A survey by a USP professor shows that Daniela Mercury's Pipoca da Rainha block drew about 20,000 people at its peak, not the 2 million claimed by organizers. The study also estimates 21,000 attendees at Michel Teló's Sertanejinho, versus 1 million reported. The analysis employs advanced technology to challenge hyperbolic crowd figures common in mass events.
On February 22, 2026, during post-Carnival celebrations in São Paulo, Daniela Mercury's Pipoca da Rainha block on Consolação street and Michel Teló's Sertanejinho at Ibirapuera park drew significant crowds, but far below official estimates. Mariana Aldrigui, a USP professor of tourism policies, in partnership with Cebrap/USP's Monitor do Debate Político, calculated a peak of 20,000 people for Mercury's event and 21,000 for Teló's, using the P2PNet method.
P2PNet integrates drone aerial photos at varied times — 3:45 p.m., 4:20 p.m., and 5:00 p.m. for Mercury; 1:09 p.m., 1:42 p.m., and 2:31 p.m. for Teló — with AI software trained to count heads, achieving 72.9% precision and 69.5% accuracy, with a 12% average error. GeoSampa georeferenced data mapped the spaces: Consolação is 30 meters wide, and for 2 million people, each taking 50 cm x 50 cm, it would require 16.6 km of length, which Aldrigui describes as 'lined up neatly, cutely'.
"Our inability to understand dimensions makes us think any gathering has thousands, even millions of people," states the researcher. She notes neither block exceeded 100,000 people, accounting for the typical carnival flow of entrants and exits.
The Public Security Secretariat stated the Military Police did not estimate crowds this year. São Paulo City Hall and Daniela Mercury's team did not respond to requests. Aldrigui criticizes exaggerated figures, like the 16 million projected for 2025 Carnival by Ricardo Nunes (MDB) administration, when the capital has 12 million residents.
Realistic data, per her, aids in planning infrastructure like portable toilets and transport, and assessing commercial needs. The hyperbole phenomenon occurs in events like the LGBT+ Parade and political protests, debunked by camera monitoring, mobile phone data, and AI.
