use the term contemplation phase to capture any stages that may lead to a suicide attempt but prior to immediate crisis. Simply providing a link to the hotline may not be the most appropriate response in the contemplation stage, and therefore, we undertook a co-design process to expand what was offered on the landing page. Further, given that the keywords searched may represent different cognitive states, we investigated whether different categories of keywords led to different levels of engagement. Finally, we used Google Ads’ targeting features and investigated whether individuals show higher engagement with landing pages tailored to specific age groups compared to a general all-ages landing page.

The advertisements ran from March 2 to 21, 2022, in Australia with a total of 120,881 impressions, 6227 clicks (a 5.15% click-through rate), and 1419 conversions, with a conversion rate of 22.79% (1419/6227) and a total conversion rate of 1.17% (1419/120,881). The breakdown per keyword group is shown in Table 1. Note that due to the low number of the clicks related to the means-related keywords (n=4, with no conversions), these keywords were excluded from the analysis.

We investigated whether the study-2 total conversion rate was higher than those of study 1 or the industry standard. The results revealed that the study-2 campaign had a higher total conversion rate than the study-1 total conversion rate (IRD 0.005619, 95% CI 0.004984-0.007436; P<.001) and industry standard total conversion rate (IRD 0.01055, 95% CI 0.01036-0.01191; P<.001), with the study-2 total conversion rate being 1.91 times greater than the total conversion rate in study 1 and 9.75 times greater than the industry standard.

We then investigated whether the click-through rate differed as a function of keywords searched. The results revealed that the click rates for individuals searching low risk keywords, high risk keywords, and help seeking keywords were all significantly different from one another, in which individuals searching for help seeking keywords had the highest click rate, followed by individuals searching for low risk keywords, followed by individuals searching for high risk keywords. Results are shown in Figure 4A (low risk vs high risk: IRD 0.01608, 95% CI 0.00836-0.04194; P<.001; low risk vs help seeking: IRD 0.01223, 95% CI 0.001711-0.00734; P<.001; help seeking vs high risk: IRD 0.02831, 95% CI 0.01898-0.03764; P<.001).

We then investigated whether conversion rates differed by keyword type. The rate comparison analysis found a significant difference between low risk and high risk keywords (IRD 0.6972, 95% CI 0.00929-0.18058; P=.009), in which there was a higher conversion rate for people who searched for high risk keywords, and a significant difference between low risk and help seeking keywords (IRD 0.07877, 95% CI 0.03793-0.1196; P<.001), in which there was a higher conversion rate for individuals who are seeking help. However, there was not enough evidence to suggest that there was a difference in conversion rates between high risk and help seeking keywords (IRD 0.01617, 95% CI 0.09235-0.12469; P=.77). A graphical representation of these results can be found in Figure 4B.

Finally, we investigated whether age tailoring had a significant effect on conversions. The analysis revealed a significant difference between tailored and general landing pages for low risk keywords, in which general landing pages had a higher conversion rate than tailored landing pages (IRD 0.03635, 95% CI 0.06116-0.01153; P=.004). However, there was no significant difference for the general versus tailored landing pages for high risk (IRD 0.08295, 95% CI –0.11938 to 0.28528; P=.42) and help seeking keywords (IRD 0.02078, 95% CI 0.11145-0.0699; P=.65; see Figure 4B).

In this study, we codeveloped a series of search advertisement campaigns targeting individuals contemplating suicide and evaluated them for levels of reach and engagement. As per study 1, we observed high engagement relative to the industry standards. Within 21 days, we were able to reach individuals searching for suicide-related terms 120,881 times, with an average cost of US $13 each time a person engaged with a behavior designed to help them. Note that cost per conversion does not differ from one conversion to another; thus, it is the role of the Google Ads client—in this case, the research team—to set conversions that are meaningful.

Given that the advertisement was designed for individuals contemplating suicide, it is consistent with the finding that there was a significantly higher click-through rate for individuals searching for low risk than high risk keywords. The help seeking keywords yielded the highest click-through rate, which is consistent with the idea that individuals responded to the ad that featured words communicating explicitly that help could be found. This is consistent with the notion that the better the advertisement matches the search term, the more likely an individual is to click the advertisement. However, the analysis found a higher conversion rate for high risk compared to low risk keywords, which may stem from the fact that not all individuals searching for low risk keywords are searching for or need immediate help.

One finding was that there were markedly lower impressions, clicks, and conversions for individuals searching for means specific keywords, suggesting that overall, there were fewer people who were searching for means specific keywords. Given the low number of impressions and clicks, we were unlikely to see any conversions.

