EmotionTrac
Politics guide

How to test attack ads for backlash before you air them

2026-09-01 · 8 min read · Rob / EmotionTrac

See how campaigns use facial coding to catch attack ad backlash before air, spotting sympathy spikes and credibility gaps second by second.

Attack ads boomerang more than campaigns want to admit

Negative ads work, until they don't. Research on campaign advertising has shown that a meaningful share of attack ads backfire, generating sympathy for the target instead of doubt.

The problem is you usually find out after the ad has run for two weeks and the tracking poll comes back wrong. By then, you've spent the media budget and handed your opponent a sympathy narrative for free.

Facial coding on opt-in panelists lets you catch the boomerang before it airs, not after.

What actually happens when an attack ad backfires

Viewers don't process attack ads the way media buyers hope. They don't just absorb the claim about the opposing candidate.

They also react to the tone of the attack, the visuals used against the target, and whether the whole thing feels fair. Those reactions show up on the face well before anyone can articulate them in a survey.

The sympathy boomerang

This is the classic failure mode. The ad is designed to trigger anger or contempt toward the opponent.

Instead, viewers show sympathy expressions right at the moment the attack lands hardest, usually during unflattering photos, ominous music cues, or exaggerated claims. That sympathy transfers to the target, not away from them.

The credibility gap

The second failure mode is skepticism directed at the ad's sponsor. Viewers show brow furrows and asymmetric mouth movement, classic skepticism markers, when a claim feels exaggerated or out of context.

That skepticism doesn't stay contained to the claim. It bleeds into how viewers feel about the candidate running the ad.

A practical testing workflow

You don't need a lab and six weeks. You need a rough cut and a panel that matches your actual target audience, down to district and demographic makeup.

  1. Recruit a matched panel. Swing voters in the district you're targeting, not a general national sample. Attack ads land differently in a suburban swing seat than in a base turnout district.
  2. Run the full cut, unedited. Second-by-second facial coding needs the real pacing, music, and voiceover timing. A script read-through won't surface the visual triggers.
  3. Map three emotion tracks. Anger and contempt directed at the target, sympathy spikes for the target, and skepticism markers tied to the sponsor.
  4. Flag the crossover moments. The moments where sympathy for the target rises faster than anger toward them are your boomerang risk points.
  5. Decide before you buy media. Recut around the flagged seconds, or kill the ad if the sympathy curve dominates the whole spot.

What to look for in the data

Three patterns matter more than overall sentiment scores.

Sympathy that outpaces anger. If viewers feel more for the target than against them, the ad is working against you regardless of the message.

Skepticism clustered on the claim, not the visuals. This tells you the problem is the argument, not the production. No amount of recutting music fixes a claim viewers don't believe.

Delayed negative reactions. If the anger or contempt spikes happen well after the claim, viewers are reacting to the tone of the attack, not the substance. That's a tell the ad reads as unfair.

When to kill the ad entirely

If sympathy for the target spikes higher than any negative emotion aimed at them, and that pattern holds across most of the panel, don't bother recutting.

You're not looking at an editing problem. You're looking at a strategy that doesn't survive contact with real viewers.

Where this fits in a campaign's testing cycle

Most campaigns test messaging in focus groups and polling, then move straight to media buy. Facial coding sits in between, giving you a second-by-second read on how the actual edited video lands emotionally before you commit budget.

EmotionTrac's Politics offerings run this exact workflow: opt-in panelists watch the finished cut on camera, and the system codes facial action units frame by frame to flag sympathy spikes, skepticism markers, and anger-versus-sympathy crossovers automatically.

You get a second-by-second emotional map instead of a gut check from six staffers in a conference room.

Try it: Schedule an EmotionTrac demo and see second-by-second emotion tracking in action. Or visit Politics for more information.

Sources

Fridkin, K. L., & Kenney, P. J. (2011). Variability in Citizens' Reactions to Different Types of Negative Campaigns. American Journal of Political Science.

EmotionTrac Politics