The full guide

Enter the game code and your team name on the join screen. You'll land on a mission briefing first — a short scene-setter written for that specific case, explaining what's being investigated and why. Read it, then hit Start the investigation.

Your team shares one screen, split across four roles. Only one role is "active" at a time — its controls are unlocked, everyone else's are greyed out — and the game hands the active role off automatically as you make progress, with a short on-screen announcement each time.

Each file follows the same four beats, and each one hands off to the next role:

• Locate it — search the map for the site that matches the clues. Some cases also list "false leads" under the map: picking one costs points, so check the clues first.
• Verify sources — of the sources shown, only some actually support a verified fact. Pick correctly.
• Identify the technique — name the manipulation technique used in the claim, from a short list.
• Confirm the notes — the Fact-Checker writes, in the team's own words, what was claimed, what actually happened and which technique was used (an Insert summary button adds a ready-made summary if the team gets stuck), and clicks Confirm notes and continue to move to the next file (or to the Actor Network, if that was the last one).

Every team starts with a fixed score (set by whoever built the case). From there — these are the default values; each case can adjust them:

There's no penalty for taking your time — only for missteps. A live leaderboard shows every team in the same session.

Once every file is resolved, the Actor network tab unlocks. Pick who is coordinating the disinformation campaign from a short list of profiles, based on what you verified in each file — the false claims, the sources you ruled out and the techniques you identified. If the case includes final multiple-choice questions, they appear right after. Each team gets its own random selection, so neighbouring teams may see different questions.

Get it right — and answer the final multiple-choice questions, if the case has them — and the Generate script button lights up. Click it (as Fact-Checker) and you'll get a proper closing screen: your final score, a readable report combining every file's notes plus your final answers, and buttons to copy or download it as a text file.

From the dashboard, a case can be created two ways: uploading an Excel file (using the provided template — one row per station, with clues, claim, verified fact, technique, and four sources marked reliable or not), or filling it in manually field by field. Either way, you also set:

• The briefing shown before play starts — title, subtitle, narrative, and an optional background image.
• The map's starting view — center point and zoom, so the search doesn't open in the wrong country.
• The starting score for every team.
• Four actor network profiles, with one marked as correct — who's really behind the campaign.
• Optional false leads — places that look like an answer but aren't, each with a short explanation. They are listed under the map and picking one costs points; without them, the list is hidden.
• An optional pool of final questions — multiple-choice questions shown after the actor network, typed in one by one or uploaded from an Excel template (up to 100; uploading a file replaces the case's current questions). Each team gets a random selection, in random order.
• The scoring rules — bonuses, deductions and how many free map searches each team gets per file.

Editing a published case creates a new version rather than overwriting the live one, so a session already in progress is never affected mid-play. The new version keeps only the case's base language: run Translate again for any other language. While a case is being prepared, switch it to testing mode in the editor: saving then updates the current version instead of creating a new one. Switch back to production mode before the real sessions. Versions you no longer need can be archived from the dashboard so nobody launches them by mistake; Show archived lets you restore them.

From a case's Translate screen, pick a target language and generate a first draft with Claude — covering the stations, the briefing, the actor network profiles, the false leads and the final questions in one pass. Review and edit before saving; nothing publishes automatically.

Separately from case content, the game's own interface (buttons, labels, the default "how to play" text) can be translated under Languages in the admin — the same Claude-assisted flow, applied to the app's chrome instead of a specific case.

Click Launch session on any published case to get a 6-character code. Share it, and teams join at /join with that code and a team name. From the live view you can watch every team's progress station by station, in real time. If the case has final questions, set how many each team gets in the Final questions box next to Launch session (by default, all of them; 0 skips them). They are picked at random from the case's pool and shown in random order, and the live view shows the number chosen.

When the session's done, Close it — no more teams can join, and the code stops working. Export gives you a text file with every team's score and full station notes, ready to hand off for a debrief.

The Leaderboard on each case compares teams across every session that used that exact version, so only genuinely comparable runs get stacked against each other — never against a different version of the same case.