Generative AI is becoming a political intermediary that works in private. Testing across eight platforms during Brazil's 2026 election found one assistant ranking candidates in 64% of conversations and indicating one in 68% — but the more common behaviour, in seven conversations out of ten, was quieter: relating candidates to whatever the user had just said they cared about, and agreeing with that premise four times out of five. This briefing sets out a six-stage answer chain, a matrix for classifying what kind of interaction actually occurred, a response stack and a standing monitoring protocol.
What is inside
- 01What Changed
- 02The Intermediary Is Private, Personalized and Generative
- 03The Artefact91 Political Answer Chain
- 04The AI Electoral Intermediation Matrix
- 05This Is Already More Than One Election
- 06What Organizations Should Not Do
- 07The Artefact91 Integrity Response Stack
- 08A Political AI Monitoring Standard