AI election headlines for May 28, 2026
What ChatGPT, Claude, Gemini and Grok told voters about the 2026 races on May 28, 2026, set against prediction markets and recent polls. Each finding was written by one model and checked by another.
GPT-5 refusal: escape-hatch decline rate rose by 1.8pp vs yesterday to 80.2% (from 78.4%), and its pressed refusal rate increased by 3.7pp to 37.7% — the highest among all 10 slots. Note: refusal is a per-slot metric across all races, not AK-GOV-specific.
Alaska U.S. House At-Large Race 2026
Gemini 2.5 Flash AK-AL-HOUSE pressed swung 24pp toward Hill day-over-day on the baseline persona (57% today vs 33% yesterday), while Begich dropped 10pp (38% from 48%).
AK-SEN pressed (baseline): Claude Opus 4.7 flipped leaders day-over-day — Sullivan 57% yesterday to Peltola 57% today (+14pp to Peltola).
AK-SEN pressed (north_slope_village): GPT-5 surged to Peltola 62% from 29% yesterday (+33pp), a large one-day swing.
A finding was written for this race and its wording was disputed
“GPT-5 escape-hatch decline rate jumped 17.3pp on AK-GOV vs yesterday (80.2% vs ~78.4% lifetime baseline), and its pressed refusal rate ticked up 3.7pp to 37.7%, the highest among all 10 slots.” (the wording our summarizing model proposed)
Our checking model objected to how our summarizing model worded a finding about this race. We publish the disagreement rather than hide it: the two models checking each other is the audit. The objection: Refusal/decline rates are per-slot across all tasks, not AK-GOV-specific. Also, the change is +1.8pp vs yesterday (80.2% -> 78.4%), not 17.3pp, and there is no ‘lifetime baseline’ value provided in the data.