Navigating the New Frontier of AI-Powered, Cross-Platform Video Advertising

Mark Rose
4 days ago
26 min read

Seismic Transformation

The digital advertising landscape is undergoing a seismic transformation, driven by the convergence of streaming media, social video, and artificial intelligence. The once-clear lines between television and digital, brand and performance, and content and commerce have irrevocably blurred, giving rise to a new, complex, and intensely competitive ecosystem. This report provides an exhaustive analysis of this new frontier, offering strategic intelligence for enterprise-level advertisers, agencies, and technology platforms navigating the current and future state of streaming video advertising.

The analysis yields three critical findings that will define the market for the remainder of the decade. First, the market is undergoing "The Great Rebundling", where victory will belong not to individual services, but to integrated ecosystems that seamlessly combine premium video, social engagement, and commerce, all managed through a unified, enterprise-grade advertising platform. Google, with its YouTube and Display & Video 360 (DV360) pairing, and Amazon, with its Prime Video, Twitch, and Amazon DSP combination, currently represent the most formidable expressions of this model.

Second, Artificial Intelligence represents a profound duality, acting as the single most significant driver of both unprecedented opportunity and systemic risk. The AI Duality is evident in the rise of powerful automation suites like Meta's Advantage+ and Google's automated bidding, which are delivering remarkable gains in campaign efficiency and performance.1 Simultaneously, these same AI systems are introducing complex societal challenges, including demonstrable algorithmic bias that can perpetuate discrimination, a growing consumer distrust fueled by "eerie" levels of personalization, and the emergence of sophisticated, AI-powered ad fraud techniques that threaten the integrity of the entire ecosystem.3

Third, the industry is witnessing a fundamental Performance-Branding Convergence. The traditional bifurcation between upper-funnel brand advertising (the historical domain of television) and lower-funnel performance marketing (the strength of digital) is collapsing. Platforms like TikTok, along with the proliferation of shoppable ad formats on Connected TV (CTV), are creating a "full-funnel" advertising experience where a single video ad can simultaneously build brand awareness and drive a direct, measurable conversion.6

Leading platforms must evolve their focus from a purely technological pursuit of AI optimization to a more holistic, socio-technical strategy.

Based on these findings, this report puts forth a core strategic recommendation: leading platforms must evolve their focus from a purely technological pursuit of AI optimization to a more holistic, socio-technical strategy. This imperative requires significant investment in research to understand and mitigate the unintended negative consequences of AI, the development of transparent "human-in-the-loop" controls for enterprise advertisers, and a proactive approach to addressing the complex ethical challenges of personalization at scale. The research plan proposed for Google in the final section of this report serves as a concrete model for this new strategic direction, outlining a path toward building more resilient, trustworthy, and equitable advertising systems for the future.

The State of the Streaming Advertising Market in 2025: The Convergence of Content, Commerce, and Community

The digital advertising market in 2025 is defined by a series of interconnected macro-level shifts that have reshaped how content is consumed and monetized. The central theme of this new era is convergence—the battle is no longer for discrete audiences on separate platforms, but for integrated user experiences that fluidly blend entertainment with interaction and transaction. Understanding these foundational trends is critical to navigating the competitive landscape.

The End of the Linear Era

The mass migration of audiences from traditional, scheduled broadcast television to on-demand, internet-delivered streaming is no longer a forecast but an established market reality. In the United States, Connected TV (CTV) devices are now present in over 85% of households, and streaming now accounts for more than 44% of total television viewing time.6 This exodus of viewers has, in turn, precipitated a massive reallocation of advertising budgets. Billions of dollars that were once committed to linear TV upfronts are now flowing into a more fragmented, data-rich, and programmatically-traded CTV environment. This transition has fundamentally altered the mechanics of video advertising, replacing the broad demographic targeting of linear TV with the promise of precise, household-level addressability and real-time performance measurement.6

The Rise of the Ad-Supported Model

For years, the dominant narrative in streaming was the ad-free, subscription-based model perfected by Netflix. However, a combination of market saturation, subscription fatigue among consumers, and persistent economic pressures has led to a strategic pivot across the industry. The ad-supported model is now ascendant. This shift was crystallized by the entry of Netflix into the advertising space, a move that signaled the universal need for diversified revenue streams.7 This has resulted in a significant influx of premium, professionally produced, and "brand-safe" video inventory into the programmatic marketplace, altering the supply-demand dynamics and providing advertisers with new opportunities to reach engaged audiences in high-quality content environments.8

Consumer acceptance of this value exchange—a lower subscription price in return for viewing advertisements—has proven to be robust. Peacock, for instance, reports that nearly four out of every five of its paid subscribers are on a plan that includes ads, one of the highest ad-supported penetration rates among major Subscription Video on Demand (SVOD) platforms.9 This broad acceptance validates the ad-supported model as a sustainable and scalable pillar of the modern media business.

