DarkBadgeDarkBadgeThe addictive patterns in your apps, exposed.

Candy Crush vs. Duolingo: Two Freemiums, One Toolkit

Same streaks, same guilt notifications, same variable rewards — one app just calls it a lesson.

DarkBadge Editorial · Aug 14, 2026

Both take your money and your time. Only one claims it's good for you. DarkBadge scores them on the identical rubric.

Candy Crush Saga and Duolingo don't look like the same product. One is a match-three game that sells "thousands of levels" with no end. The other is a language app with an owl mascot and a reputation for making people feel virtuous about opening it. Both are free to download. Both make most of their money from the small fraction of users who pay to remove friction — extra lives in one case, a subscription and streak freezes in the other. Both want you back tomorrow, and the day after that, for as long as they can manage it.

▶ Read the full article

DarkBadge scored both on the identical five-dimension rubric. The result is not what the marketing on either app would predict.

The scores

| Dimension | Candy Crush Saga | Duolingo | |---|---|---| | Bottomlessness | 4.8 | 4.2 | | Difficult to self-regulate | 4.5 | 4.5 | | Speed | 4.2 | 4.0 | | Teasing | 4.0 | 4.3 | | Social pressure | 3.5 | 4.8 | | Total (mean) | 21 / 25 (4.2) | 21.8 / 25 (4.36) |

Duolingo scores higher. The gap is almost entirely social pressure: public streak counts, a weekly league reset that forces continuous grinding to avoid dropping a rank, and push notifications written to sound like a disappointed friend ("Duo is sad"). None of that is language pedagogy. It's retention tooling, bolted onto a core activity that — unlike Candy Crush's matching loop — has a genuine, independently useful skill on the other side of it.

Same toolkit, different claim

That's the actual difference between these two apps, and DarkBadge's current scoring doesn't capture it: Candy Crush's manipulative mechanics are the product — there's nothing underneath the loop. Duolingo's manipulative mechanics are wrapped around something that would still have value with the loop removed. The badge, as it exists today, treats both the same, because it's built to answer one question — how hard is this app engineered to be to put down — and both apps answer it about equally.

Whether that's the right question to stop at is a separate one. It touches evidence DarkBadge doesn't currently verify (does Duolingo actually teach a language, for whom, how well — and who gets to adjudicate that for every app that claims a benefit), and a scope question about whether "does this app deliver on its claimed value" belongs on a badge that otherwise sticks to describing mechanism, not judging worth.

That's the open question — not answered here. The follow-up piece in Explore takes it on directly: whether "outcome value" belongs on the badge as a second axis, and what it would take to score it responsibly rather than just assert it.

From the research

Apps on DarkBadge

Related