Agentic ad tech, CTV measurement, and impulse commerce data converge
DoubleVerify DV Neura, LiveRamp LAB, Mediaocean H2 survey, Adobe impulse buying, Pixalate OpenEPG Index, and Trade Desk location attribution in one briefing.
Two infrastructure platforms launched on the same day. A bi-annual survey documented the widening gap between AI investment intent and operational readiness. A consumer study showed that impulse buying has become the baseline rather than the exception for US online shoppers. And a new measurement index arrived with a structural critique of how the industry has been counting streaming audiences for years. All of it published on or before June 18, 2026. All of it sourced from PPC Land.
DoubleVerify’s DV Neura and the architecture of agentic verification
DoubleVerify on June 17, 2026 launched DV Neura, a cognitive AI engine embedded across its DV Media AdVantage Platform. The announcement from New York is one of the most technically detailed public disclosures of how a major verification and measurement company is approaching autonomous campaign management. DV Neura is not a single product but an architecture organised into four pillars: Media Intelligence, Adaptive Performance, Open Connectivity, and Agentic Execution.
The classification scale makes the ambition concrete. DoubleVerify says the platform has increased its content classification output by nearly 300 times since the start of 2026. That figure reflects the volume demand created by AI-generated content proliferating across the open web and the need to process it at speed. The hybrid architecture combining large language models, specialised machine learning, and deterministic rules is what enables classification at that throughput. Since the beginning of 2026, DoubleVerify says the platform has monitored or blocked more than 500 million impressions across AI slop sites and other low-quality generative AI environments. DV Scibids AI currently optimises 25 billion impressions each month.
The Media Intelligence pillar handles fraud detection and content classification. Adaptive Performance covers bidding optimisation and campaign measurement, building on DV Scibids AI and DV Rockerbox multi-touch attribution. Open Connectivity defines how external systems and AI tools access DV data. Agentic Execution connects insight to action.
The MCP integration is the most technically consequential piece. DoubleVerify is embedding support for the Model Context Protocol, the open standard originally developed by Anthropic, into its Open Connectivity pillar. The DV Neura Insight Agent, which clients can access through Anthropic Claude, allows campaign and media quality data to be queried through natural language rather than structured dashboard navigation. Integrations with Google Gemini and Microsoft Copilot are described as forthcoming. Additional support for the Ad Context Protocol (ADCP), an advertising-specific MCP extension launched in October 2025, is included in the Open Connectivity pillar. ADCP enables AI agents to discover inventory, compare pricing, and activate campaigns through a standardised interface, which means DV Neura is designed not just for human-initiated queries but for agent-to-agent workflows.
The timeline distinction matters. The DV Neura Insight Agent is available now. The DV Neura Activation Agent, which will autonomously execute approved campaign changes within advertiser-defined guardrails, is scheduled for Q3 2026. That sequencing separates the read layer from the write layer deliberately, while the industry works through governance questions about how much operational authority to delegate to automated systems. Once the Activation Agent launches, a workflow becomes possible where a planning agent from a DSP queries DoubleVerify’s measurement data through the ADCP interface, receives a brand safety response, and triggers a bid adjustment, all without human intermediation at any step.
What that means structurally for advertisers is a redefinition of where human attention is required. Rather than monitoring dashboards and adjusting bids in response to performance data, the operational job shifts to setting parameters, defining guardrails, reviewing audit logs, and adjudicating edge cases that automated systems surface for human judgement. The number of decisions a human makes per campaign does not necessarily decrease. The nature of those decisions changes: from reactive execution to prospective governance. Whether marketing organisations are equipped to make that transition is a separate question from whether the technology is ready, and it is the more uncertain of the two. The Mediaocean survey data released the same day suggests that 81% of marketers do not yet believe AI is transforming their workflows at a major level, which implies that the governance muscles required for agentic systems are not yet widely developed.
LiveRamp’s LAB program and the open agent marketplace
On the same day, June 17, 2026, LiveRamp launched the LiveRamp Agent Builders (LAB) program, a formal initiative allowing third-party AI companies to build and deploy purpose-built agents on the LiveRamp data collaboration platform and make them available to the entire LiveRamp customer base. The move extends a strategy that LiveRamp began executing in March 2026 when it first deployed active agentic AI capabilities.