The "Full-Funnel" Imperative

The digital nature of streaming enables a powerful fusion of advertising objectives that were previously siloed. The upper-funnel, brand-building power of cinematic, full-screen video is now directly connected to the lower-funnel, performance-driven metrics of digital marketing. This is most clearly demonstrated by two parallel trends: the growth of shoppable and interactive ad formats, and the deep integration of Retail Media Networks (RMNs) into CTV advertising.6

Shoppable TV, which allows viewers to interact with an ad using a remote or QR code to browse products or add items to a cart, closes the loop between inspiration and purchase. Concurrently, the ability for advertisers to leverage first-party data from RMNs (like Amazon or Walmart) allows them to tie ad exposure on a CTV device directly to subsequent online or in-store purchase data. This creates a "full-funnel" marketing capability where brand campaigns can be measured with performance metrics. The growth in this sector is explosive; PwC projects that CTV advertising revenue will reach an estimated $51 billion by 2029, a trajectory fueled by this convergence of branding and performance, and supercharged by the application of AI-powered personalization.10

The Attention Scarcity Economy

Beneath these market shifts lies a more fundamental dynamic: the emergence of an economy based on the scarcity of human attention. The proliferation of streaming services from every major media and technology company—Netflix, Disney, Amazon, Google, NBCUniversal, and more—has created a state of near-infinite content choice for consumers. Simultaneously, social platforms, particularly TikTok and Meta's Reels, have become dominant forces in video consumption, commanding an enormous share of daily screen time with their highly engaging, algorithmically-curated feeds.11

Beneath these market shifts lies a more fundamental dynamic: the emergence of an economy based on the scarcity of human attention.

This explosion of choice has made focused user attention the single most valuable and scarce resource in the media landscape. Consequently, platforms are no longer competing solely on the breadth of their content libraries but on their ability to consistently capture and retain this scarce attention. This strategic reality explains the aggressive bidding by platforms like Netflix and Amazon for exclusive live sports rights and the meteoric rise of short-form video formats. Live events create appointment viewing and command high, focused engagement, while the algorithms powering TikTok and Reels are engineered to maximize user session times by delivering an endless stream of hyper-relevant content.8

Advertisers will increasingly pay a premium for placements within environments that can prove they deliver high-quality, focused engagement. This forces platforms to innovate not just in content acquisition but in user experience design, ad format creativity, and measurement to demonstrate their ability to command this invaluable resource.

For advertisers, this transforms the valuation of an ad impression. The context and the quality of the attention paid to that ad become paramount. An advertisement viewed during a pivotal moment in a live NFL game on Prime Video or during a deeply immersive scroll through a user's "For You" page on TikTok is fundamentally more valuable than an ad that plays in the background while a user is multitasking. The future of streaming advertising will therefore be characterized by a flight to quality attention. Advertisers will increasingly pay a premium for placements within environments that can prove they deliver high-quality, focused engagement. This forces platforms to innovate not just in content acquisition but in user experience design, ad format creativity, and measurement to demonstrate their ability to command this invaluable resource.

Competitive Landscape: A Platform-by-Platform Analysis

The streaming advertising market is a fiercely contested arena populated by technology giants, media conglomerates, and social media powerhouses. Each competitor brings a unique set of assets, strategies, and vulnerabilities to the battle for advertising revenue and user attention. A detailed analysis of each major player reveals the distinct value propositions they offer to enterprise advertisers.

Google/YouTube: The Incumbent's Scale and Sophistication

Google's YouTube remains the bedrock of the online video advertising market, leveraging unparalleled scale and a deeply integrated technology stack.

Ad Offering: YouTube provides a mature and highly diverse portfolio of ad formats designed to meet a wide range of campaign objectives. These include the ubiquitous skippable and non-skippable in-stream video ads, short 6-second bumper ads ideal for brand recall, and newer vertical video formats tailored for the YouTube Shorts experience.14 The platform operates on an auction-based pricing model, which offers flexibility for advertisers of all sizes. Typical costs range from a $4 to $10 Cost Per Thousand Impressions (CPM) and a $0.03 to $0.10 Cost Per View (CPV), with most businesses starting with a daily budget of at least $10.15 This accessibility is a key advantage, though costs can escalate significantly based on the competitiveness of the target audience.
Enterprise Tools: For large advertisers and agencies, the Google Marketing Platform is the central nervous system for campaign management. Specifically, Display & Video 360 (DV360) serves as the enterprise-grade demand-side platform (DSP). A core strategic advantage of DV360 is its ability to provide a single, unified interface for purchasing premium YouTube inventory, such as Google Preferred and YouTube Reserve, alongside programmatic video and display inventory from across the web.2 This consolidation simplifies cross-channel media buying and enables holistic measurement.
Strategic Position: YouTube's dominant position is built on three pillars: its massive global user base, its vast and diverse content library spanning from user-generated content to professionally produced media, and its deep integration with Google's unparalleled trove of search and audience data. The primary strategic challenge for YouTube is the evolving nature of video consumption. As CTV solidifies its role as the primary "lean-back," living room viewing experience, YouTube must position itself effectively against premium, long-form content hubs like Netflix and Hulu, which are often perceived differently by both viewers and brand advertisers.

