Political reporting is the domain where AI tools are most capable and most dangerous simultaneously. Capable because political information is abundant in training data; dangerous because AI models reflect the distribution of perspectives in that training data — which skews toward published, establishment, English-language viewpoints, and systematically under-represents minority political perspectives, non-Western politics, and dissenting views.
Three Rules for AI in Political Reporting
Rule 1: Use AI for logistics, not judgement. Use AI for date verification, background research, and stakeholder mapping — not for assessing political significance, framing contested issues, or characterising political actors. Rule 2: Always identify whose perspective is missing. Ask: who is systematically underrepresented in this AI-assisted draft? Add those perspectives deliberately. Rule 3: Never use AI-generated text for politically sensitive attributions. Every quote, every characterisation, every claim about a political actor must be human-verified against primary sources.