Four founding partners were named at launch, each addressing a distinct workflow segment. SemantIQ focuses on health and life sciences. The SemantIQ integration enables AI-native workflows inside the LiveRamp Clean Room for healthcare marketing teams building and activating healthcare provider audiences, running media analytics, and generating faster planning intelligence. The clean room setting is not incidental. Manik Khanna, co-founder and CEO of SemantIQ, said agents can only transform healthcare marketing data use “if they operate within governed, auditable environments”, a reference to HIPAA constraints that make data access logs and audit trails a regulatory requirement rather than a product feature.
Newton Research connects to LiveRamp’s Cross-Media Intelligence system to turn measurement data into actionable planning insights. Albertsons Media Collective is already running the Newton Research agent in production. Liz Roche of Albertsons described the integration as “unlocking comprehensive analytics across data environments.” Newton Research was founded by entrepreneurs who previously sold Data Plus Math to LiveRamp, which gives the production deployment more credibility than a typical launch-day partnership announcement. The same people who built the measurement architecture that LiveRamp acquired are now building agents on top of it.
Akkio handles the audience workflow from discovery through activation within a single agentic flow. A fourth partner, DataFleets, covers data transformation. During the pilot period, brands receive access to agents from all LAB participants without committing to a single vendor upfront. That access model is a structural choice. LiveRamp is positioning its platform as a marketplace of specialised capabilities rather than a proprietary end-to-end solution. The question of which model wins, marketplace or vertically integrated, will likely be determined by which approach delivers measurable outcomes faster for the buyers who are now planning 60% spend increases in AI media during H2 2026.
Mediaocean’s H2 2026 survey and the 41-point gap
Mediaocean on June 17, 2026 released its 2026 H2 Market Report, the tenth in a bi-annual research series that has accumulated more than 6,400 total respondents since 2021. The May 2026 survey covered 312 marketing professionals spanning brands, agencies, media companies, and technology providers.
AI media leads planned spend growth at 60% for H2 2026, the highest figure recorded for any channel across the history of the series. That follows AI media already topping planned spend growth in the H1 edition at 54%. CTV and digital display tied at 63% planning increases. Social followed at 61%. AI media at 60% sits above retail media at 40% and search at 47%, marking the second consecutive period in which a nascent category surpassed traditional performance channels in planned investment growth. Traditional channels face the sharpest contractions: print recorded 49% of respondents planning decreases, local television 36%, national television 34%.
The figure that sits in uncomfortable contrast with all of that: only 19% of marketers believe AI is causing a major workflow transformation, down from 28% in the previous period. That 41-point gap between investment intent and perceived operational impact is not a contradiction in the data. It is a description of what is actually happening. Money is moving toward AI media placements while the organisations executing those placements have not yet restructured their workflows, integrated their data stacks, or hired the skills to manage automated systems at scale.
The capability priority data underlines this. AI-powered automation and cross-channel orchestration entered the top five capability priorities for the first time in the survey’s history. The named barriers to scaling AI deployment include data gaps, lack of integration with existing marketing stacks, talent shortages, and unclear ROI. None of those barriers are technology problems. They are organisational constraints that infrastructure launches like DV Neura and LiveRamp LAB are designed to lower. The timing of both announcements, one week before Cannes Lions begins on June 23, 2026, is not incidental.
The channel-level data from H1 2026, drawn from the November 2025 survey of 320 respondents, adds granularity to the trajectory. CTV and digital display and video tied at 63% planning increases in H1, with social platforms at 61%. AI media was already at 54% in H1, above search at 47%. That sequencing tells a specific story: AI media surpassed search in planned spend growth before it had even been in the survey series for a full year. The Mediaocean survey first included AI media as a tracked category in late 2025, and within two survey cycles it had become the fastest-growing line item in the dataset. The speed of that shift is what makes the 19% workflow transformation figure so striking in comparison. Investment is moving faster than operations are adapting.