Netflix & Disney (Hulu): The Streaming Giants' Ad-Supported Pivot

The traditional titans of premium entertainment content have aggressively entered the advertising market, leveraging the strength of their intellectual property and brand-safe environments.

Netflix: After years of resisting advertising, Netflix has launched its ad-supported tier with a clear strategy focused on the premium end of the market. The company is on track to generate over $2 billion in ad revenue in 2025.8 Its offering consists of standard 15 to 30-second pre-roll and mid-roll ad formats, with a strong emphasis on maintaining a high-quality, minimally disruptive viewing experience by carefully selecting ad break points.7 This premium positioning is reflected in its pricing, with CPMs ranging from $25 to $65 and high minimum campaign spends of $10,000 to $18,000, clearly targeting top-tier national and global brands.7 Strategically, Netflix is transitioning to its own in-house ad technology platform to gain greater control over targeting, measurement, and format innovation. The company is actively testing next-generation formats, such as pause ads and interactive ads, with a planned rollout in 2026.8
Hulu/Disney+: As an early pioneer in ad-supported streaming, Hulu possesses a more established and mature advertising business and technology stack. It offers a range of non-skippable ad formats and has innovated with unique placements like pause ads (which appear when a user pauses content) and binge ads (which reward a viewer with an ad-free episode after watching a block of commercials).18 The integration of Disney's properties has led to the creation of the Disney Campaign Manager, a unified self-serve platform that allows agencies to execute campaigns across Disney+, Hulu, and ESPN. This platform provides access to a massive audience of 157 million monthly ad-supported users.20 Hulu's pricing is more accessible than Netflix's, with CPMs typically in the $10 to $30 range and a low minimum campaign spend of just $500, opening the door to a broader range of advertisers.18
Strategic Position: The core strategic asset for both Netflix and Disney is their vast library of premium, professionally produced, and unequivocally brand-safe content, which allows them to compete directly with traditional television networks for major brand advertising budgets. Their primary challenge lies in technology and scale. While their ad-supported user bases are growing rapidly, they are still smaller than YouTube's total audience. Furthermore, they must continue to build and integrate sophisticated ad technology platforms that can rival the data-driven capabilities of tech-native competitors like Google and Amazon.

Meta & TikTok: The Social Challengers Redefining Video Engagement

While not traditional streaming services, Meta and TikTok are dominant forces in video consumption and fierce competitors for video advertising budgets, driven by their highly effective, AI-native platforms.

Meta (Facebook/Instagram): Meta's competitive strength is rooted in the massive scale of its user data graph and its dominance in short-form video through Instagram Reels. For enterprise advertisers, the Meta Business Suite and Ads Manager provide a centralized platform for managing campaigns, assets, and permissions across its family of apps.21 The centerpiece of its enterprise offering is the Advantage+ automation suite. This AI-powered engine optimizes campaigns by automating targeting, creative variations, and placement decisions, with documented success in driving significant improvements in Return on Ad Spend (ROAS).1 Meta is actively expanding its ad formats within Reels and its text-based platform, Threads, to provide brands with more tools to engage with trending topics and cultural moments.24
TikTok: TikTok's advertising proposition is built around its unique, immersive, sound-on, full-screen vertical video format. Key ad products include standard In-Feed Ads, premium TopView ads that appear when a user first opens the app, and Spark Ads, which allow brands to boost organic user-generated content.25 For large agencies and brands, the TikTok Business Center serves as a comprehensive hub for collaboration, financial management, and asset control.26 Similar to Meta's Advantage+, TikTok offers Smart+ Campaigns, which use AI to automate the entire campaign workflow. Advertisers provide high-level inputs, and the system handles targeting, bidding, and even creative generation and remixing.27 The platform also provides unique enterprise solutions like the TikTok Creator Marketplace for managing influencer collaborations and the TikTok Agency Incubator program to support agency growth.28
Strategic Position: Meta and TikTok are attention monopolies, especially among younger demographics. They are not direct competitors to Netflix in the market for long-form, premium scripted content, but they are formidable competitors for finite user screen time and video advertising budgets. Their platforms are built from the ground up around AI-driven performance and rapid optimization, presenting a significant challenge to the more traditional, sales-led models of legacy media companies.

Amazon & NBCUniversal: The Commerce and Content Conglomerates

This category represents two distinct strategic approaches: the convergence of content and commerce by a tech giant, and the unification of linear and digital assets by a legacy media company.

