Upgrading Our Political Award Season Ballots

We have better vetting for movies than political candidates! Ever notice that? You can go to Rotten Tomatoes and discover gems in the rough, but when it comes to primaries and general elections the political award season for candidates, the nominations are… Who?

Determining the best candidate for each voter should be possible based on the quality of today’s AI, and the amount of public data available, from state candidacy filings, official sites, pundits on social media, and so forth.

But if you’re like every other American, we’re still in the dark, asking our wonky friends or scanning neighborhood road signs for inspiration while recognizing we still have no clue who is qualified to be our representative in government.

I posed this challenge to Perplexity AI, and it returned a ton of random, obtuse data from not-trustworthy websites on the current field. So I pressed on:

It’s strange that your references are random law firms and such, when I would think the state government websites would have a vested interest in presenting all viable candidates with profiles, policies, PAC information, endorsements, and so on. Doesn’t the state have any resources to inform the public on candidates?

Perplexity Responded that I was right to be skeptical of various sources “but it’s not where the state itself sends voters for official information. There are limits on what the state will and won’t provide.

What the state actually provides

• The [state] Secretary of State is the official record-keeper for who has qualified to run.
  • The SoS systems (especially the “My Voter Page”) show who is on your ballot and provide official candidate names and offices.
  • They do not generally write candidate bios, issue-by-issue policy breakdowns, or track endorsements and PACs, because the state has to remain neutral.
• The [state] government portal points people to elections info and to the Secretary of State for candidate services and qualifying information.
  • It has detailed instructions on how to run, what forms and fees are required, and the fact that you should use the Secretary of State’s “Candidate Services” page for official candidate/qualifying documentation.

So: the state gives you “who is officially running and where they filed,” but it stops short of saying “here’s who supports them, here’s their platform, here’s how they compare.”

Why you keep seeing non-government sites

Because the state deliberately avoids anything that might look like campaigning or favoritism, the richer information (platforms, endorsements, PAC activity, etc.) gets filled in by:

• Nonprofit civic tools like BallotReady and Ballotpedia, which combine official ballot data with basic bios and links to campaign sites.
• Media outlets and issue groups that aggregate candidate positions, debates, and questionnaires.

That’s why, when you ask “what are their policies and how do they compare?,” the trail often leads to those intermediaries, not to a .gov domain.

The lack of compiled candidate data is a growing concern because there’s no authoritative source for such information. We can’t trust the press. They lack the influence they once had anyway. But social media is even more untrustworthy. But that means every citizen has to do the SAME RESEARCH.

I would say there’s a business opportunity, much like a Rotten Tomatoes movie site can rank movies by category and other factors. With AI research models, it seems like something that could be developed. What would it take to get communities to move from disparate news sources and road signs to a single website that uses AI to determine the best candidates per voter input?

Perplexity: “You’re putting your finger on a real gap: there’s no trusted, neutral “Rotten Tomatoes for candidates”, even though the tech and data largely exist to build one. Existing civic tools (BallotReady, VOTE411, etc.) are partial stabs at this idea, but they stop short of ranking “best candidates per voter input,” especially with AI-driven personalization.

Why this is hard (but doable)

To replicate a Rotten Tomatoes model in politics, you have to solve three big problems at once:

• Data fragmentation and bias
  • Candidate info is scattered across campaign sites, state filings, social media, news stories, debates, and questionnaires, all with their own spin.
  • An “aggregator” (like Rotten Tomatoes’ Tomatometer) works by collecting lots of inputs and applying consistent rules so people trust the score.editorial.rottentomatoes+1
• Trust and neutrality
  • Movies are relatively low-stakes; if critics skew left or right, most people shrug.
  • In politics, if your system looks biased—even slightly—half the electorate will dismiss it. You’d need transparent criteria, open data sources, and clear explanations of how scores are generated.
• Ethics and legality
  • You’re not just aggregating opinions; you’re influencing elections.
  • You’d need to respect campaign-finance rules, disclaimers (“this is not a governmental source”), data privacy, and avoid becoming an opaque micro-targeting engine.aitechinsights+1

None of those is a show-stopper, but they dictate how you’d design and position a platform.

What an “Candidate Match” Would Need

Think in layers, like a stack.