Retail media at 40% planning increases shows durability rather than acceleration. The category has been in the survey for several cycles and has consistently shown elevated planned growth. What has changed in H2 2026 is that print, local TV, and national TV are now showing net negative momentum of a kind that suggests structural reallocation rather than cyclical softening. Print at 49% planning decreases is not a year-over-year dip. It is a directional statement about where marketers believe those audiences have gone. Traditional broadcast television at 34% to 36% decreases reflects the same structural logic: the addressable audience has migrated to streaming, and the Mediaocean survey is documenting the budget following it.
Adobe’s impulse buying data and the collapse of the purchase funnel
Adobe on June 17, 2026 published research showing that 86% of 1,003 surveyed US consumers make at least one unplanned online purchase per month, at 95% confidence with a 3% margin of error. More than 20% of respondents complete five or more such purchases within a single month. Nearly one in five spends over $1,800 annually on unplanned buys, a figure that sits entirely outside of planned household budgets.
The channel driving this is social video, followed by flash sales and product recommendations. Gen Z respondents showed a distinctive pattern: a significant proportion cited impulse purchases explicitly as stress relief, framing unplanned spending as an emotional response rather than a discovery event. That framing is relevant to campaign strategy because it changes what creative and messaging work for that audience segment. Aspiration and identity signals perform differently from product-led communication when the trigger is emotional rather than rational.
The speed dimension is where the data connects most directly to infrastructure. The study found that 48% of beauty category consumers complete a purchase within minutes or seconds of first encountering a product. The equivalent figures are 37% for health and wellness, 33% for apparel, 25% for toys and games, and 15% for software. For the fast categories, the funnel model of awareness-consideration-conversion has collapsed into a single moment. A viewer encounters a product in a short-form video, evaluates it in seconds, and either buys or does not. The gap between impression and transaction is measured in seconds, not days.
That speed requirement is precisely what agentic systems are built to address. A human campaign manager reviewing yesterday’s performance data and adjusting bids manually cannot operate at the latency required when a beauty brand’s conversion window is seconds wide. The systems that DoubleVerify and LiveRamp were building on June 17 are, in part, answers to the consumer behaviour Adobe documented on the same day.
The measurement problem that impulse buying creates is addressed by two related announcements published June 17 on PPC Land. Cint merged brand lift and sales lift into a single live dashboard inside Lucid Measurement, allowing US advertisers to optimise active campaigns against purchase data simultaneously rather than sequentially. The old workflow involved running a campaign, then measuring brand lift as a leading indicator, then waiting for sales data as a lagging confirmation. The Cint update collapses that sequence into a concurrent view. A related announcement the same day: Basis integrated Cint brand lift into omnichannel campaigns, giving advertisers real-time awareness, ad recall, and purchase intent data while ads are still running, across display, video, audio, CTV, and native.
The Trade Desk’s integration of Adsquare real-world location data into Audience Unlimited, also announced June 17, closes a complementary measurement gap. Audience Unlimited, which launched in September 2025 and restructured how The Trade Desk handles third-party data, previously helped advertisers identify and activate audiences with better precision. The Adsquare integration adds a feedback loop: location-derived store visit data can now tell advertisers whether those audiences converted in physical retail. The integration is available through The Trade Desk’s Kokai platform and connects to what the company describes as real-world business outcomes, specifically store visits and in-store conversions. For impulse-driven categories like beauty and food where the purchase often happens in a physical store after a digital impression, that attribution bridge is the only mechanism for closing the measurement loop.
Pixalate’s OpenEPG and what the bidstream reveals about streaming audiences
On June 18, 2026, Pixalate launched the OpenEPG Index 1.0, a public ranking of streaming TV content built from real-world open programmatic ad spend and audience reach data, covering both mobile and connected television. The index covers 5,108 unique shows across 224 streaming channels in all 210 US Nielsen Designated Market Areas, drawing on bidstream data from January through May 2026. The underlying OpenEPG 1.0 Analytics infrastructure has catalogued 12,875 unique television shows across 318 channels.
The measurement architecture sets it apart from every existing benchmark. The system operates entirely from standard bidstream Bundle ID signals in the open exchange, requiring no publisher opt-in and no custom data agreements. Every ad call in the open programmatic ecosystem carries a Bundle ID identifying the app or channel where the impression is served. Pixalate maps those Bundle IDs to specific shows and ranks shows by the volume of open-exchange ad spend they attract and the consumer reach they deliver. The measurement happens from the buy side, not the sell side. No publisher needs to agree to participate, and no platform needs to share proprietary viewership data.