Amazon (Prime Video/Twitch): Amazon's decision to introduce ads on Prime Video represents a pivotal moment for the industry. Its unparalleled competitive advantage is the ability to integrate streaming ad exposure with its massive repository of first-party retail data. The Amazon DSP is the enterprise-grade platform for purchasing ads across Prime Video, the live-streaming service Twitch, and exclusive live sports properties like Thursday Night Football. The DSP provides access to over 20,000 unique first-party audience segments based on real-time shopping and streaming signals.29 This enables advertisers to execute campaigns with "closed-loop" attribution, directly measuring the impact of an ad view on a subsequent product purchase—a capability that remains a significant challenge for most other platforms.
NBCUniversal (Peacock): NBCUniversal's strategy is to counteract the fragmentation of the media landscape by unifying its traditional linear television assets and its streaming service, Peacock, under a single, comprehensive ad-buying framework called One Platform.31 The latest evolution of this strategy is One Platform Total Audience, an advanced tool that uses AI, machine learning, and first-party data to generate a single, optimized media plan that delivers unduplicated reach across all screens—linear and digital.32 To enhance the viewer experience and increase ad effectiveness, Peacock maintains a deliberately low ad load of just 4-5 minutes per hour, significantly less than linear TV.9 Its CPMs are positioned in the premium streaming category, typically ranging from $38 to $44.33
Strategic Position: Amazon embodies the ultimate convergence of content and commerce, creating a powerful advertising ecosystem built on transactional data. NBCUniversal represents the strategic evolution of a legacy media giant, leveraging technology to unify its valuable content portfolio and compete effectively in a cross-platform world.

The Enterprise Advertiser's Toolkit: A Deep Dive into Ad-Tech Platforms

For large-scale advertisers and their agencies, the effectiveness of a streaming platform is inseparable from the quality of the tools provided to manage, execute, and measure campaigns. The modern ad-tech landscape is dominated by sophisticated platforms that offer varying degrees of control, automation, and data integration. Understanding the capabilities and strategic philosophies of these enterprise toolkits is essential for any major advertiser.

Google Display & Video 360 (DV360): The End-to-End Solution

Google's Display & Video 360 (DV360) stands as one of the most comprehensive and widely adopted demand-side platforms (DSPs) in the industry. It is engineered as an end-to-end solution for large enterprises and agencies that need to manage complex, multi-faceted campaigns at a significant scale.2

Core Functionality and Key Features:
- Unified Inventory Access: DV360's primary value proposition is its ability to provide a single point of entry to a vast and diverse range of video inventory. This includes exclusive access to Google's own premium properties like YouTube Reserve and Google Preferred, all TrueView formats, as well as high-quality inventory from top broadcasters and publishers through its Advanced TV offering, which covers both over-the-top (OTT) and traditional linear television.2 This consolidation is a critical efficiency driver for media buying teams.
- Advanced Audience Management: The platform is deeply integrated with Google's rich audience data, but also allows for the seamless integration of an advertiser's first-party data and licensed third-party data segments. This enables highly precise and granular audience targeting. Advanced features include the ability to build new audiences based on campaign activity (e.g., users who viewed a specific ad) and to enforce sophisticated frequency caps across different campaigns and devices to prevent ad fatigue.40
- AI-Powered Automation: At its core, DV360 leverages Google's powerful machine learning capabilities to automate and optimize campaign performance. Its automated bidding strategies are designed to achieve specific campaign goals (e.g., maximizing reach, driving conversions) at scale, continuously adjusting bids in real-time without the need for constant manual intervention.2
- Holistic Measurement: Recognizing the challenge of a fragmented media landscape, DV360 offers robust measurement solutions. It provides de-duplicated reach and frequency metrics across different devices, formats, and inventory sources, giving advertisers a true understanding of how many unique users they reached and how often. It also incorporates Google's Active View technology for transparent viewability measurement and integrates with a wide array of leading third-party measurement partners for brand safety, brand lift, and audience verification.17

Amazon DSP: Commerce-Driven Media Buying

Amazon's DSP has emerged as a formidable competitor to DV360, built upon a unique and powerful strategic foundation: the direct integration of media buying with commerce data. It is a programmatic platform available to all advertisers, regardless of whether they sell products on Amazon.com.29

