FreeWheel this quarter published a report showing that media buyers and sellers in the television advertising business consistently overestimate how much viewers dislike artificial intelligence in advertising, while the two sides of the market remain split on whether AI actually improves the medium.

In Short

FreeWheel, the Comcast-owned ad technology company, asked TV ad buyers, TV ad sellers, and regular viewers what they think about AI in advertising, and the three groups gave very different answers. Buyers and sellers guessed that around four in ten viewers would be bothered by AI-made commercials, but when FreeWheel actually asked viewers, only one in ten said they disliked the AI ads they saw. This matters because advertisers may be holding back AI-generated ads out of a fear that does not match what audiences actually feel, and because buyers and sellers still disagree on basic questions like whether AI is making TV advertising better or harder to manage.

A gap between assumption and experience

Comcast Advertising, the parent unit behind FreeWheel, ran three separate surveys to build the report, titled "AI in TV Advertising: The Buyer, Seller, and Viewer Perspectives." The document compiles findings from a survey of 226 media buyers and 50 media sellers fielded by AdExchanger in April 2026, a viewer survey of 2,496 United States adults conducted by Dynata in the same month, and an earlier survey of 216 marketers and agencies run by AdExchanger in November 2025. Together, the three data sets let FreeWheel compare what buyers and sellers believe about viewer sentiment against what viewers actually reported.

The most striking figure in the report concerns creative perception. Forty-eight percent of buyers and 39 percent of sellers told AdExchanger they assume viewers find AI-generated creative off-putting. When Dynata asked the viewers themselves, only 10 percent said they disliked the AI-made advertising they had seen. That leaves a gap of 38 percentage points between what buyers expect and what viewers report feeling, and a gap of 29 points on the seller side. Neither industry side underestimated viewer tolerance; both overestimated the backlash by a wide margin.

That mismatch is not unique to this report. PPC Land has tracked a similar pattern across several separate studies through 2026. Magnite found that 78 percent of live-content streamers said AI-generated commercials would not negatively affect their perception of a brand, with 57 percent saying such ads could be as engaging as traditional ones. A companion Magnite study covering Hong Kong streamers found 89 percent reporting no negative brand impact from AI-made commercials, with only 16 percent reporting discomfort. A field study Taboola ran with researchers from Columbia, Harvard, the Technical University of Munich, and Carnegie Mellon, analyzing 500 million impressions, found that AI-generated advertising matched human creative performance when audiences could not tell the creative had been machine-made.

The pattern is not universal, however, and FreeWheel's report does not resolve the disagreement in the wider research record. Other studies have found meaningfully higher resistance. NielsenIQ presented research at CES 2025 showing that consumer brains process AI-generated advertising differently at a neurological level, with viewers describing such content in the study as annoying, boring, and confusing. A Raptive survey of 3,000 United States adults found that suspected AI content cut reader trust by close to half and reduced brand advertisement effectiveness by 14 percent in purchase consideration metrics. The IAB has also found a generational split: Gen Z respondents in one 2026 study reported 39 percent negative sentiment toward AI advertisements against 20 percent for Millennials. What most of these studies agree on, including FreeWheel's, is that the industry side of the market consistently rates viewer hostility to AI creative higher than viewers report it themselves.

Buyers and sellers disagree on whether AI helps

Beyond the creative question, FreeWheel's survey data shows buyers and sellers split on a more basic point: whether AI is making TV advertising better at all. Fifty-three percent of media sellers agreed that AI will make TV advertising better, compared with 41 percent of media buyers, a 29 percent gap in favor of sellers' greater optimism. The relationship reverses on frustration. Forty-eight percent of buyers said AI has glitches that make their jobs harder, versus 42 percent of sellers, making buyers 14 percent more likely to report AI-related friction in day-to-day work.

The disagreement extends to who actually benefits from AI's presence in the market. Asked who benefits most from AI's integration into media buying, 57 percent of sellers said everyone benefits equally, compared with 32 percent of buyers holding that view. Buyers were more likely to name specific winners: 24 percent pointed to advertisers and clients, 22 percent to agencies, and 22 percent to sellers and publishers. Sellers, by contrast, mostly saw the gains as shared, with only 19 percent naming advertisers and clients, 17 percent naming agencies, and 8 percent naming sellers and publishers as the primary beneficiaries. FreeWheel's own framing of that finding states plainly that buyers and sellers need to agree on what they want from AI in order for adoption to accelerate industry-wide.

That kind of buyer-seller misalignment is a recurring theme in ad tech reporting on agentic systems more broadly. FreeWheel itself has been an active participant in the shift toward machine-operated buying and selling. In March 2026, FreeWheel connected its platform to a Model Context Protocol server, piloting the integration with agency PMG's Alli platform to let external AI tools query and act on FreeWheel's deal and inventory data directly. That move positioned FreeWheel among the earliest premium video ad technology platforms to embed agentic AI capabilities into the transactional layer of its business rather than as a bolt-on interface. The buyer-seller survey results in the new report suggest that even as the infrastructure for machine-to-machine advertising transactions matures, the human stakeholders operating around that infrastructure have not converged on what they expect it to deliver.