Nielsen’s public Streaming TV Top 10, the dominant industry benchmark, ranks on-demand content by minutes viewed on television sets. It does not cover mobile viewing, FAST channel audiences, or the substantial portion of CTV inventory that transacts through the open exchange. The OpenEPG index covers all three simultaneously. The top five shows in the OpenEPG Index 1.0 for May 2026 by open programmatic spend were dominated by news programming and weather content, a finding that diverges from the entertainment-heavy Nielsen Streaming Top 10 and reflects where open programmatic dollars actually flow versus where direct-deal premium video spending concentrates.
The practical value of that divergence for media buyers is significant. A planner allocating CTV budget across the open exchange using Nielsen’s top-10 as a proxy for audience value is looking at a map of premium direct-deal inventory. The OpenEPG index shows what is actually trading in the open marketplace, at what price, and with what reach. Those two pictures are not the same.
The bidstream approach also creates a different kind of competitive intelligence. Because the index is built from actual ad spend signals rather than declared viewership data, it reflects where advertisers have chosen to put money, not where platforms say audiences are watching. That is a meaningful distinction in an environment where streaming platforms have commercial incentives to present their viewership as broadly as possible. An index built from buy-side spend signals cannot be inflated by a publisher’s measurement methodology. The spend either happened at the bundle ID level or it did not.
Pixalate’s coverage of 210 US Nielsen Designated Market Areas simultaneously, rather than providing only national averages, is also significant for local and regional campaign planning. A national DMA breakdown allows buyers to see not only which shows are attracting open programmatic spend nationally but which shows are capturing spend in specific local markets. That granularity is not available in any existing streaming measurement framework. For advertisers whose campaigns have regional concentration, whether because of retail footprint, regulatory variation by state, or local market competitive dynamics, show-level spend data at the DMA level is a fundamentally different planning input than national viewership averages., which PPC Land covered in detail on June 15 and June 17, gives the Pixalate launch additional structural relevance. When Fox Corporation closes its acquisition of Roku, a single buyer will control The Roku Channel, Tubi, and Fox’s live broadcast inventory, with first-party data from over 100 million global streaming households. That consolidation makes independent measurement from bidstream signals a more consequential tool. Bidstream-based measurement from the buy side functions as an independent check on self-reported audience data from a platform with a commercial interest in its own numbers. Advertisers buying Roku or Tubi inventory who can cross-reference OpenEPG data against what the platform reports will have independent leverage in planning and negotiation.
AI search citations, CTV attention, and the content strategy implications
Two data releases on June 17 from PPC Land describe the information environment that now shapes where marketing budgets can and cannot work.
NP Digital’s survey of 500 marketers, conducted in May 2026, ranks 11 content types by AI search citation performance. Original research scores 82%. Comparison content scores 76%. Rankings and best lists score 57%. FAQs score 41%. How-to content scores 39%. Community and forum participation scores 28%. Generic blog posts score 25%. Definition and explanation pages score 22%. Opinion and thought leadership scores 16%. Product pages score 14%. Video content scores 2%.
The 80-percentage-point gap between first and last is not a gentle slope. There is a cliff between the top two formats and everything else, and a second cliff at the bottom where video sits almost entirely alone. The structural logic, as NP Digital frames it, is that AI search engines preferentially cite content they cannot generate themselves. Original research contains data, methodology, and proprietary conclusions. Comparison content makes claims about relative performance that require empirical backing. Both formats have something to offer that a language model cannot fabricate without a traceable source. Video, from the perspective of a text-based AI system processing queries, offers almost nothing citable regardless of the quality of the underlying content.
The implication for marketers who have oriented their content investment toward YouTube, short-form social video, and podcast production is that their AI search visibility may be close to zero, regardless of how well that content performs on its native platform. A brand with 500,000 YouTube subscribers and a library of product explainer videos is not building AI search presence in any meaningful sense. The organisations building that presence are those running original surveys, publishing proprietary data sets, and producing comparison content with documented methodology.