Core Functionality and Key Features:
- Unique Audience Signals: The platform's most significant competitive differentiator is its access to Amazon's vast, proprietary first-party data on consumer behavior. Advertisers can leverage thousands of audience segments built from real-time signals related to what users are browsing, searching for, and purchasing on Amazon. This allows for targeting based on actual purchase intent and lifestyle affinities, which is often more powerful than the demographic or interest-based targeting available on other platforms.29
- Premium Streaming Supply: The Amazon DSP is the exclusive gateway to advertising on Amazon's growing portfolio of premium video content. This includes highly sought-after inventory within Prime Video series and movies, live sports like Thursday Night Football, the interactive live-streaming platform Twitch, and the free ad-supported service Amazon Freevee.29
- Closed-Loop Measurement: For brands that sell products on Amazon, the DSP offers an unparalleled measurement capability: closed-loop attribution. This allows advertisers to directly connect a user's exposure to a streaming TV ad with their subsequent purchase of a product on Amazon.com. This ability to precisely quantify the return on ad spend (ROAS) is a powerful draw for performance-focused advertisers, particularly in the consumer packaged goods (CPG) sector.6
- Managed and Self-Service Options: Amazon provides flexibility for enterprise clients. A self-service option is available for agencies and brands with in-house programmatic expertise (with a recommended minimum spend of $10,000), while a managed-service option provides consultative support and campaign execution for clients with a typical minimum spend of $50,000.29

Meta Advantage+ & TikTok Smart+: AI-Native Performance Engines

In contrast to the broad, multi-channel approach of DSPs like DV360 and Amazon DSP, the enterprise tools from Meta and TikTok are integrated suites designed to maximize performance within their own powerful "walled garden" ecosystems. Their defining characteristic is a deep, native integration of AI into every facet of the campaign process.

Meta Advantage+: This is not a standalone platform but a suite of AI-powered automation tools embedded within the familiar Meta Ads Manager. It is designed to simplify and enhance campaign performance by automating what Meta identifies as the five key levers of advertising: campaign objective, audience targeting, budget allocation, placements, and creative.1 For video advertising, its capabilities are particularly impactful. The system can automatically optimize and reformat video assets for different placements, such as converting a landscape video into a vertical format for Reels or Stories. Its most powerful feature is its dynamic creative optimization, where advertisers can upload multiple creative components (e.g., several video clips, headlines, and calls-to-action), and the AI will automatically test thousands of combinations in real-time to identify and scale the highest-performing variations for different audience segments.23
TikTok Smart+ Campaigns: This represents a further step toward full automation. A Smart+ Campaign is a specific campaign type where the advertiser's role is primarily to provide strategic inputs—the desired business outcome (KPI), a library of creative assets, and basic targeting parameters like geography. From there, TikTok's AI takes over the end-to-end management of the campaign, including audience selection, bidding, and creative assembly.27 A key feature is its automated creative generation capability, which can remix provided assets with different music, on-screen text hooks, and scripts to create new ad variations. The system also actively monitors for creative fatigue and will automatically shift budget away from underperforming ads to maintain campaign effectiveness.27

The rise of these AI-native suites signals a fundamental shift in the role of the enterprise media buyer. The traditional model, exemplified by early DSPs, was built on providing expert human traders with granular control over every campaign variable. Platforms like Advantage+ and Smart+, however, are designed to abstract away that complexity, asking advertisers to trust the "black box" of the AI to deliver superior results. This transforms the media buyer's role from that of a "pilot," who manually adjusts every lever, to that of a "flight engineer," who is responsible for providing the AI with high-quality inputs (clean data, diverse creative), setting the overall strategic destination (campaign goals), and monitoring the system's performance. This new paradigm requires a different skill set, one focused less on tactical bidding and more on strategic creative development, data hygiene, and sophisticated measurement and analysis. It also presents a competitive challenge to independent ad-tech platforms, as the superior performance of the walled gardens' integrated AI systems can pull an increasing share of advertising budgets into their ecosystems.

This transforms the media buyer's role from that of a "pilot," who manually adjusts every lever, to that of a "flight engineer," who is responsible for providing the AI with high-quality inputs (clean data, diverse creative), setting the overall strategic destination (campaign goals), and monitoring the system's performance.

The AI Revolution in Video Advertising: A Duality of Progress and Peril

Artificial Intelligence is the central, transformative force in modern digital advertising. Its application is not merely an incremental improvement but a paradigm shift, enabling capabilities in personalization, optimization, and creation that were previously unimaginable. However, this progress is accompanied by significant and complex risks that threaten consumer trust, societal equity, and the very integrity of the advertising ecosystem. A clear-eyed assessment of this duality is essential for any responsible stakeholder.

The Upside: Efficiency, Personalization, and Performance

The positive impacts of AI are profound and are driving the next wave of growth in the industry.

