The Tavern Take: Week of August 17, 2026 – Kansas Ads
August 17, 2026
As we’ve seen in high profile contests like the Wisconsin governor’s primary and the Michigan US Senate primary, huge late breaks driven by paid media are where races are decided.
This summer, Tavern proved that our ads move votes, and move them better than our competitors. The first example happened in Kansas.
Tavern worked with a group of allies supporting Noah Taylor in the US Senate primary. Though Noah ultimately didn’t win the primary, he moved from 3% in our first poll to 13.5% in a field of 11 Democrats over a few short weeks.
That movement came from paid media -- Tavern’s paid media.
The Ad That Attached Trump to a Price Dominated
This was our best performing ad in the Kansas Senate primary with a wooping 45.8% completion rate on Google. It followed all of our best practices: 15 seconds, our candidate’s name multiple times, Trump on screen, one focused line doing the work: Noah Taylor knows the real cost of war.
We picked that angle because of something we found back in March. Our AI driven winning machine tested hundreds of messages to find the strongest one. It rewrote the messages over and over trying to make them better. About half the rewrites got better and about half got worse.
One kind of message was the exception. When a message connected Trump to something specific that costs people something at home, the rewrite reliably improved. Trump on his own didn't do it. A local issue on its own didn't do it. The two together did.
That's the whole idea behind this ad: tie Noah’s story to Donald Trump and rising costs and do it all in 15 seconds.
The 3 Top-Performing Attention Ads
We also won the completion battle with our mesmerizing ads, see: here, here, and here. The same style worked for each, and a diverse set of visuals kept them interesting enough to run a continuous program.
Message wise, cost of living beat every other message head to head. No surprise there.
The look of these ads came from a study we ran where we took more than 1,000 political ads and tried to predict how well each one performed two ways: from the script alone, and from the footage alone. The footage was the better predictor. If you only test scripts, you're grading half the ad.
We also know what kind of footage to shoot. Across 801 political ads, the ones with tense music and a fearful tone persuaded people less. Calm ones did better. And when we looked at everything at once, the usual marks of a polished ad stopped mattering much. How the ad made people feel is what kept showing up.
What’s Next
In the weeks ahead, we’ll be highlighting the ads we’re running for campaigns and organizations across the country. Showing definitive proof of their power in moving voters and winning races.
— Want the full studies behind any of these findings? Email: data@tavernresearch.com
