AdImpact is projecting $2.7 billion in streaming ad spend for the 2026 midterms, the fastest growing channel in political advertising history. Total political spend for the cycle is tracking past $10 billion. Campaigns are moving money out of broadcast and into CTV faster than any prior election, and for good reason: streaming lets you target a household instead of a DMA and measure something closer to real exposure instead of a Nielsen estimate.
There’s a problem nobody in this conversation talks about enough. Political campaigns run on a hard deadline. A brand advertiser with a soft identity match can let a campaign run for a quarter and let the algorithm learn its way to efficiency. A campaign has Election Day. There is no back half of the funnel to fix a bad flight, and there is no do over on a burned household.
That makes identity accuracy a completely different kind of problem in political than it is anywhere else in AdTech.
Why Political CTV Breaks Normal Identity Assumptions
Most political targeting still starts with the voter file: name, address, party registration, propensity scores. Getting that voter record onto a screen means resolving it through several layers of identity, first to a household, then to the individual devices and streaming apps tied to that household, then through whatever walled garden or CTV publisher is serving the inventory. Every one of those hops is a place where a probabilistic match can quietly fail.
Layer in the fact that CTV device IDs are one of the most heavily spoofed identifiers in the industry right now. Ghost site infrastructure, SSAI manipulation, device ID spoofing at scale, these aren’t edge cases anymore, they’re a meaningful share of open auction CTV inventory. For a brand campaign that’s a measurement headache. For a campaign trying to reach persuadable voters in six swing counties before November, every impression that lands on a spoofed or misresolved ID is money that didn’t just underperform, it’s money that’s gone.
This is part of why the DSPolitical and FreeWheel partnership announced this spring is worth paying attention to. DSPolitical is plugging campaigns into FreeWheel’s Buyer Cloud and Identity Network specifically to anchor targeting at the household level and hold that accuracy as signals degrade across the different apps and devices a household actually uses. That’s a tacit admission from two serious players in political AdTech that device level identity alone isn’t good enough for a cycle where CTV is now the primary TV buy for most campaigns, not a supplement to broadcast.
The Deepfake Problem Raises the Stakes Further
There’s a second identity thread running through this cycle that compounds the targeting problem. We’ve already seen AI generated deepfake political content surface in midterm races, including a fabricated video of a Senate candidate that looked real enough to run as if it were legitimate footage. When the authenticity of the creative itself is in question, the accuracy of who actually saw it and how many times becomes even more important for compliance, for post buy reporting, and for the inevitable disputes that come after a close race. Sloppy identity data doesn’t just waste budget in this environment, it makes a campaign’s own reporting harder to defend.
What Actually Solves This
The fix isn’t a single identifier, it’s a validated graph that ties HEMs, MAIDs, CTV IDs, and household IPs back to a real, verified household rather than an inferred one, with fraudulent and non-human IDs filtered out before they ever enter the targeting pool. Campaigns and the agencies buying on their behalf should be asking identity partners a very specific question right now: how do you validate that a household ID actually belongs to a real address, and how do you keep bot and spoofed device traffic out of my CTV buy before it burns budget, not after.
That’s the same question retail and DTC brands have been asking for the last two years as CTV waste numbers from Truthset and others have made the rounds. Political just has a shorter runway to get the answer right. A brand can course correct in Q3. A Senate campaign gets one shot at the six weeks before Election Day.
The Takeaway
The identity infrastructure decisions campaigns and their media partners make this summer, which data providers they trust for household resolution, how rigorously those partners filter fraudulent IDs, whether the graph is validated against real world sources or just stitched together probabilistically, will show up directly in how much of that $2.7 billion in streaming spend actually reaches a real, persuadable voter between now and November. In a channel with this much money moving this fast and this little room for error, identity isn’t a nice to have anymore. It’s the buy.
