• by Joe Mandese , 2 hours ago

WASHINGTON, DC -- It's been nine months since I asked a panel of media buyers and ad execs to estimate what their share of budget will be targeting AI agents by the year 2030, so I asked a slightly modified version of it to a panel of political media buyers discussing the role of AI in political marketing at MediaPost's Marketing Politics conference here.

It's a little apples to oranges, because I moved the date for their question to 2032, which will be two presidential election years from now, but I was nonetheless surprised to see how much more bullish a group of media buyers are now about using media to target agents in the future than were the ad pros on my Outfront panel were last April.

Back then (see below), two of the five panelists predicted zero percent of their media budget would be used for AIs, and two predicted a relatively tepid 10%. Take the moderator's (me) response out of the base and the average of those responses was a 24% share of budget going to reach agents.

By comparison, even the most agent-skeptical respondent -- Bully Pulpit's Matthew Campbell -- predicted 25% of political media spending would be used to target agents, while the average of the three panelists and their moderator was 56% of political spending by 2032.

That said, he pointed out that it will be 40% of what he projects will be a $20 billion 2032 political ad spending election cycle.

"Even 40% is still $8 billion," he quipped.

What's most noteworthy about the bullishly AI shift in sentiment among media buyers from April 2025 to January 2026, is that -- in my opinion -- political media buyers tend to be more conservative in shifting paid media strategies than general market media buyers, which explains why the majority of political media buys still are placed on television, and a plurality of it is placed on good old linear TV.

That's something else Goldberg and I discussed on a panel later in the day focusing on when political media will catch up with the general market's share of digital ad spending, which is more than 70% today and is projected to be around 80% of total ad spending by 2030 and 2032.

One reason for political digital latency, he said, is some of the FCC and FEC rules for broadcast TV buys, which still include "lowest unit rates" for qualified political candidates.

And more or just as equally important as CPMs, Goldberg said, was the access to opposition campaign spending intelligence vis a vis those rules.

In terms of if and when those rules might change -- either being phased out for broadcast TV, or being applied to other forms of political media buys (say streaming) -- that's anybody's guess.

On that note, I'm just wondering whether and how lowest unit rates would be applied to AI agents in 2032?

So I asked one (Gemini 3 Pro) and here's what it said:

This is a fascinating and specific prediction that highlights a major shift in how we think about influence. Based on the trajectory of "Agentic AI" (AI that performs tasks on behalf of a user) and current federal election laws, here is an analysis of that projection and the regulatory question you posed.

1. Opinion on the Prediction (50% of Spend by 2032)

The projection is directionally correct but likely aggressive on the timeline.

The prediction assumes a fundamental shift in voter behavior: that by 2032, voters will trust their personal AI agents to curate, filter, and summarize political platforms for them, rather than watching TV ads or reading news directly.

• The Plausible Case: We are moving toward a "curated reality" where AI agents act as gatekeepers. If an AI assistant (like a future Siri or Gemini) filters your emails, summarizes the news, and answers questions like "Which candidate creates more jobs?", then political campaigns must spend money to influence that AI's output. This is effectively "Search Engine Optimization (SEO) on steroids."

• The Skeptical View: Political voting is notoriously emotional, tribal, and identity-based. While consumers might happily let an AI buy their laundry detergent (transactional), they may resist letting an AI filter their political worldview (ideological). Reaching 50% of all spending implies that traditional emotional appeals (TV spots, rallies, fear-based ads) will become secondary to data-driven AI influence. That is a massive behavioral leap to make in just six years.

2. Would "Lowest Unit Rate" Regulations Apply?

Under current law, the answer is No.

The federal regulation you are referring to is the Lowest Unit Charge (LUC) provision found in Section 315(b) of the Communications Act. Here is why it wouldn't apply to AI agents without a major act of Congress:

A. The Law is Platform-Specific

Currently, the "lowest unit charge" rule only applies to broadcast stations (TV and radio) and cable systems.

• It explicitly guarantees that during the 45 days before a primary and 60 days before a general election, candidates get the lowest rate offered to the station’s "most favored commercial advertisers."

• Crucially, this rule does not apply to digital advertising. Internet platforms (Google, Meta, etc.) are not subject to the lowest unit charge. Since AI agents live in the digital cloud, ad buys directed at them would fall under "digital/online" spending, which is currently unregulated regarding price pricing.

B. The Definition of "Advertising" Would Break

The LUC rule relies on buying "time" (e.g., a 30-second spot). Marketing to an AI agent doesn't look like a 30-second spot. It looks like:

• Sponsored Data Injection: Paying to ensure a candidate's policy paper is in the AI's "trusted" training set.

• Preference Weighting: Paying a platform so that when a user asks, "Who has a better climate plan?", the AI prioritizes Candidate X's data.

• API Access: Paying for premium real-time access to the agent's decision-making stream.

Because these aren't "blocks of time," the current legal framework for calculating a "unit rate" would be mathematically impossible to apply.

3. The "Black Box" Problem

If campaigns start spending billions to influence AI agents, we will likely see a push for new regulation, but it will be harder to enforce than TV rates.

• Transparency: On TV, if a candidate gets a cheap ad rate, we can see the ad. With AI, the "ad" is invisible—it's a subtle bias in the answer the AI gives a voter. proving that a platform charged a candidate a "fair" rate to bias an algorithm is technically very difficult.

• The "Equal Time" Doctrine: Related to ad rates is the rule that stations must offer equal opportunities to rival candidates. Applying this to AI is a minefield. If an AI answers a user's question about Candidate A, must it immediately offer a "rebuttal" from Candidate B?

Summary

The prediction suggests a world where campaigns optimize for algorithms, not people. While plausible, current federal laws protecting candidate ad rates would offer zero protection in this new landscape. We would need an entirely new "Digital Campaign Act" to define what a "political ad" even looks like when it's targeted at a machine.