What buyers actually want from automation

FreeWheel's buyer-side data draws a clear line between openness to AI assistance and reluctance toward full automation. Sixty-eight percent of buyers rated monitoring campaign performance as a high priority for agentic tools, the single most popular use case in the survey. Strong interest clustered around a second tier of tasks: 53 percent wanted AI ensuring deals target the most relevant inventory, 48 percent wanted help tracking top-performing supply-side platforms and publishers, 47 percent wanted assistance drafting requests for proposals, another 47 percent wanted help setting up campaigns and deal parameters, and 46 percent wanted AI monitoring spend against publisher commitments. Negotiating directly with publishers ranked lowest at 19 percent, the one task buyers were least willing to hand to a machine.

Buyers separately identified campaign planning and optimization as the area where AI can make the most immediate impact, a view held by 43 percent of respondents. But that appetite for assistance comes with limits. Only 22 percent of buyers expressed strong openness to AI autonomously managing campaigns without human oversight, and just 13 percent said they see AI as a near-term threat to their jobs. Taken together, the data describes a buyer population that wants AI functioning as what FreeWheel characterizes as an assistant rather than a replacement: useful for monitoring, drafting, and flagging, but not yet trusted to negotiate or run campaigns unsupervised.

That caution sits alongside broader industry findings about how automation is changing advertising roles rather than eliminating them outright. PPC Land reported in November 2025 that McKinsey's Technology Trends Outlook found job postings related to agentic AI increased 985 percent from 2023 to 2024, alongside 1.1 billion dollars in equity investment flowing into the category during 2024. Separately, PPC Land examined a paper citing Stanford data showing young workers lost 16 percent of AI-exposed jobs, a trend the report connected to advertising's growing reliance on agentic systems for targeting, bidding, and creative selection decisions.

Sellers prioritize pricing and inventory intelligence over agentic selling

On the sell side, FreeWheel's survey found comparatively muted enthusiasm for agentic selling tools. Only 26 percent of sellers said agentic capabilities represent where AI will have the most immediate impact on their business, a considerably lower figure than the 43 percent of buyers who named campaign planning and optimization as their top AI priority. Sellers instead concentrated their interest on inventory and pricing intelligence. Sixty percent named inventory allocation as a capability that would most help them maximize yield and increase revenue, followed by 55 percent for automated inventory classification, 53 percent for AI-assisted pricing strategies, and 48 percent each for audience value scoring and AI-driven packaging.

That pattern reflects a sell-side preference for AI applied to the operational mechanics of running a marketplace, rather than to the negotiation or transaction layer where agentic buying tools are increasingly being deployed. The gap between what buyers want automated (planning and optimization) and what sellers want automated (inventory and pricing) illustrates one structural reason the two sides continue to report different views on how much value AI has already delivered.

Targeting is where both sides agree

Audience and content targeting emerged as the clearest area of convergence between buyers and sellers in the report. Sixty percent of buyers said AI improves contextual matching, meaning the alignment between an advertisement and the content surrounding it, and 59 percent said they are open to AI supporting the creation of audience segments. On the seller side, 67 percent said AI-driven audience targeting helps them compete with other advertising platforms. FreeWheel's report frames sellers as already deploying AI-powered context engines that automatically detect themes across video and transcripts to support this kind of matching.

Separate FreeWheel survey data cited in the report found that 43 percent of buyers used AI in the past year specifically for audience segmentation and identification. That figure sits within a wider body of PPC Land coverage documenting persistent gaps in how connected television inventory gets classified and targeted. Gracenote, the Nielsen content intelligence unit, published a report in May 2026 finding that 86 percent of media planners would shift more linear TV budget into CTV if show-level targeting and reporting were consistently available, with 47 percent of planners naming limited show-level data as a primary barrier to reallocating spend. An earlier Gracenote study found that only 45.2 percent of respondents described themselves as very familiar with contextual targeting techniques, despite the technique's relevance to the brand-safety and relevance concerns advertisers routinely raise.

Attribution remains the shared bottleneck

If targeting is where buyers and sellers found common ground, attribution is where both sides identified the same unresolved problem, though they weighted its urgency differently. Forty-seven percent of buyers, drawing on the earlier November 2025 AdExchanger survey of 216 marketers and agencies, said proving return on investment remains the biggest barrier to increased spending on connected television. On the seller side, 69 percent said advanced attribution and incrementality measurement are the most important AI capabilities needed to compete with other advertising platforms, a considerably higher share than any single capability sellers named for inventory or pricing.

