In the following sections, we describe the elements of the 11 games. However, first, it is important to note how little information about the games themselves was readily available in published articles. Authors often devoted little space to describing how the games looked and how they were played. Considering that playing games is a visual activity, the inclusion of images could have been a valuable way of providing this information. However, with some noteworthy exceptions [27,30,35,41], this was rarely performed; most articles did not include any form of visual aid to explain the game. Some articles included links to demos or webpages where the content was said to be available; however, in >1 case, the links were broken, or the page had already ceased to exist. Owing to the rapid nature of this review, no author and game developers were contacted during the process.

Regarding the age of the games themselves, we identified a clear bimodal distribution; 3 games were created in the wake of the digital revolution ≥30 years ago, whereas the remaining group was developed more recently. Earlier games were graphically simpler but in no way less informative. We found it striking that from very early on, video games were seen as valuable tools to promote learning and attitude change.

The interventions used different platforms to host the games. Considering the time span across which the games were developed, it is not surprising that the most commonly used platform was that with the longest history, the PC. The changes in the games reflect the technological evolution that has affected this platform. Although popular, tablets were only mentioned in one of the games [33]. As might be expected, when mobile phones became widespread, the idea of using smartphones as platforms for serious gaming became an option. Of the 4 reviewed games for mobile phones, 2 (50%) were exclusively developed for the iPhone [35,37], whereas the other 2 (50%) did not specify which operating system they worked with [38,42]. These later games used mobile connectivity to increase participant engagement through push notifications and messages [35,36]. However, it is noteworthy that most interventions did not appear to use patches and updates to freshen their content, even when they were constantly connected to the internet.

A decision that is implicit in platform selection is accessibility. Most interventions aspire to be easily scalable. To achieve this, they must operate on an already popular and ubiquitous platform and use relatively little graphic processing power. We do not have access to the system requirements for any of the listed games. However, based on their description, it appears most operated on or below the considered average computing power at the time of their release.

With regard to game difficulty, all interventions can be placed on a continuum. On one end, there are games that provide an easy experience; that is, the game is seen as a vehicle through which information can be provided in an entertaining fashion. On the other end, we have games that provide a very demanding experience. The best example of this in the games we identified is VODO. The developers included a section of the game in which the player needed to answer 2 questions about sexual health. If either of these answers was incorrect, the game was over, and the player had to start all over again. This type of failed outcome was a very common scenario of text adventures at the time and motivated the player to replay the game several times, learning all its intricacies. An even more challenging aspect of these questions is that the answers were not provided in-game. The player was meant to search elsewhere for information; or even better, the topic should become a conversation starter for youth to discuss with their friends, parents, or teachers. This is an example of how a severely challenging task was used to frustrate the players into action.

As could be expected, these extremes in game difficulty were not without their issues. A nonchallenging game can be boring and can inhibit engagement. At the same time, a too-challenging game can tire a proportion of the players out of the intervention. In an effort to resolve these 2 positions, some game developers opted for an increasing level of difficulty; that is, instead of having a flat, low-level difficulty throughout the experience or a series of extreme spikes every so often, these games ramp up the difficulty with every level [35,37].

There are 2 related elements to consider here: expected playthroughs and expected contact time. In other words, how many times are players expected to play the game, and for how long? As we will discuss in the following sections, repetition, rehearsal, and feedback play an essential role in learning through games; therefore, it is expected that most games were designed to keep the player engaged in >1 run. Unfortunately, there is little information about these aspects in most of the included articles. Among the games that specified an expected playing time, we found periods of <1 hour of gaming. However, it was not always clear whether these times were for single or multiple playthroughs.

Game settings can be broadly divided into 2 types: realistic and science fiction. There was a clear preference for the latter in our sample of games. Most games were situated in locations and environments to which the player could directly relate, such as hanging out with friends after school or going to a house party. The idea behind this is that a greater similarity between the simulated situation and a plausible real-life event might make it more likely that players will relate to the content and act upon what they have learned. Game developers have gone into great efforts to create content grounded in reality, where situations that the players have directly experienced are portrayed both didactically and accurately. For example, SOLVE allowed players to personalize their avatars, and The Baby Game used actual prices when they calculated the costs of raising a child.

The types of games in the realistic group were highly diverse. Table 1 presents a list of the different styles of games that were included. We want to highlight 3 features. First, the variety is notable; as can be seen, games ranged from management simulators to role-playing games. The second feature is the relative preponderance of the dating simulators. Approximately 44% (4/9) of the nonfiction games were dating simulators. However, even within this specific setting, we found different styles (eg, text adventures, 3D, and chat simulators). Finally, the role of complementary activities in each game should be mentioned. Minigames are found in several games, particularly in the form of quizzes [29,34,36,37]. This seems to have been one of the main strategies through which game developers delivered specific sexual health knowledge.

