AI Finds Climate Posts Feel Happier During Big Events — But Fewer People Act

A new study from researcher Wentao Xu looked at what people actually say online during big environmental awareness days. The AI scanned 364,118 public X (Twitter) posts tied to four recurring campaigns — Earth Day, Earth Hour, Global Climate Action Day and World Environment Day — across 19 separate event-years. A language model scored each post for happiness, hope about the future, confidence in group efforts, distress, and calls to do something. The goal was simple: find out whether these campaigns make people feel good, and whether feeling good turns into action.

The answer was mixed. During event periods, happy language ran about 9 percentage points higher than in the 30 days before. That sounds like a win. But talk about collective action — joining, organizing, demanding change — fell by nearly 11 points. Think of a stadium: the crowd is roaring, but almost nobody is signing up for the team. Even more surprising, posts that sounded happiest were less likely to be reshared widely. Warm feelings spread, but not far, and not into doing.

Why should you care? Because a huge share of modern activism happens in this exact way — a hashtag, a repost, a temporary burst of good feeling. If celebration quietly crowds out calls to act, campaigns can look successful while changing little. That matters whether you care about climate, elections, or charity drives. There's also a quieter implication: AI can now estimate how millions of people feel at once by reading their public posts, which raises real questions about how that mood data might be used.

The caveats are worth knowing. This is one researcher's analysis of one platform, and X's users have shifted over the years, so the numbers may not represent the general public. The happiness finding held up under several checks, but was less precise under one stricter statistical test. And AI mood-reading is an estimate, not a truth serum — sarcasm and nuance can fool it.