FreeWheel's own framing in the report links the two figures directly: if sellers can use AI to deliver better attribution and incrementality measurement, buyers gain the confidence needed to increase spend, and sellers in turn unlock faster, more defensible revenue growth. The report also notes that AI is only as reliable as the data feeding it, arguing that source-level, seller-owned data is necessary for attribution claims to be credible.

That framing lands inside a well-documented measurement gap that PPC Land has tracked across multiple vendors and research organizations throughout 2026. The IAB's State of Data 2026 report found that up to 75 percent of buy-side decision-makers rate attribution, incrementality tests, and marketing mix models as underperforming. Separately, AppsFlyer found that 58.6 percent of marketers believe their organizations underinvest in channels they cannot adequately measure, with connected television named among the worst attribution blind spots at 47.1 percent. A March 2026 IAB Europe discussion among five CTV Working Group members similarly identified cross-device attribution and fragmented identifiers as the primary technical blockers preventing connected television from completing its shift from a brand channel to a measurable performance channel. FreeWheel's own research from February 2026 warned separately that IP-address-based targeting can miss as much as 87 percent of households, a finding the company used to argue for more deterministic identity infrastructure across its marketplace.

Viewers want fewer repeats, less disruption

The viewer-facing portion of FreeWheel's report describes a population that is broadly receptive to AI use, provided it improves the day-to-day experience of watching ads rather than simply enabling more of them. Asked how open they were to companies using AI for specific purposes, 89 percent of viewers said they were open to AI reducing ad repetition within a single episode, and the same share, 89 percent, said they were open to AI choosing when ads are shown to minimize disruption to the content being watched. Openness to AI creating personalized advertising was somewhat lower but still a majority position, at 76 percent.

Those figures echo complaints about ad frequency and repetition that have surfaced elsewhere in the trade press. A separate PPC Land report on marketer approaches to AI and demographic data noted that DoubleVerify survey data found 40 percent of viewers keep watching a repeated advertisement, while 44 percent agreed that repetition helps brand awareness, leaving the majority position on both underlying questions unresolved. FreeWheel's figures suggest that when the question shifts from tolerating repetition to actively wanting AI to fix it, viewer support becomes much more one-sided.

Why this matters for the marketing community

The report lands at a moment when PPC Land has documented a wave of agentic infrastructure entering the CTV supply chain on both the buy and sell side, from FreeWheel's own MCP pilot to Magnite's Orchestration platform and PubMatic's AgenticOS. FreeWheel's data suggests that infrastructure investment and internal sentiment are running on separate tracks. Sellers are, on average, more optimistic that AI will improve the medium, yet less interested in agentic selling specifically than buyers are in agentic buying. Buyers want monitoring and drafting help but are unwilling to hand over negotiation. And both sides are, by FreeWheel's own measurement, significantly misjudging how viewers feel about the creative output already reaching them.

That misjudgment carries a practical cost. If buyers are withholding AI-generated creative out of a belief that viewers will react badly, and viewer data across multiple independent studies, including FreeWheel's own, suggests the actual reaction is muted, the caution may be constraining creative testing and production efficiency more than viewer sentiment justifies. At the same time, the disagreement over who benefits from AI, and the sell side's comparative lack of appetite for agentic selling tools, points to a marketplace where the technical rails for automated transactions are being built faster than commercial consensus on their value.

The attribution findings carry a separate, more immediate weight. With 47 percent of buyers naming ROI proof as the top barrier to increased CTV investment and 69 percent of sellers naming attribution and incrementality as the AI capability most needed to stay competitive, the report adds another data point to a measurement debate that the IAB, Gracenote, AppsFlyer, and IAB Europe have all been documenting through separate research efforts across 2026. Whether AI-driven attribution tools can close that gap faster than the industry's broader measurement standards evolve remains an open question the report does not resolve.

Timeline

Summary

Who: FreeWheel, the Comcast-owned advertising technology company, surveyed 226 media buyers, 50 media sellers, and 2,496 United States viewers, with fieldwork conducted by AdExchanger and Dynata.

What: A report finding that buyers and sellers substantially overestimate viewer dislike of AI-generated advertising creative, alongside continued disagreement between buyers and sellers over whether AI improves TV advertising, who benefits from its use, and where automation gets applied.

When: FreeWheel published the report today, September 5, 2026, drawing on survey fieldwork conducted in November 2025 and April 2026.

Where: The buyer and seller surveys covered marketers, agencies, and media sellers operating in the United States television and streaming advertising market; the viewer survey covered United States adults with access to traditional TV, paid streaming, or free ad-supported streaming television.

Why: The findings matter because advertisers may be withholding AI-generated creative and cautious automation based on assumptions about viewer reaction that the same report's own viewer data does not support, while unresolved attribution gaps continue to constrain how much buyers are willing to spend on connected television regardless of AI's other capabilities.