However, not all games followed a realistic route. BattleViro and Viral Combat opted instead for science fiction in the action setting [35,37]. In both games, the characters are shrunk down to a microscopic scale and are meant to protect the human body from infections by shooting down viruses, bacteria, and vectors of disease. Here, the developers tried to create a power fantasy in which the player can take control of their actions and reach a desirable healthy state. The content and settings were still related to the topic of focus (antiretroviral treatment and PrEP adherence); however, the developers avoided making direct or explicit statements about them.

We can see that there is >1 valid strategy for promoting engagement with the material. One school of thought aims to create easy-to-relate experiences, whereas the other uses a fast-paced game to empower participants into action.

An element mentioned throughout the different interventions is the contrast between creating a single or multiplayer experience. In other words, was the game designed to be played alone or with a group? Most of the games reviewed appear to have been designed for single-player use. However, it should be noted that all single-player experiences can be turned into multiplayer experiences by the players themselves. Researchers reported that, in several cases, games that were not meant to be social experiences were transformed into a group activity when a player spontaneously brought their friends or partner to play the game together and comment on it.

The decision to develop a single- or multiplayer game is affected by several factors. The first is the target population. Interventions focused on MSM were very keen on not outing their players involuntarily or having them openly disclose their health information; hence, single-player games might have been preferred. Second, the game’s topic of focus is an important factor to consider. Some topics are easier to work with at the individual level than at the group level. For example, SOLVE was a game that tried to decrease the feelings of shame that gay or bisexual men might experience regarding their sexual preferences [35]. Considering that many players had strong feelings of shame and were reticent to disclose information about their sexual interests, they may not have been comfortable playing a game with others.

The third influence is logistical, technological, or economic restriction. When PCs were not ordinary household items but specialized pieces of hardware, they were not as commonly available as they are now. For this reason, older games tended to be a social experience; many people had to use the same computer to make it viable for enough players to play the game [27,28].

The final reason for choosing single- or multiplayer games relates to the learning strategy of choice. Some game developers opted to purposely promote out-of-game discussions of sexual health topics [28]. The aim was to make the game a topic for discussion with family members, teachers, and friends.

There are several ways in which we could try to classify the underlying mechanisms used in video games to change specific behaviors in users. Here, we divided the mechanisms into 3 categories: those based on knowledge, those focused on enhancing skills and self-efficacy, and those that motivate change through emotions. These groups are not mutually exclusive; 1 intervention might have >1 underlying mechanism.

Some games aim to provide knowledge, expecting that it will generate behavior change. For example, some games share facts about contraceptives, their efficacy, and the risks involved in not using them. In such cases, one of the most critical elements is to provide a clear and easy-to-understand message. It has been noted that most interventions try to make the message grounded in a specific element or situation in the game. Many games focusing on providing knowledge prioritized the provision of immediate and clear feedback, specifying where and when an error was made and what its potential outcomes might be. The same applies to decisions that have a positive outcome. For example, in SOLVE, when a player chose to engage (or not) in in-game sex, they were offered a quick recap of all the previous decisions that drove them to their current state (decisions that were not always apparent at the time they were made).

The final element of knowledge is how it is constructed. We have previously stated that the message must be clearly stated. However, for some interventions, this did not necessarily mean that the message had to be explicitly delivered. For example, the HIV risk game had a clear message that needed to be delivered: older people were more likely to have HIV than younger people. Players played 10 rounds of the game in which they made a judgment about which character was more likely to have HIV. As feedback, the players did not receive the correct answer; they only knew whether they were right or wrong. This key message was supposed to be inferred (constructed) by the participants based on their in-game experience.

Two of the most frequently used strategies in games focused on increasing the participants’ skills and self-efficacy, which was achieved by a mixture of relatability and rehearsal. By relatability, we mean all the different factors that can make the situation in a game similar to the ones players face or think they will face. The developers made great efforts to provide experiences grounded in those that the players have had or will experience. The assumption is that, in general, the closer a setting and its characters are to the real world, the easier it will be for the player to assimilate the lesson and put it into practice. This is one of the reasons why several games designed characters with different personalities and stories so that the player can easily associate 1 or several of them with their friends and acquaintances. Similarly, one of the reasons why some avatars were customizable was to make it easier for players to empathize with their in-game presence. The same can be said of the setting in which the interactions occur. In several cases, the setting was very similar to that currently experienced by the players. One of the clearest cases of relatability is in the FPSG game. The player learns about the risks of dating apps by playing a game that uses an instant messaging app as one of its primary interfaces.