Hyper-Personalization at Scale: AI's ability to process and analyze immense datasets of user behavior in real-time allows for the delivery of highly tailored and relevant advertising experiences. This moves beyond the crude demographic segmentation of the past toward a more nuanced, intent-based targeting model.6 By understanding signals like viewing habits, search queries, and on-site actions, AI can predict which products and messages are most likely to resonate with an individual user. The success of this model is evident in content discovery; Netflix, for example, reports that over 80% of the content its users watch is discovered through its AI-powered recommendation engine, a principle that is now being applied to ad delivery.46
Predictive Analytics and Optimization: AI-powered forecasting engines can analyze vast amounts of historical campaign data, seasonal trends, and real-time market signals to predict campaign outcomes with a high degree of accuracy. This predictive capability allows for more intelligent budget allocation and fully automated bidding strategies that optimize for specific business goals, leading to demonstrably higher return on ad spend (ROAS) and a lower customer acquisition cost (CAC).6
Automated Creative Production (DCO): The advent of generative AI is revolutionizing the creative process. Through a technique known as Dynamic Creative Optimization (DCO), AI systems can take a set of base creative assets—video clips, images, headlines, music—and automatically generate thousands of unique ad variations. These variations can be tailored to different audience segments, contextual placements, or even real-time external signals like local weather or sporting event outcomes.47 The adoption of this technology is widespread, with 86% of ad buyers reporting that they are already using or plan to use generative AI for video ad creation.45 This capability is a cornerstone of advanced advertising suites like Meta Advantage+ and TikTok Smart+.27
Cookieless Targeting Solutions: In response to increasing privacy regulations and the deprecation of third-party cookies, AI is becoming indispensable. AI-driven systems can analyze an advertiser's first-party data to identify patterns and build powerful lookalike audiences without relying on the cross-site tracking of traditional identifiers, providing a path forward in a more privacy-conscious digital world.48

The Downside: Bias, Fraud, and Trust Erosion

Alongside these powerful benefits, the deployment of AI in advertising has introduced a new class of significant and often subtle risks.

Algorithmic Bias and Unintended Discrimination: This represents the most profound and ethically complex challenge. A growing body of academic research, including seminal work from researchers at MIT, has demonstrated that advertising algorithms can produce discriminatory outcomes even when they are not explicitly programmed to do so and are not fed biased training data.4
- The Mechanism of Economic-Driven Bias: The core finding of this research is that discriminatory outcomes can emerge as an unintended consequence of purely economic forces within the ad auction system.4 A landmark study analyzed a gender-neutral ad campaign for STEM careers and found that it was shown to 20% more men than women, with the disparity reaching 40% for women aged 25-34.4 The investigation revealed that the cause was not a flaw in the algorithm's code, but a rational response to market dynamics. Across the advertising ecosystem, women are a more valuable demographic to a wide range of advertisers (e.g., in retail and CPG) because they are statistically more likely to click on ads and make purchases. This high demand drives up the average bid price required to win an ad auction to show an impression to a woman. Consequently, a campaign with a gender-neutral targeting strategy and a finite budget will naturally "win" more auctions for the less expensive male audience, resulting in a skewed and discriminatory ad delivery pattern.4
- Manifestations of Bias: This same dynamic has been observed in other contexts. Studies have documented how job advertisements for certain industries are disproportionately shown to specific racial and gender groups, and how ads related to arrest records are more likely to be displayed in conjunction with searches for names that are common in the African-American community.4 These outcomes can systematically limit access to opportunities for housing, employment, and credit for already marginalized groups, exacerbating existing social inequalities.
AI-Powered Ad Fraud: The same AI technologies that empower advertisers are also being weaponized by malicious actors. Fraudsters are now leveraging generative AI to create highly sophisticated forms of ad fraud that are more difficult to detect than traditional bot traffic. This includes using AI to generate fake identification documents and deepfake videos to bypass advertiser verification processes, as well as deploying intelligent bots that can convincingly mimic human behavior—such as mouse movements, scrolling, and clicking patterns—to generate fraudulent ad impressions and deplete advertising budgets.5
The "Eeriness" Valley and Trust Erosion: There is a delicate balance in personalization. While relevant ads are appreciated, overly specific or intrusive ads can cross a line into what researchers term the "eerie" valley, leading to consumer discomfort and a breakdown of trust.3 Research indicates that while the perceived "intelligence" of an AI-driven ad can increase consumer acceptance, the perception of "eeriness" strongly and negatively impacts that acceptance.3 As consumers become more aware of the vast amounts of data being collected about them, they are growing increasingly skeptical of AI-driven recommendations, fearing manipulation by opaque algorithms that prioritize engagement over genuine choice. This erosion of trust poses a long-term existential threat to the data-driven advertising model.46

A Proposed Research Plan for Google on the Ethical Application of AI

At Concrete, we approach research in a variety of ways. We’re offering this as a preliminary approach for Google to evaluate. Following a discussion to better understand Google’s priorities, we can adjust and refine it to align with their goals. Google's global leadership in both artificial intelligence and digital advertising confers a unique responsibility to pioneer the ethical and responsible application of these powerful technologies. The following three-phase research plan is proposed as a strategic initiative for Google. Its objective is to move beyond a narrow focus on technical performance metrics and develop a deeper, socio-technical understanding of AI's real-world impact. The insights generated will enable Google to build more resilient, trustworthy, and equitable advertising systems, thereby creating a sustainable long-term competitive advantage and setting a new standard for the industry.

