Future Audiences/Roblox game

The objective of the Future Audiences team for FY24-5 is to, based on insights from experiments that sharpen our understanding of how knowledge is shared and consumed online, provide recommendations on strategic investments for the Wikimedia Foundation to pursue that help our movement serve new audiences in a changing internet. To accomplish this, we are exploring strategies to expand beyond our existing audiences of readers and contributors so we can truly reach everyone in the world as the essential infrastructure of the ecosystem of free knowledge. As a team that pays close attention to evolving trends, we saw the increase in participation in gaming communities and spaces as an important area to better understand and learn how to position Wikimedia content and brand in.

To do this, we developed Wikispeedia, a game on Roblox that is inspired by Wikipedia speedruns. We validated this concept by promoting our game on Roblox and reaching out to content creators who actively participate in and make content about Wikipedia speedrunning on YouTube and Twitch.

Overview & Key Facts

What is a Wikipedia speedrun?

A Wikipedia speedrun is a game in which players try to navigate from one Wikipedia article to another using only internal links in as short a time as possible. For example, a player could start from the Wikipedia article for rabbit,” find a link to molecular biology, use that to navigate to x-ray crystallography, and finally reach the target article: Sir William Henry Bragg.

Why did we build a game? Why did we build it on Roblox?

Gen Z and Alpha have developed new forms of interacting and socializing. In order to effectively capture these generations’ attention, we need to act on their current habits. One major avenue of that is gaming. Not only does 90% of Gen Z play video games, but 73% of Gen Z has a video game console, 7% up from millennials. In fact, each new generation only plays games more: 94% of Gen Alpha plays video games. This means that an investment into gaming will enjoy longevity in relevance.

Roblox is a platform of user-created games that boasts 380M monthly active users (MAUs) as of 2024, up 19% from the previous year. Of this user base, 42% are under 13 years old. As a platform with a user base that mirrors our target audience, Roblox presents a suitable playground for us to learn how we can create a space for Wikipedia in gaming.

We set out to create a game that:

1. Captures the attention to all players on Roblox,
2. Introduces these players to Wikipedia, and
3. Teaches us which interactive features are most effective in engaging with our target audience.

With this game, we aimed to:

1. Determine to what extent (if at all) gaming is a viable path for Wikipedia to attract audiences that otherwise wouldn’t necessarily be on Wikipedia,
2. Better tailor our experiments to our target audience of Gen Z and Alpha, and
3. Provide more informative learnings and make more viable recommendations to other Foundation members.

What does our game do?

For our initial game, we set out to enable exploration and learning for users. We wanted users to:

1. Enter the Future Audiences Roblox game with their own credentials,
2. Start from a starting article and end their rounds by traveling to the target article,
3. Travel between different rooms that are linked together by article relations,
4. Influence the content of the map through in-game travel,
5. Play multiple rounds with different starting/target articles, and
6. Enjoy custom-made music and art assets as they play.

An early, low-fidelity version of our experience was demoed on 10 February. In this version, players were able to navigate a grid of rooms that each represent a relevant internal link/article in an aim to find the target room (and corresponding article).

For our initial release, we hand-curated 52 levels. A level would be selected for players depending on when they entered the game. Responding to feedback from internal testing, we also implemented three difficulty modes: Chill, Classic, and Chaotic. Our Chill mode was designed to create a low-pressure environment for play, while Classic was intended to be a fun and semi-challenging exercise. Chaotic, on the other hand, had an extra spin: in Chaotic, floors fall after players step on them, adding a layer of tension for extra fun.

Version One Learnings

We released and ran an initial course of promotions for Wikispeedia in April 2025 worth $236.31 in on-platform ads. This investment netted us 2.6M impressions, 9.8k plays, and 644 hours of playtime. Our game's retention was around 1%, but our stickiness floated between 6-30%. This likely meant that our initial game was fun, but could be more consistently compelling to users. It also meant that we likely needed additional Roblox-tailored mechanics (ex. dailies, limited items and events) to retain users.

To compare our three play modes, Chill, Classic, and Chaotic: players generally gravitated towards Classic, possibly because the entry point for Classic was located in the middle, between Chill and Chaotic. 38.5% of players who started a round did so with Classic, with Chill following at 32.57%. Chaotic experienced a start rate of 28.93%. Completion rates are even more separated: Classic is completed at a rate of 43.11%, while Chill and Chaotic fall behind with completion rates of 37.29% and 19.60%, respectively.

That being said, when we examine plays per user, Chaotic leads the pack. On average, each player who played our game on Chaotic does so 8 times, while Classic and Chill players played 7 and 6 times, respectively. It is possible that the more exciting nature of playing on Chaotic led to greater retention on such levels.

Overall, due to the promising results of our first round of promotion, we decided to investigate how to ensure this level of growth and retention. Our improvements were centered around making a compelling experience that naturally incentivizes players to play for longer periods of time and return. We wanted to ask our players two questions: 1) did they learn something new? and 2) did they learn more about Wikipedia? As such, our improvements were also aimed at incentivising deeper learning.

Version Two Learnings

How was Version Two different from Version One?

For our Version Two, we set out to learn more about retention and deeper engagement with learning via the game. Our version one successfully captured attention, but did not retain it, and we weren’t sure to what extent players were learning about or from Wikipedia (our primary goal). Using collected feedback, we planned and executed on a few improvements.