An unexpected finding was that there was a higher conversion rate for general landing pages compared to tailored landing pages for low risk keywords, suggesting that our tailoring was not effective in application and rather reduced engagement. This may indicate that despite there being clear, mutually exclusive preferences indicated by different age groups in the co-design process [10], these preferences may have limited generalizability or how we operationalized these preferences may be limited despite having approval from the co-design team. Other types of tailoring have yet to be explored.

Across 2 studies, we investigated whether using Google Ads would allow us to reach individual searching for suicide-related terms for both crisis, in which we presented a suicide helpline, and contemplation, in which we presented a landing page offering help beyond presenting a helpline, such as hearing lived experience stories and calming exercises. Across both studies, we have observed exceptionally high click-through and conversion rates compared to the industry standard, with study-2 total conversion rate being 9.75 times higher than the industry standard. Overall, the data show that individuals are engaging well with both the advertisement and the pages themselves.

Given that for much of the time, it is likely that the hotline banners were present in both studies, future studies should investigate why individuals were still engaging with a search result if the helpline number was present on the search page. One possibility in study 1 is that within the few lines of text, the advertisement was able to provide more information on the hotline compared to the banner—that is, the hotline is anonymous and available 24/7. Lived experience advisors noted that it would be helpful to include descriptions on what it is like to call a hotline or engage in other services to help remove apprehension and promote help seeking [10]. In the second study, the help seeking keywords yielded the highest click-through rate—keywords that are likely to trigger a hotline banner given they explicitly indicate help seeking for suicide—further suggesting that there are individuals who are seeking help for suicide who do not want to immediately call a hotline, which is consistent with findings from the co-design study [10].

Future studies should investigate whether other advertising services can be used for suicide prevention. Although Google has a majority market share [12] for search engine advertisements, the advertising industry spans across multiple platforms. For example, there is data collection and advertising on social media (such as Twitter, Instagram, and TikTok) and streaming sites (eg, YouTube) and on advertising banners on websites (eg, Google AdSense). Future studies should investigate whether the findings from this study generalize to other platforms and whether using more than one platform for data collection and advertising increases reach and promotes help seeking. We propose, given the wealth of research showing web-based behaviors reflecting suicidality outside of search (eg, social media [19]), that we should see similarly fruitful findings outside of search engine advertisements.

Future work should also investigate how to integrate digital advertising such as Google Ads into routine practice. For example, given that individuals may discover existing resources through internet searches, partnerships with local health providers could help to place ads to link individuals to their local area’s most validated and well-resourced services. To implement this into routine practice, we would need to ascertain (1) the cost and cost-effectiveness of integrating digital advertising into existing health systems, (2) what services could be highlight on the landing pages, and (3) the funding requirements to sustain this approach.

Our investigation has several limitations. One limitation is that we did not record data on whether persons clicking on the ads were experiencing suicidal ideation or had engaged in suicidal behaviors, nor did we assess change in these outcomes. Future studies could investigate whether help-promoting web pages, such as those designed for this study, increase the number of individuals who seek help and reduce the suicide rate within a specific geographic region using Google Ads’ geographic targeting. Another limitation is that we do not know when the helpline was also triggered. Thus, we are unable to investigate how individuals behave when there is both the helpline and the advertisement, compared with just the advertisement.

Another limitation is that in study 2, we only included individuals whose ages could be ascertained by Google as 18 years or older. This was to ensure that the general versus age-tailored groups were comparable. However, we cannot determine what proportion of the population were excluded due to their ages being undetermined from past browsing history. Thus, future studies should also include individuals whose ages cannot be determined, where the analysis permits. Another key limitation in study 2 is that none of the advertisements explicitly used the word “suicide.” Although this is the result of the rigorous co-design process [10], there were advisors who suggested that in some settings, explicit use of the word “suicide” may lead to higher engagement. Thus, future studies should compare engagement with the campaign when advertisements use and do not use explicit suicide wording.

Although the true effect of this intervention remains to be seen, the engagement metrics suggest that using internet ads may be useful in reaching out to individuals searching for suicide-related terms as a service delivery platform. Beyond suicidality, using internet ads may be helpful to reach out to individuals across a wide range of different health applications. Given that internet search is perhaps one of the most primary methods of information seeking and the ranking on search page can shape behavior, health-based organizations can use internet ads to ensure that what is presented is not only what is the most engaging but the most authoritative and valid.