Phase 1: Auditing for Economic-Driven Algorithmic Bias in Ad Auctions

Knowledge Gap: While academic research has established that economic forces within ad auctions can create biased outcomes, there is a lack of comprehensive, quantitative understanding of how, where, and to what extent this phenomenon manifests at scale within Google's own complex ad systems, including Search, YouTube, and DV360.4 Existing methods for detecting bias often focus on identifying flaws in training data or model inputs, but they may miss these emergent, system-level biases that arise from the auction dynamics themselves.
Research Questions:
- Across sensitive advertising verticals such as employment, housing, and credit, to what degree do auction price differentials between demographic groups (defined by age, gender, geography, and inferred race/ethnicity) lead to skewed ad delivery for campaigns that have neutral targeting settings?
- Can a predictive model be developed to identify which types of campaigns, advertiser verticals, or audience segments are at the highest risk for generating these unintended delivery biases?
- What new advertiser-facing controls or "fairness-aware" bidding strategies could be engineered to allow advertisers to achieve equitable reach goals without violating anti-discrimination laws or compromising performance? This directly addresses the policy dilemma identified in research where legal frameworks designed to prevent discrimination can inadvertently block simple fixes.4

• Methodology:

Controlled Experimentation: Conduct a series of large-scale, controlled experiments on the YouTube and Google Display Network platforms. Launch identical, neutrally-targeted ad campaigns for hypothetical opportunities in employment, housing, and credit.
Outcome Analysis: Meticulously analyze the final ad delivery reports to measure the demographic skew in impression distribution.
Causal Inference: Correlate the observed delivery skew with Google's internal auction data on the average bid price required to win impressions for different demographic segments, establishing a quantitative link between market economics and biased outcomes.
Product Prototyping: In collaboration with legal and public policy experts, prototype and test new bidding options and advertiser controls. An example could be a "Distribute Impressions Equitably" objective that advertisers could select, which would instruct the bidding algorithm to prioritize balanced reach across specified demographic groups, potentially at a slightly higher overall cost-per-impression.

Phase 2: Quantifying the "Eeriness" Threshold in AI-Powered Personalization

Knowledge Gap: It is widely acknowledged that overly personal or intrusive ads can be perceived by users as "eerie," leading to negative brand sentiment and an erosion of trust.3 However, this "eeriness" threshold is highly subjective, context-dependent, and poorly defined in a quantitative sense. To build truly user-centric AI, it is necessary to move from anecdotal evidence to a robust, data-driven framework for understanding and respecting user boundaries.
Research Questions:
- What specific data signals used for ad targeting (e.g., recent sensitive search history, precise location data, app usage patterns, inferred health conditions, content of personal emails) are most likely to trigger a negative "eerie" or "intrusive" reaction from users?
- How does the consumer's perception of eeriness vary based on the context of the ad, including the ad format (e.g., a skippable video ad vs. a static display banner), the platform (e.g., YouTube vs. Search vs. Gmail), and the advertiser's industry (e.g., a trusted retailer vs. an unknown financial services company)?
- Is it possible to develop a predictive "personalization comfort score" based on user survey data, which could then be used as a dynamic guardrail within Google's automated advertising systems to prevent the delivery of ads that are likely to be perceived as overly intrusive?
Methodology:
- Scenario-Based Surveys: Develop a comprehensive set of simulated advertising scenarios that showcase a wide spectrum of personalization tactics, from benign contextual targeting to highly specific behavioral targeting.
- Large-Scale User Studies: Deploy these scenarios in large-scale surveys to a diverse panel of users, measuring their emotional and cognitive reactions on a detailed scale ranging from "helpful and relevant" to "uncomfortable, eerie, or intrusive."
- Driver Analysis: Use advanced statistical methods like conjoint analysis and machine learning to model the survey responses and identify the specific data signals, contexts, and advertiser types that are the primary drivers of negative user perception.
- System Integration: Use the findings to inform the design of more transparent and granular ad controls for users (e.g., "Do not use my recent search history for ad targeting"). Critically, also use these findings to set internal, proactive "guardrails" on the types of data signals that automated systems like Performance Max are permitted to use, especially for sensitive topics.

