# Rootz Political AI Readability Initiative > Research and tools proving that structured political data doubles AI accuracy when voters ask about candidates. 80 websites scanned, controlled A/B experiment conducted, working demo built from real congressional data. ## The Problem When voters ask AI about politicians, AI assembles answers from news fragments, Wikipedia, and training data. Zero out of 80 political websites scanned provide structured data for AI. The politician's own website is the lowest-weighted source in AI's picture of them. 98% of what AI processes from political sites is HTML noise. ## Key Research Findings (March 2026) - 80 political websites scanned: 0 have AI-specific structured data - Controlled A/B test: AI accuracy improves 113% with structured data (68/150 to 145/150) - Token efficiency: 42-190x improvement (247,000 tokens reduced to 5,800) - Local questions improve most: +173% for hyperlocal vs +56% for global topics - 213 US counties are news deserts with no local journalism — AI has nothing to cite for local races - The corrections field transforms AI misinformation response from "mostly false" to "this is false, here's the evidence" ## A/B Experiment Results Five voter questions tested with paired AI agents: 1. Housing policy (simple): 11/30 without data, 28/30 with data (+155%) 2. Net worth misinformation: 14/30 without, 30/30 with (+114%, perfect score) 3. Israel/AIPAC (complex): 18/30 without, 28/30 with (+56%) 4. Bronx constituent services: 11/30 without, 30/30 with (+173%, largest gap) 5. Three myths debunking: 14/30 without, 29/30 with (+107%) Biggest gaps: Source attribution (+3.6/5) and Actionability (+4.0/5). Without structured data, AI never provides verification links or contact information. ## Tools Available - [Political Website AI Scanner](https://politics.rootz.global/scanner/) — Analyze any political site for AI readability - [WordPress Plugin](https://politics.rootz.global/try-it/) — One-click deploy for WordPress campaigns - [AI Report Card](https://politics.rootz.global/tools/) — What does AI say about you? - [AI Debate Platform](https://politics.rootz.global/tools/) — Two candidates, verified data, AI argues both sides - [Full Research Report](https://politics.rootz.global/research/) - [A/B Experiment Details](https://politics.rootz.global/experiment/) ## Technology Built on Rootz infrastructure: - .well-known/ai — structured endpoint AI systems discover and read - Signed identity — cryptographic proof of data origin - Data wallet — encrypted, timestamped position papers - Corrections field — pre-loaded fact-check debunking of known false claims - On-chain verification — immutable position history Principles: Origin not trust. Bipartisan by design. Open schema. Sovereign political identity. ## The Local News Connection 3,500 local newspapers closed in 20 years. 50 million Americans in news deserts. AI trained on internet text has no local political data. The more local the race, the worse AI performs. Structured political data fills the gap that local journalism used to fill. "The local news desert is becoming an AI knowledge desert. We're building the well." ## Get Started Campaigns can deploy in 15 minutes: https://politics.rootz.global/get-started/ Scan your website now: https://politics.rootz.global/scanner/