Seven days ago we wrote that the press release is dead and most campaigns hadn't noticed. The response was predictable: agreement from operatives who'd already moved on, quiet discomfort from everyone else.
This is what we didn't have room to say the first time.
The press release problem is a symptom. The disease is allocation. And the data on that is embarrassing.
The Number You Need to Know
Political campaigns in 2024 spent 36% of their ad budgets on digital. Commercial advertisers spent 78%.
That's a 42-point gap. And it's growing. In 2020, the gap was 36 points. Four years later, commercial advertisers pulled further ahead while political campaigns shuffled incrementally in the right direction and called it progress.
Total political ad spend crossed
That's not an indictment of the candidates. It's an indictment of who they hired.
What the Best Operations Actually Did
The industry average is a lagging indicator. It includes every county commissioner's race that ran a broadcast-only program, every legacy firm still trafficking spots to local cable affiliates, and every campaign that hired the wrong shop because they had a famous client on their reel. The average tells you where the floor is. It doesn't tell you where the ceiling is.
Look past the average.
ColdSpark ran Senator Dave McCormick's Pennsylvania Senate campaign. McCormick won. ColdSpark won 10 Reed Awards for that race, including Best Statewide Campaign from AAPC. The channel split isn't public. The result is.
On the presidential side, the numbers are public and instructive. Harris spent north of 50% digital in the final stretch, with roughly
The lesson is not "digital doesn't matter." The lesson is that Trump's team matched channel to objective. Meta for mobilization, with 9% of Trump's digital budget going to GOTV versus only 4.6% for Harris. CTV for broad persuasion. Google for intent capture. They spent less. They disciplined their deployment. Harris's operation spent on volume. Trump's spent on function.
That is the gap between the best operators and the average firm. Not just how much digital. How it's deployed and why.
Connected TV Is Where the Serious Money Went
CTV political spend grew 506% from 2020 to 2024. It now represents 45 cents of every digital political dollar.
The reason is basic math. Broadcast TV carries an effective CPM of $43. CTV runs
Broadcast buys reach everyone in the DMA regardless of party registration, district lines, or voter file status. CTV delivers to a zip code. When you account for the waste on non-district voters, digital is 14 to 65 times more cost-effective than broadcast depending on the geography. In rural districts where a broadcast signal covers three counties but the race is only one of them, a third of every TV dollar is buying impressions in precincts you don't need.
For down-ballot campaigns, CTV changes the math structurally. Broadcast minimums for a meaningful program have historically started at $5,000 or more. CTV campaigns launch at $50. A state legislative candidate who has never been able to afford TV-quality video communication now can. That's not an incremental improvement. That's a new channel.
The firms that understood this in 2022 are three cycles ahead of the firms that understood it last November. The 57% of political advertisers who now prefer to buy CTV programmatically over direct are not trendsetters. They are the mainstream. The ones still negotiating broadcast packages are the outliers. They just don't know it yet.
Why Your Consultant Hasn't Told You This
Traditional media consultants earn a commission on the buy. The historical rate is 10 to 15%. Some campaigns have negotiated it to 3%. But the structure holds: bigger broadcast buy, bigger check.
The Center for Public Integrity documented the conflict plainly: "The commission structure creates a conflict of interest because there's an incentive to recommend sizable television campaigns, driving up costs."
Run the math. A consultant on a 5% commission who recommends $3 million in broadcast earns
The Grassley campaigns in Iowa rejected the commission model explicitly: "Grassley campaigns have never believed in the arrangement where media advisors get a share of the buy and therefore have an incentive to have a bigger media campaign. It's a bad idea."
Most campaigns never push back on it. Most candidates never ask how their consultant gets paid relative to the size of the TV buy. They should start asking. The answer explains a lot about why the 42-point gap has widened for four consecutive cycles while firms collected commissions on the wrong side of it.
What AI-Augmented Operations Actually Look Like
The AAPC surveyed its members in 2026. 83% use AI tools at least once a week, up from 59% one year earlier. 48% use AI multiple times a day. Those numbers are moving fast.
But the distribution matters. The 83% headline includes people using ChatGPT to write a press release once a week. That is not what the early adopters are doing.
The early adopters are running A/B testing at a speed and scale that wasn't previously possible. Blue Dot Consulting used Battleground AI on the Kiana Fields state senate race in Kentucky, analyzing voter data and Meta user behavior to generate social ad campaigns. Faster content, lower cost, tighter targeting.
Selfie-style AI-produced videos ran 13 to 40% lower cost-per-click than traditionally produced content. They cost 26% less per voter recall. A production that took three days and a crew of five now takes three hours and a laptop.
The research on persuasion is direct. A study published in PNAS Nexus found that political ads personalized to individual voter personalities outperform generic ads measurably, and demonstrated that generating and validating them at scale with AI is feasible. The combination of AI generation with human curation produced the strongest results. That is not a lab finding. It's being deployed in the field this cycle.
The teams doing this produce more creative, test more variations, and reach more precisely targeted voters per dollar than any traditionally staffed operation can match. The gap is not a function of budget. It's a function of who's running the tools.
The staffing implication is real. Experts cited by CNBC have stated that AI absorbs enough production, testing, and optimization work that "staffing campaigns to the brim will no longer be necessary." A digital-first AI-augmented team produces more output with fewer bodies. That's not a future prediction. That's what the early adopters reported in the 2026 AAPC survey when 35% said AI has been "very effective" in their work and the top use cases were cost reduction and efficiency.
Where This Ends Up by 2028
At least 15 campaign ads featuring AI-generated content have run in 2026 midterm races as of this writing. Republicans are leading adoption.
The NRSC ran a computer-altered video using a candidate's actual voice to read old social media posts. Senator Cornyn posted an AI-generated synthetic video attacking a congressional opponent. The Texas primaries featured multiple candidates with disclosed or identifiable AI-generated creative. This is not experiment. This is production.
There is no federal regulation constraining AI in political advertising. A patchwork of state laws exists. None of it is tested. The field is open.
By 2028, three things will be true:
AI-generated creative will be the default for social and digital ads. Not a novelty. Not a line item to explain to the finance committee. The default. Campaigns that haven't built the production workflow will buy it from vendors at a markup instead of running it in-house.
Personalized video at scale will be standard in top-tier races. The same ad, localized by district, tailored by voter segment, delivered across CTV and pre-roll, generated and versioned by AI. The campaigns doing this now will have refined the process through two cycles. Everyone else will be buying the capability for the first time in a primary window.
Real-time creative optimization will make mid-cycle course correction automatic. The ad that isn't performing gets rewritten during the cycle, not after the election. The campaigns with AI workflows built in will course-correct in days. The campaigns running traditional production will course-correct in the next race.
The campaigns that built these capabilities in 2024 and 2025 go into 2028 with three cycles of institutional knowledge. The ones that hire a legacy firm for the 2028 primary and discover digital in October go into 2028 with zero.
The 42-point gap between political and commercial digital allocation is not an accident of timing. It's a function of incentive, inertia, and the advisory relationship between candidates and consultants who benefit from the status quo.
AI doesn't fix the incentive problem. AI makes the gap between early adopters and laggards permanent.
If your current media plan still treats broadcast as the primary vehicle and digital as the amplifier, you are running the 2014 playbook with 2026 money. Someone on the other side of your next race already knows it.
Christopher Paul Gergen
Founder, Dark Horse Political