Case studyGetting a #1 puzzle game recommended by AI
In July 2026 the Angry Birds Dream Blast team started a pilot with Targetlytics to find out how ChatGPT, Gemini, Perplexity and other AI assistants answer when players ask for the best match-3 or puzzle game, and to turn those answers in Dream Blast's favour. By the end of the pilot, Dream Blast was named in 47% of tracked prompts, up from 19% at the start.
What the team got
- A repeatable view of how AI assistants recommend puzzle games, instead of one-off manual checks
- A single, accurate source of truth about the game that both AI engines and the team's own content work from
- A concrete list of third-party sources to win, with the outreach already drafted
- One prioritised task list a week, sent on Slack and shared with product and marketing
About the client
Angry Birds Dream Blast is Rovio's bubble-popping puzzle game, part of one of the most recognised mobile franchises in the world. It competes in the crowded match-3 and casual puzzle category against titles such as Candy Crush Saga, Royal Match and Toon Blast.
The Dream Blast team came to Targetlytics with a clear question: when a player asks an AI assistant which puzzle game to download, does Dream Blast come up, and if not, why not?
The challenge
More and more players ask AI assistants for game recommendations instead of browsing app-store charts. Those answers are built from the sources AI engines trust most: "best puzzle games" listicles, review sites, community threads and the brand's own pages.
The team had no visibility into three things:
- Which prompts players use, and how often Dream Blast appears in the answers
- Which sources the AI engines pull from, and which competitors those sources recommend instead
- What the brand's own material actually tells an AI model about the game, and whether that information was accurate and consistent
Standard SEO reporting did not cover any of this, and checking prompts by hand across several engines was not repeatable.
What Targetlytics did
Mapped the prompts and the people behind them. On day one the platform generated the list of prompts to track, from broad asks like "best puzzle games for iPhone" to niche ones about bubble-pop mechanics and Angry Birds titles, together with the player personas asking them. Every prompt was tied to a persona, intent, country and engine, and tracked on each run across the AI engines we monitor.
Gave the AI engines a reliable source of truth. The team uploaded Rovio's brand documentation into the Brand Constitution. Targetlytics scanned it and stored the game's positioning, naming, key features, tone and visual style. This became the reference for every recommendation and the basis for checking how accurately AI models describe the game.
Found where Dream Blast was missing. The Sources module showed which pages the AI engines cite when they answer puzzle-game prompts. Several popular "best puzzle games" listicles recommended Candy Crush and other competitors but left Dream Blast out entirely. The Outreach module drafted personalised emails to those publishers, asking for the game to be included.
Turned findings into weekly tasks. The recommendation engine produced a prioritised weekly task list: listicle placements to pursue, brand pages to update, questions the game's own content did not yet answer, each with an expected lift, so the team always knew what to do next. The list was posted to the team's Slack channel every week, so recommendations reached people where they already worked.
Onboarded the whole team. Onboarding calls brought in the Dream Blast product team and members of Rovio's marketing team, so the tasks landed with the people who could ship them.
How the pilot ran
- Week 1: setup. Pilot kicked off in late July. Prompt list, personas and competitor set generated; brand documentation uploaded to the Brand Constitution; first full visibility run across engines.
- Weeks 2 to 3: onboarding. Calls with the Dream Blast team and Rovio marketing to walk through the dashboard, agree on priorities and assign owners for recommendations.
- Weeks 3 to 4: gap analysis. Source and citation review surfaced the listicles and review pages where competitors were recommended and Dream Blast was absent. First outreach emails drafted.
- Week 4 onward: weekly cadence. Recommendation engine delivered a fresh task list each week, sent to the team on Slack; shipped tasks were tracked before and after over 30 days to measure their effect on mentions.
Results
Numbers from the Dream Blast workspace over the pilot.
| Metric | Start of pilot | End of pilot |
|---|---|---|
| Prompts tracked | 32 | 49 |
| Share of tracked prompts mentioning Dream Blast | 19% | 47% |
| Listicles and sources identified where Dream Blast was missing | — | 162 |
| Weekly recommendations delivered on Slack | — | 15 to 20 per week |
Want to know how AI assistants recommend your game or product?
Start a pilot with Targetlytics and get your prompt map, persona list and first recommendations in the first week.