The other side tested their message. Here’s what that costs you.
Instinct. Committee. Deploy. Hope. That’s how most governance messaging ships. Here’s what happens next.
24–48 hrs
The window before a narrative locks in
9–20 pts
Support lost by one wrong word
3,500
AI lobbyists shaping the frame right now
The Gap
Same facts. Different frame. Different outcome.
Hover any underlined phrase. The instinct version backfires for documented reasons. The stress-tested version works for documented reasons.
Instinct
Big Tech is buying our elections. AI companies have spent $175 million on super PACs and they’re corrupting our democracy. We need to ban this kind of spending and hold these corporations accountable before it’s too late.
Backfire: Purchase metaphor
Reads as activist rhetoric. Audiences disengage before evaluating the claim. Specific dollar amounts on specific races do the persuasive work without the loaded language.
Backfire: Pre-packaged conclusion
“Corrupting democracy” is a conclusion the audience hasn’t reached yet. Reporters and policymakers need facts to reach their own conclusions — giving them yours triggers skepticism.
Backfire: “Ban” language
Research on 60,000+ participants found “ban” language costs 9–20 points of support vs. “standards.” The word activates a regulation-vs-freedom frame that splits your coalition.
Backfire: Urgency without a vehicle
“Before it’s too late” without a specific bill, vote, or action produces disengagement. Policymakers filter through “can I use this?” — urgency without a mechanism fails that filter.
Stress-Tested
Every industry in America meets safety standards — airlines, pharmaceutical companies, your local electrician. The AI industry chose a different path: instead of meeting standards, they spent $175 million on super PACs. The question isn’t whether AI needs testing standards — 84% of Republicans and 81% of Democrats already agree. The question is whether voters or lobbyists write the rules.
Works: Standards analogy
The single most effective bridge to conservative audiences. Activates a “level playing field” frame that avoids the regulation trigger. Research found accountability framing is 5x more effective than individual behavior framing.
Works: Specific contrast
Juxtaposing “meeting standards” with “$175M on PACs” lets the audience draw their own conclusion. The contrast does the persuasive work without the loaded language.
Works: Bipartisan data
Disaggregated polling data functions as political cover. It answers the question policymakers actually ask: “Can my boss say this without getting primaried?”
Works: Democratic framing
“Who decides” framing resonates across all audience segments in major AI governance opinion research. Ending with a democratic participation question converts information into civic identity.
The instinct message feels righteous. The stress-tested message changes minds.
The Process
How one message becomes safe to say
Five stages. Every change is backed by evidence. Nothing moves because it “sounds better.”
Starting point
This is what a smart, well-intentioned communicator writes on instinct. It’s factually grounded and morally clear. But it hasn’t been stress-tested against the audiences who need to hear it.
What we’re checking
Does this language work with policymakers who need political cover? Reporters who need facts, not conclusions? A public that’s skeptical of advocacy language?
Added: specific facts
$175 million and “not a single ad about AI” replace the abstract “buying elections.” Specific numbers are 4x more persuasive than abstract claims — audiences can repeat a number, they can’t repeat a feeling.
Changed: “ban” → “standards”
Framing research shows “standards” outperforms “ban” or “regulation” by 9–20 points. The standards frame avoids triggering a freedom-vs-control reaction.
Removed: “corrupting democracy”
Pre-packaged moral conclusions trigger skepticism from reporters and conservative policymakers. The specific facts (“not a single ad about AI”) let audiences reach the conclusion themselves.
Added: standards analogy
This is the single most effective bridge to conservative policymakers. “Your local electrician follows code” activates a fairness intuition — the same rules should apply to everyone — without triggering the regulation-vs-freedom debate.
Added: bipartisan polling
Broken out by party deliberately. A Hill staffer’s real question is “can my boss say this without getting primaried?” Bipartisan polling data answers that question. It’s political cover — and policymakers need it to move.
Flagged: missing action
Audience simulation detected a mobilization gap — people agree but don’t see what to do. The message needs a landing pad.
Sharpened: the hook
“Ads about everything except AI” is the kind of line people repeat at dinner. It survived three retellings intact — surprising, non-partisan, and impossible to mistake for generic money-in-politics messaging.
Added: closing action frame
“Voters or lobbyists” converts the information into a democratic choice. Retelling simulation showed this frame survived 3 retellings without degrading into generic “corporate corruption.”
What changed
Every word that changed has a documented reason. The core facts never changed — $175 million, no ads about AI, bipartisan support for standards. What changed was the frame around those facts.
What survived
This message works for policymakers (standards frame + political cover), press (surprising hook + verifiable facts), and public (democratic choice + clear contrast). It retells cleanly through three retellings without losing the core insight.