Testing retention improvements

In our first round of Wikispeedia, players would play levels that were presented to them in a repeating loop. Players were unable to select which level to play. There were no in-game methods for completion, tracking, and progression. A lack of longer-standing feeling of accomplishment was cited as a major reason for drop-off.

In response, in our second round, we introduced a distinct set of levels with a level selector. The level selector informed players of the levels they finished, how well they beat those levels through a best time and star rating, and the levels they still had yet to finish. The level selector also had a leaderboard for each level, instead of just appearing at the end of a level.

Testing deeper learning engagement features

To further create a sense of accomplishment, we also introduced puzzle pieces and fun facts. Players would encounter puzzle pieces while running through mazes, and could turn in these puzzle pieces in the lobby for a fun fact. Each fun fact was tied to the level in which the player found the puzzle piece. This version of Wikispeedia also included a lobby for players to hang out in without having to play.

To have a point of comparison, we created Trivia Game Show with Friends, a player-vs-player game that tested players' speed and accuracy when answering trivia questions, with significantly less investment. We wanted to put the same amount of adspend behind a different game that was created with much less effort to see how strong the effect of additional dev investment was.

Both games were released in September 2025.

How did we promote Version Two?

Future Audiences team members collectively posted on Reddit, Discord, TikTok, and HackerNews. Of these platforms, Reddit was the most successful, and we were able to acquire feedback from Redditors. We also conducted a live testing session on Discord, though there we acquired a much smaller pool of playtesters using this method. Our Communications team posted about Wikispeedia on Wikipedia's TikTok.

We also reached out to popular content creators who play and make videos about Wikipedia speedruns. Via this outreach, we found one YouTube creator and Wikipedia speedrunner RoyalPear, who was interested in playing the game and making a video about it. RoyalPear was paid for his time and work in trying out our Roblox game and sharing his experience in a video he created. In addition to content creators, we reached out to 26 libraries with Roblox clubs. 3 replied expressing interest, and we acquired some feedback from 1.

In terms of paid promotions, we set the same budget for Wikispeedia Version Two and Trivia Game Show with Friends that we did for Wikispeedia Version One, which came out to $260.24 and $260.61 in adspend, respectively.

What results did we achieve for Version Two?

Wikispeedia Version Two achieved 4,172,615 impressions, 70,240 clicks, 13,457 plays (0.32% play rate), 0.019 USD cost per play (CPP), an 81% thumbs up rate, and 864.7 hours of playtime. Trivia Game Show with Friends achieved 4,178,756 impressions, 64,446 clicks, 10,676 plays (0.26% play rate), 0.024 USD CPP, a 64% thumbs up rate, and 392.0 hours of playtime. We collected 66 pieces of feedback.

Our retention, however, which is what we set out to improve, dropped significantly. The feedback we collected indicates that we would likely need to either better position our game for the Roblox economy (adding PVP elements, introducing cosmetics, updating significantly at least once a week) or find a different game concept that would be better aligned with what Roblox players tend to gravitate towards.

What did we learn as a whole?

Our overall results can be summarized through the attached image. Our Roblox games altogether achieved over 11 million impressions and 34 thousand plays, leading to over 1900 hours played.

Overall Wikipedia Roblox Game Results

To compare our three games, Wikispeedia Version One, Wikispeedia Version Two, and Trivia Game Show with Friends: more effort did translate to more playtime, as can be seen in the difference in playtime between the three games, which indicates player interest and engagement levels. This interest level, however, is perhaps not to scale with effort: Wikispeedia Version Two incentivized around twice the amount of playtime that Trivia Game Show with Friends did, but took more than twice the time to build.

Most notably, our retention dropped significantly. Our improvements were certainly aimed at retention, but were not the typical retention drivers that Roblox games employ, which is likely why we were not able to positively influence our retention.

On incentivizing deeper learning engagement, we found promising signals that we were able to reach an audience of young knowledge-lovers with these games. For Wikispeedia Version Two’s puzzle pieces, 8,220 fun fact puzzle pieces were collected, and of these puzzle pieces, 1,260 were redeemed for fun facts. This points to a smaller percentage of players who appeared to deeply enjoy our game’s education elements. Qualitatively, we observed these players consistently engaging with our core game loop: selecting a level, collecting a puzzle piece, beating the level, redeeming the collected puzzle piece, and repeating. Further, we saw relative success with Trivia Game Show with Friends. Players would play for an average of 15 minutes.

Overall, our hypothesis was partially supported: we were able to create a Wikipedia experience on Roblox, and successfully used this experience to engage some Gen Alpha audiences in a new Wikipedia learning experience. However, we were not able to retain these players longer term even with additional engagement features.

To conclude, Roblox is a relatively effective platform for engaging with Gen Alpha audiences in short-term periods, but will require more integration with its ecosystem for long-term engagement and retention. As a platform, Roblox encourages player-vs-player games, constant updates, and monetization, which are not elements that necessarily naturally mesh with Wikipedia's mission.

How to stay updated on insights from this experiment

As usual, we will be sharing updates on this and other Future Audiences experiments during our monthly open community calls. Please sign up here if you’d like to be notified for upcoming calls.

If you have any further questions/inputs please get in touch on the talk page!