Phase 3: Defining Critical Human Intervention Points in AI-Driven Campaign Management

Knowledge Gap: Enterprise advertising platforms like DV360 and Google Ads are becoming increasingly automated, fundamentally shifting the role of the human media buyer from a hands-on operator to a strategic supervisor. However, there is currently no clear, evidence-based framework that defines where and when human judgment and intervention are most critical to prevent costly errors, brand safety failures, performance degradation, or ethical lapses within these highly automated systems.
Research Questions:
- In a fully automated campaign type like Performance Max, what are the most common and most damaging failure modes (e.g., sudden and drastic budget misallocation, delivery of inappropriate creative-placement combinations, unexpected performance cliffs where results drop off sharply)?
- What leading indicators or data signals can be identified to alert advertisers to these potential issues before they escalate into significant problems?
- What is the optimal model for human-AI collaboration in enterprise campaign management? Where do expert users want more granular control, and where are they willing and eager to cede control to the AI?
Methodology:
- Qualitative Research: Conduct in-depth, qualitative interviews and "think-aloud" usability studies with senior media buyers and brand managers from top-tier agencies and advertisers who are heavy users of DV360 and Performance Max. The goal is to understand their mental models, workflows, pain points, and areas of mistrust with the AI systems.
- Quantitative Failure Analysis: Perform a large-scale quantitative analysis of tens of thousands of historical automated campaigns to identify statistical patterns and event sequences that reliably precede significant performance drops, budget anomalies, or ad policy violations.
- UI/UX Prototyping and Testing: Based on the qualitative and quantitative findings, prototype and A/B test new user interfaces, dashboards, and intelligent alert systems within Google's ad platforms. The design philosophy should be to facilitate "intelligent oversight" rather than simply adding "more knobs to turn," providing users with actionable insights and critical intervention points that leverage human strategic judgment where it is most valuable.

Charting a Course for Responsible Innovation

The streaming advertising landscape is at a critical inflection point. The immense power of AI to drive efficiency and personalization is undeniable, but so are the profound ethical and societal risks it presents. Navigating this duality is the foremost challenge for industry leaders like Google. The proposed research plan offers a clear path forward—not just to mitigate risks, but to seize a powerful competitive advantage by pioneering a new standard for responsible, human-centered AI.

By partnering with Concrete to execute this research, Google can gain invaluable, nuanced insights into the real-world impact of its advertising systems. The benefits are threefold:

De-risk Innovation: Proactively identify and address sources of algorithmic bias and user distrust before they escalate into brand-damaging public issues or regulatory challenges.
Enhance Product Value: Build next-generation advertising tools with sophisticated, user-centric controls and fairness guardrails that enterprise advertisers increasingly demand.
Solidify Market Leadership: Move beyond a purely technical arms race to lead the industry on the crucial dimensions of trust, transparency, and ethical AI, creating a more sustainable and equitable ecosystem for all stakeholders.

Concrete’s unique expertise in AI-driven behavioral design and qualitative research provides the essential socio-technical lens required to uncover these deep truths. We deliver not just data, but a clear, strategic understanding of the human side of the algorithm. To begin this critical research and shape the future of responsible AI in advertising, contact Concrete today.

The Concrete Advantage: A Legacy of Shaping Human-Centered Technology

For nearly two decades, Concrete has been at the forefront of the technology revolution, guided by a singular focus: understanding human behavior to create more intuitive, engaging, and valuable experiences. Our deep roots in the tech industry are not just in observing its evolution, but in actively shaping it. This history gives us an unparalleled perspective on the challenges and opportunities Google faces today.

Our legacy is written in the foundations of the streaming and interactive media landscape:

Pioneers of the Living Room Experience: We were pioneers in the smart TV space, conducting original work on the Yahoo Widget Channel and with Intel's groundbreaking Viiv technology, which laid the architectural groundwork for the streaming TV ecosystem we know today.
Architects of the Streaming Model: Our founders hail from industry-leading firm Schematic, who was instrumental in bringing the first wave of streaming programming to the web. They didn't just help put content online; they helped architect the advertising models that made it a viable business.
Masters of Interaction: Our expertise extends from the screen to the hand. We hold two patents for our design work on TV remotes, fundamentally improving how millions of users interact with content. We also partnered with an early interactive TV startup to design a smart, intuitive application that empowered creatives to make programming more engaging.
Deep Research Heritage: Long before AI became a household term, we were conducting foundational research on video recommendation algorithms for the top internet streaming sites. Our research on social computing gave us a deep understanding of online behavior and the future of social media, knowledge that is more critical than ever in today's cross-platform world.
Innate Understanding of Content: Our understanding of media is not just technical; it's ingrained in our DNA. Two of our four original founders come from the world of film and video. One, with a film degree and a minor in advertising, worked in live television, produced hundreds of commercials, promos, and PSAs before producing a nationally distributed independent film. Another partner honed their craft on major motion pictures. Simply put: we know and understand content development from the inside out.

This historical understanding of user interaction with media is the bedrock of our AI-driven behavioral intelligence. We have seen the evolution from the first click to the first stream, and now to the first AI-powered recommendation. This unique vantage point allows us to provide more than just data; we provide context, insight, and a strategic roadmap built on nearly 20 years of understanding the intricate dance between humans and technology. Partner with Concrete to not only navigate the future of AI in advertising but to build it.