Similarly, rehearsal and repetition also played a significant role in improving self-efficacy. The idea is that players will train themselves to make safe decisions in real life because they have made the same correct decisions in a virtual world before. The more times a player does something, the more likely it is for him or her to feel (and be) proficient in it.

The final strategy relies on using emotions to generate a reaction in the player. There are several methods in which this has been performed in different games. Some developers opted to generate negative emotions that frustrated or scared players into action. For example, in Romance, if the players initiated unprotected sex, they would have a baby that would cry randomly during the game, negatively affecting their relationships with their friends. Other games used positive emotions to inspire players to act. BattleViro and Viral Combat are good examples of this practice, having used fast action, increasingly challenging shooter-style games to empower their players to take control of their treatment. Finally, there were games that aimed to reduce the negative emotions that inhibit players’ ability to do something. The best example of this practice is SOLVE, a game whose main aim was to reduce the feelings of shame that MSM might experience. Through a series of stories and vignettes, the intention was that the player might consciously acknowledge their desires as something normal, which carries no stigma.

We cover 3 main topics in this section. We begin with a general description of the development process of the games. We then assess the involvement of stakeholders in the creation of the game: who was invited, when, and in what capacity.

A final element to discuss is that although all interventions relied considerably on video games to achieve their goals, it was not necessarily the only method they used. Approximately 27% (3/11) of the games were meant to be played in conjunction with other activities.

For these specific interventions, the games seem to be one of the few activities in which the participants could take agency and act upon the knowledge they received. For example, when the participants are completing scales or watching videos, they are fairly passive, and the moments in which they play the games are the only times when they really take control, make decisions, and see their results. Although no intervention specified the playing time, or the time used in the other modules, it appeared that the games were the activities that comprised most of the participants’ time.

In this section, we discuss the effectiveness of the interventions in achieving their goals, organized by the outcomes adopted in our search criteria. Multimedia Appendix 3 provides a summary of the results.

Rather unexpectedly, none of the studies assessed the number of pregnancies. There are 2 possible explanations for this. First, almost half of the chosen games were not marketed to women but to MSM. Second, the sample sizes were too small, and the follow-up periods were too short, for the relatively low occurrence of pregnancy to become a viable measure of the success of an intervention.

Only Keep it up! used STI biomarkers to assess changes in STI incidence. The researchers tested for chlamydia and gonorrhea through self-collection of rectal swabs. Through matched odds ratios, the control group showed a 55% increase in STI incidence, whereas the treatment group showed a decrease of 51%. These results were significant; however, we must be mindful that this intervention had multiple components, and the video game was only one of them.

Approximately 27% (3/11) of interventions measured changes in reported contraceptive use. Unfortunately, they did so in very different ways, which limited our ability to compare them. Choose Your Own Adventure asked about condom use at the last intercourse and found no effect of the intervention. The other 2 studies assessed the frequency of UAI with nonprimary partners during the past 3 months. Although SOLVE was unable to show significant differences between the treatment and control groups, Keep it up! reported a significant decrease in the number of UAI events 1 year after the start of the intervention.

In summary, studies on the effects of video game interventions on contraceptive use have shown inconsistent results.

The acquisition of new knowledge was one of the most commonly measured outcomes; however, the topics and measures varied significantly among the different interventions. Frequently, ad hoc questionnaires were created to assess differences between the treatment and control groups. The Baby Game quizzed participants on the costs (both time and money) involved in taking care of a baby. Romance used the same methodology but compared knowledge about the efficacy of different contraceptive methods. Studies on both of these interventions suggested improved knowledge in the treatment group compared with the control group. However, we should keep in mind that knowledge was assessed only immediately after the game ended and that no effect size measure was presented. Other interventions also relied on ad hoc tests; however, it was unclear exactly what topics they explored. Choose Your Own Adventure showed positive results (of medium effect size); however, VODO failed to do so. Both interventions followed a pretest-posttest design.

A comparable example is that of BattleViro and Viral Combat. They both tested their participants’ HIV knowledge, and although they were very similar games, only BattleViro showed positive results. It should be noted that BattleViro tested their participants 16 weeks after the intervention started, whereas Viral Combat did so at weeks 12 and 24. The extended period between intervention and data collection in Viral Combat, compounded by attrition of 32% of the original sample, might have biased the intervention results. However, one would also expect that participants who stayed longer would be more engaged, would have clocked in more hours in the game, and would have a better overall performance.

In conclusion, we do not have enough evidence to clearly state that games have a significant effect on increasing knowledge of sexual health topics. The variety of topics assessed, the limited amount of information regarding the content of the tests, and the large variability in the time between intervention and postintervention assessment precludes our ability to establish a clear causal relationship between playing and learning.