1. Raw data layer (sources to ingest)
   • Official: ballot lists, candidate filings, campaign finance and PAC data, voting records, bill sponsorships.
   • Semi-official: candidate campaign sites (issues pages), questionnaires (e.g., from League of Women Voters–style orgs), and public statements.
   • Media / commentary: articles, interviews, debate transcripts and fact checks.
   • User input: ratings, “this person matches what they promised,” feedback on issue explanations.
2. Normalization & fact layer
   • Extract core facts: issue positions, vote history, donors, endorsements, etc.
   • Map them into a standard schema: e.g., a “tax policy” dimension, “immigration enforcement” dimension, “abortion,” “guns,” “education,” “crime,” “governance,” etc.
   • Use AI to summarize, but ground every claim in cited source documents so a user can click and verify.
3. Voter preference engine
   • Ask each voter a short but sharp questionnaire:
     • What are your top 3–5 issues?
     • Which direction on each (e.g., lower taxes vs. more services, stricter immigration enforcement vs. more leniency)?
     • How do you trade competence vs. ideology vs. character?
   • Turn that into an issue-weight profile and match candidates on a distance metric (how close their record and platform are to your stated preferences).
4. Scoring & explanation layer
   • Instead of one dumb “overall score,” you’d show:
     • Issue match score (how well they line up with your answers).
     • Integrity/consistency score (how closely their current rhetoric matches their record).
     • Transparency score (how much verifiable data exists vs. fluff).
   • For each match, AI should generate: “We think this candidate is an 84% match for you because: A, B, C,” with links to the evidence.
5. Interface and trust design
   • Simple front door: address → election → your ballot → your personal match list (like how VOTE411 and BallotReady let you see ballot-specific candidate info, but more personalized).
   • Extreme clarity: “We don’t endorse candidates. We show how well they match what you tell us you want, based on public records, not rumors.”

How to get communities to adopt it

People will only leave yard-sign politics if they feel this new tool is:

1. Nonpartisan and transparent
   • Publish your methodology, scoring formulas, and data sources.
   • Offer a “view without personalization” mode for skeptics.
   • Allow candidates to claim their profile and challenge factual errors with a documented correction process.
2. Backed by credible partners
   • Early partnerships with civic organizations (e.g., League of Women Voters–type groups), local universities, and maybe local newsrooms to co-host voter guides.organizations.ballotready+1
   • Treat it as infrastructure: “We power the candidate info widget on your local paper, your county GOP/Dem site, and student groups,” rather than another silo.
3. Directly useful at decision time
   • Integrate with state tools like sample ballot lookups (even if you’re not official): “Paste your address or precinct → here’s your real ballot and matches.”
   • Make it mobile-first so people can literally pull their personalized cheat sheet up in line at the polling place.
4. Community feedback loop
   • Let users flag when a score or issue summary “feels wrong,” then review and either correct or explain.
   • Consider separate “expert” and “citizen” scores like Rotten Tomatoes’ critic vs. audience split, but with much stronger identity and anti-astroturf safeguards.

Business opportunity and model

There really is a viable business angle here, especially if you pitch it as infrastructure rather than a media brand:

• B2B / SaaS angle
  • License the platform as a white-labeled voter guide to local news, civic groups, unions, churches, and advocacy orgs.
  • Provide APIs they can embed on their sites (e.g., “see how candidates match our group’s values”).
• Freemium civic product
  • Free core for voters (ballot lookup + basic match), paid tiers for campaigns and orgs to see aggregate, anonymized issue trends in their districts (while respecting privacy and avoiding micro-targeting abuse).
• Funding & credibility
  • Early on, this is almost certainly philanthropy + grants + maybe a low-burn startup, not “hyper-growth VC” money, if you want to remain trusted.

Practical Next Steps

If you wanted to move from “idea” to “MVP”:

1. Start small (one state, one race level)
   • Example: “Virginia 2026 state-level offices” with governor + a few statewide positions.
   • Manually build curated profiles for each candidate, with transparent references.
2. Design a pilot questionnaire + match logic
   • Pick ~10 issues and build a simple weighting and scoring algorithm.
   • Use AI mainly to summarize and communicate, not to decide who’s “good.”
3. Run a local pilot with a civic partner
   • Partner with a local civic group, or college to test the tool with a few hundred or thousand users.
   • Measure: trust, perceived bias, and whether it actually changes how informed people feel (like the research on AI voter assistants suggests is possible).
4. Iterate on trust and transparency
   • Make “why this score?” explainable at every step.
   • Publish a “bias and limitations” page that is brutally honest.

---

From there, Perplexity wanted to help me sketch a minimal MVP (so helpful!) but I thought I’d pose it to the public instead. Any takers?

The bottom line is, if we want better candidates in office, we need to identify them, and the current political process favors incumbents and the wealthy who can throw pallets of cash at big media players and social media.

Instead of outlawing “money in politics,” why not improve the candidate nomination process?