Know how your message lands — then ship the version that wins.
Drop in an ad, email, speech, debate answer, fundraising ask, GOTV text, or landing page. Coldread reacts with hundreds of district-accurate voter personas, diagnoses what's costing you votes and donors, and hands you the rewrites — by segment — ready to deploy. Testing, diagnosis, and the fix, in one place.
Add the message first, then choose the audience and data before review.
Step 1
Add the message
Paste the copy, fetch a campaign URL, or attach a PDF/image/video.
Examples:
Start with one message.
public benchmark
A message preflight should recover the big human themes.
We benchmarked Coldread against a public MIT focus group study on AI and deepfake technology. The human study used 3 focus groups and 39 participants. Coldread recovered 18 of the 20 human insight points, generated 546 reactions, and surfaced the same major themes for $0.097 in model spend.
For campaigns, that means a fast first read before the ad buy, blast, debate prep, or press response. It is not polling. It is a preflight layer for catching confusion, low trust, weak proof, and predictable objections before the message goes live.
more reads to inspect objections, trust gaps, and splits
preflight spend$0.097
benchmark run id: run_2026-05-04T15-01-20-207Z_529d3cde
Cost · before media spend
Coldread message read
~$0.15
510 simulated American voter personas · ~1 minute · no recruiting, facility, or honoraria
Traditional research project
$30,000
Single 6-group study in a US metro · 48–80 participants
~$0.15 — Coldread$30,000 — research project
Campaigns still need real voter research. Coldread sits before it: catch weak proof, bad framing, and avoidable backlash before spending real money on media, polling, or field testing.
Message in → voter reactions → objections, splits, and rewrite direction
Traditional research
12 weeks
Plan · Recruit · Field · Analyze · Report · Buffer
~1 min — Coldread12 weeks — research project
Campaign teams do not always have weeks. Use Coldread for the first pass: will this land, confuse people, lower trust, trigger objections, or need a stronger proof point before anyone sends it?
“The cost-of-living point lands, but I need one local example before I trust the promise.”
Trust risk · 3/10
Walt Finch · retiree, rural PA
“It sounds like campaign talk. Show me the vote, the price, or the tradeoff, otherwise I tune it out.”
Plus 508 more named simulated voter reactions, recurring objections, segment splits, trust and confusion risks, and ranked rewrite moves. Exports as CSV, JSON, and Markdown. ~$0.15 and ~1 minute for the full panel.
Frequently asked, before you run.
+Are these real voters?
No. Coldread uses simulated voter personas: named, written profiles with age, occupation, region, voice, and biographical detail. The default 510-person Americans panel was built against U.S. Census-matched demographic targets and reviewed before shipping. Custom audience runs can target any simulated population size you enter; the hosted website generates a 240-respondent sample to represent that population from the audience and source context you provide. The /methodology page explains the construction.
+Is this polling?
No. Coldread does not predict election outcomes, vote share, turnout, persuasion lift, or fundraising performance. It is a message preflight: a fast way to find confusion, trust risk, objections, weak proof, and segment splits before the campaign spends money or puts a candidate on message.
+What does the score mean for a campaign?
Each persona gives a 1–10 first read and a short reaction in their voice. The aggregate is directional, not predictive. The useful parts are the named reactions, repeated objections, confusion risk, trust risk, polarization, and rewrite suggestions by audience segment.
+When should a campaign use it?
Before launching an ad, sending a fundraising email, prepping a debate answer, releasing a statement, testing a policy frame, or pushing GOTV copy. Use it to improve the draft before polling, canvassing, field testing, or live media spend.
+What does this cost?
Three tiers: Early Access at $19/mo, Pro at $29/mo, and Max at $120/mo. Every tier includes the full Americans panel, hosted generated-audience reads with any simulated population size, follow-up Q&A, exports, and MCP/CLI access. Your first message read is free — no card required.
+What happens to campaign material?
Your message is sent to the model provider for the run, then discarded. We log a minimal run record for billing and free-run limits. We do not store the message content, reactions, or free-text run output on our servers. Full disclosure at /privacy.
+Read the full method?
Full breakdown — panel construction, scoring, what this is not, privacy, and cost — at /methodology.
Run message testing from the campaign stack.
MCP · CLI
The Coldread MCP server puts 20 tools inside Claude, ChatGPT, Codex, Cursor, or any campaign workflow that can call MCP. Test a draft, compare two versions, ask follow-ups, and export the readout without leaving the workspace your team already uses.
2
6
2
2
5
2
1
Intake
Panels
Preview
Run
Readout
Follow-up
Export
Set up MCP →Included in every paid tier · token shown once at mint
GOV