Only 2 games assessed this variable. Romance asked participants to assess the odds of becoming pregnant when having unprotected sex. Improvements in favor of the treatment group immediately after they finished playing the game were reported. Unexpectedly, a study on Choose Your Own Adventure found significant differences in favor of the control group. However, the researchers measured a construct called susceptibility, which merged the perceived risk of pregnancy with the perceived risk of STIs; thus, there might have been some cross-contamination in the assessment. The authors speculated that their results might be related to the fact that fewer people initiated sexual activities in the treatment group (ie, they were abstinent) than in the control group; hence, they did not feel at risk of any adverse outcomes related to having sex. They also considered that as the treatment group was more aware of the risks and the measures they could take against them, they felt better able to protect themselves.

Approximately 36% (4/11) of games addressed this topic. We have already discussed the findings of Choose Your Own Adventure. The entire intervention of the HIV risk game was centered on assessing the risk of someone having HIV based on their age and gender. Positive results were obtained for both men and women. The assessment was performed immediately after the intervention and 3 months later (the last time only for male participants).

BattleViro and MyPEEPS also assessed the perceived risk of STIs but in an indirect fashion by asking about STI testing. BattleViro measured the types of sexual behavior, frequency of sex, and number and gender of partners reported in the past 3 months. The authors found no differences between the control and treatment groups. MyPEEPS measured the frequency of STI testing and found that after the intervention, those in the treatment group were more likely to get tested than those in the control group.

Approximately 18% (2/11) of interventions targeted attitudinal changes toward safe sex. VODO measured participants’ attitudes toward sex on 2 axes: liberal versus conservative and positive versus negative. Both the control and treatment groups shifted to a more liberal position; however, the change was greater in the treatment group. The change in this group was sufficient for it to move, on average, from a conservative perspective toward a liberal one. No significant changes were observed in the positive and negative axes.

Choose Your Own Adventure assessed the predisposition toward waiting to have sex and found a significant effect of their intervention among students aged between 15 and 16 years. After completing ≥1 of its modules, participants were more likely than those in the control group to postpone sexual initiation.

Self-efficacy, be it general or specific to sexuality, was one of the most frequently chosen outcome variables across the different games. Choose Your Own Adventure considered 4 domains of sexual self-efficacy: condom negotiation self-efficacy, condom use self-efficacy, situational self-efficacy (the ability to control a situation that might be conducive to sex), and refusal self-efficacy (the ability to say no to sexual intercourse). However, the findings were mixed. The intervention increased participants’ self-efficacy toward condom negotiation and situational self-efficacy; however, no effects were found regarding condom use or refusal. It is not surprising that condom use self-efficacy did not change considerably between groups as the intervention did not include any components that directly taught students how to apply and use a condom. However, it did have 1 activity specifically focused on improving refusal skills, which was tightly tied to the game itself. The fact that this activity failed to produce the desired results for this variable is noteworthy.

BattleViro and Viral Combat assessed a similar domain of self-efficacy: participants’ belief that they would be able to adhere to a treatment regime. Both interventions found no significant change in either the short (12 weeks) or long-term (16 weeks and 24 weeks) assessments. MyPEEPS reported positive results when assessing HIV self-efficacy in the short term (3 months). Researchers have yet to publish their results for the long-term assessments (6 months).

We consider that the results on self-efficacy are mixed. Considering that the interventions are varied in methodology and topics and that they have worked on different domains of self-efficacy, this is not a particularly surprising result.

Approximately 27% (3/11) of interventions aimed at reducing the number of sexual partners, all of them for MSM—Keep it up!, Viral Combat, and MyPEEPS—and none resulted in a reduction in the number of sexual partners in their samples.

Approximately 27% (3/11) of interventions focused on increasing adherence to either treatment or PrEP. BattleViro and Viral Combat used a mixture of bioindicators, self-reported behavior, and electronic device follow-ups to assess this outcome. A total of 2 bioindicators were used: HIV-1 viral load in BattleViro and 1ARV (activator protein 1) levels in Viral Combat. BattleViro produced equivalent decreases in HIV-1 viral load in both treatment and control groups. Viral Combat reported results that favored the treatment group at both 12 weeks and 24 weeks after the beginning of the intervention; however, these findings were not statistically significant. Treatment adherence, measured by self-report in Viral Combat and by self-report plus electronic device records in BattleViro, showed similar results. In this regard, no intervention showed better results than the usual treatment. MyPEEPS also measured PrEP and postexposure prophylaxis adherence using self-reported measures and found no significant differences between the treatment and control groups.