Synthetic Electorate Intelligence

Ask the electorate.
Before you spend on it.

Test your message, your mailer, your attack, or your turnout push against the voters of any district in the country — and get a read on which way they're likely to move, and who moves, in minutes. No fielding window, no five-figure poll. The read is grounded in real Census, voter-file, and survey data, so it holds up.

Campaign intelligence and messaging optimization — not election prediction. What that means →

Live query — NC House District 36
Audience
Independents Suburban Turnout ≥ 2 of 4 n = 600
Prompt

“Does the property-tax-relief message move these voters more than the school-funding message?”

Net favorable shift
Property tax +11.4
School funding +6.0

Illustrative. Read reactions relatively — which message appears to move more, and for whom — not as a forecast.

Minutes
To test a message, not weeks to field one
ZIP → District
Any race, at hyperlocal resolution
Nationwide
Census + voter-file data, to the census tract
91%
Correct leading answer vs. public polls
Why it matters

There are 500,000+ elections a year. Almost none of them can afford to poll.

City council, school board, county offices, judges, state legislature, ballot initiatives, advocacy campaigns. Most run blind — no professional polling, no focus groups, no modeling team. ElectorateIQ gives them a way to test messaging and understand voters that's inexpensive, iterative, conversational, and hyperlocal, at any race size.

Sharpen the message

Put a mailer, a script, or two framings in front of the exact audience you're trying to move, and see which one appears to land — before the buy.

Spend where it counts

Direct scarce dollars toward the messages and geographies where a read suggests they'll do the most — and away from the ones that fall flat.

Move at campaign speed

No fielding window and no recruitment. Ask a question, iterate on the answer the same afternoon, and ask ten follow-ups.

How it works

Select a geography. Define an audience. Ask.

1

Select geography

Pick ZIPs, precincts, counties, or districts. ElectorateIQ assembles a panel for that area on demand from nationwide Census and voter-file data — nothing is pre-stored.

2

Define audience

Likely voters, independents, suburban women, low-turnout voters, parents, retirees, an issue-specific slice — the panel re-mints to match.

3

Ask questions

Test a message, upload a speech or mailer, compare framings, simulate an attack and a response, or ask open-ended.

4

Analyze & iterate

Estimated support levels, relative persuasion, subgroup differences, turnout effects, synthetic verbatims — then follow up conversationally.

Coherent twins, not one-off prompts. Each synthetic voter is a consistent individual — real demographics, real party and turnout, and a real respondent's attitudes held together — not a throwaway persona a chatbot improvises for a single question.

What you can do

Built for the decisions campaigns make every week.

See the concrete questions campaigns bring to it on the In Practice page.

Message testing

Upload speeches, mailers, social posts, and TV/radio scripts. See relative resonance, backlash risk, and subgroup differences. More →

Attack & response

Test opposition attacks and your defensive framings before they run. Find the counterargument that appears to hold with the voters who matter.

Turnout activation

Explore likely enthusiasm shifts, issue-driven mobilization, and demobilization risk before committing a field program.

Issue framing

Compare two framings of the same policy and see which appears to lift support — and where it may cost you.

Subgroup analysis

Drill any result into demographic and geographic slices — the same panel, conditioned every way you need.

Geographic heatmaps

Visualize issue intensity, persuasion opportunity, and turnout vulnerability across the areas you're contesting.

Why it should be trusted

Scored against live public polls.

In a 25-question battery across 12 audiences — benchmarked against Gallup, Pew, YouGov, Marist and AP-NORC — ElectorateIQ named the leading position on 91% of questions and pointed the right direction on 87%. It's strongest on who voters are and what they prioritize. A sample of passing questions:

Question & audience ElectorateIQ vs. poll
Student-loan forgiveness — young voters64% vs. 55–65%
Reproductive rights salience — suburban women55% vs. 55–65%
Congestion pricing — New York City45% vs. 45–50%
Top issue is crime — Chicago#1 · matches

Directional, not a substitute for a final poll — and honest about where it compresses. Read the full validation & its limits →

How it works, under the hood

Where they stand — and how they react.

Every result runs through the twins. The platform routes each question to the right engine.

Where they stand

Served from real answers

Vote, party, turnout, and standard issue positions are each twin's own donated survey answer, aggregated over the panel. No model call — accurate, per-twin, coherent, and drillable.

  • Base-rate standings & issue positions
  • Weighted by who actually lives in the area
  • Fully drillable to subgroups
How they react

Estimated by the Oracle

Reactions to something new — a message, ad, candidate, or current event — go to the Oracle: an ensemble of Simsurveys' proprietary Consumer and Social models that reads each twin's most relevant attitudes plus the stimulus and estimates the likely response.

  • Message, ad & framing testing
  • Read relatively, not as a forecast
  • Directional lift, not an exact number
Why it can't be cloned

Not a prompt. A proprietary engine.

Anyone can ask a chatbot to role-play a voter — and get a confident average that means nothing. ElectorateIQ is three proprietary layers, built on validated population data and years of methodology, that a general-purpose model can't reproduce.

The Consumer model

A custom-built Simsurveys population model, trained on validated survey data to reproduce how real people actually respond. It carries the political reactions.

The Social model

A second proprietary model trained on decades of social-attitude research. It answers the values and moral questions the political record never asked.

Custom voter twins

The conditioning layer no off-the-shelf model has: donor-matched, locality-aware twins, assembled on demand, that anchor every answer to real demographics, real party and turnout, and a real respondent's attitudes.

The two models run combined, as an ensemble — social on values, consumer on political — and base-rate standings are served straight from each twin's real donated answers. It's the models and the twins together that make the read trustworthy, and neither is something you can spin up in a weekend.

Who it's for

Intelligence infrastructure for the underserved 99% of races.

Traditional polling and modeling are built for the handful of races that can afford a six-figure research budget. ElectorateIQ is built for everyone else — smaller budgets, faster cycles, higher demand for cheap iteration.

See it in practice →
Built for
Local campaigns State legislative races Political consultants Ballot initiatives Advocacy organizations Unions PACs Media organizations
Responsible use

What ElectorateIQ is — and isn't.

It is
  • Synthetic electorate intelligence — messaging and strategy optimization
  • A continuous, hyperlocal voter-understanding environment
  • Read relatively: which message appears to move more, which audience responds best
  • Validated at the aggregate level against real results
It isn't
  • Not an election-prediction engine or a replacement for a final poll
  • Not a reconstruction of any real, identifiable voter or their vote history
  • Not a deterministic per-person forecast
  • Never positioned to overclaim exact accuracy

Poll your race as often as you think about it.

ElectorateIQ is rolling out to campaigns, consultants, and advocacy organizations. Request access and we'll set you up with your states and districts.