The structural logic, as NP Digital frames it, is that AI search engines preferentially cite content they cannot generate themselves. Original research at 82% and comparison content at 76% both require data, methodology, and conclusions that a language model cannot fabricate without a traceable source. Generic blog posts at 25%, definition pages at 22%, and product pages at 14% are all content categories that a language model can produce at scale without external attribution. The AI system has no incentive to cite what it can generate. It has every reason to cite what it cannot.
The community and forum category scoring 28% reflects a different dynamic. Forums contain first-person experience, local knowledge, and unstructured opinion that language models cannot replicate at the granularity of a specific product used in a specific context. Reddit discussions, niche community platforms, and professional forums carry authenticity signals that AI systems have learned to weight. That finding is consistent with the observable behaviour of AI search interfaces that regularly surface Reddit threads in their citations for consumer product queries.
Opinion and thought leadership scoring 16% is perhaps the most surprising result for the many organisations that have invested in executive ghostwriting, LinkedIn thought leadership programmes, and editorial commentary as a form of content marketing. Opinion content is, by definition, content that expresses a perspective a language model could construct without an external source. Its low citation rate in AI search may reflect not the quality of the thinking but the structural indistinguishability of generic opinion content from AI-generated opinion content.
The practical implication is a significant reorientation of content investment priorities for any brand that cares about AI search visibility. The survey does not suggest that video or thought leadership content has no value. It suggests that those formats serve social distribution and brand building on their native platforms, while doing almost nothing for AI search citability. Brands that want both will need two separate content strategies with separate investment lines, rather than assuming that good content in one format translates to presence across all surfaces.
That finding connects back to the impulse buying data from Adobe in a specific way. If beauty and apparel consumers are converting from social video in seconds, social video content is still commercially essential. But if the pre-purchase research phase, however abbreviated, now runs through AI search interfaces, then the content that earns a citation in those interfaces becomes the first touchpoint in a funnel that closes seconds after a short-form video. Owning the AI search moment does not replace owning the social discovery moment. Both matter. The content formats that serve each are almost entirely different.
The WunderKIND attention benchmark data, published June 17 by PPC Land, addresses a format-specific question within CTV rather than AI search. WunderKIND Ads released the first programmatic CTV pause ad benchmarks, drawing on millions of impressions from campaigns running across Dish, Philo, and Plex, with TVision providing second-by-second attention measurement. TVision, now a Viant Technology company after Viant closed its acquisition in May 2026, captures eyes-on-screen attention at the person level, covering four variables per second: content, delivery method, individuals present, and attention level.
The core finding is that pause ads delivered nearly 2x higher attention time than standard 60-second CTV spots across all 15 tested verticals. The category breakdown shows how consistent that premium is. Automotive pause ads generated 34.2 seconds of attention time versus 12.2 seconds for standard CTV spots, a difference of 180.3%. Technology pause ads generated 33.3 seconds versus 12.8 seconds, a difference of 160.2%. Restaurants generated 34.0 seconds versus 13.5 seconds, a difference of 151.9%. The premium is present across every vertical, which means it is not an artefact of a particular content environment. Pause ads capture attention because they appear at the moment a viewer is actively looking at the screen and choosing to interact. That attention quality is distinct from the passive exposure that a mid-roll ad produces when a viewer is waiting for content to resume.
The format availability across the open exchange, through publishers like Dish, Philo, and Plex rather than only walled gardens, means the attention premium does not require a direct deal to access. That is where Smartly’s June 17 integration with Roku Ads Manager via API becomes relevant. If performance marketers can now manage CTV campaigns through the same workflow tools they use for social, and if bidstream-based measurement from Pixalate gives them show-level performance data comparable to what social analytics provide, then the remaining friction in open-exchange CTV adoption is primarily internal budget classification rather than a technology or measurement gap.
The agency payment trap: a structural problem with a number attached
The Kaplan Group report, published April 13, 2026 and covered by PPC Land on June 17, draws on invoice data from more than 100,000 freelance engagements to map a cash-flow contradiction that has been widening for a decade. Average payment terms for agency fees rose from 45.7 days to 58.1 days between 2013 and 2019, a 27% increase in six years. The post-pandemic acceleration pushed a meaningful share of marketers to 90- or 120-day terms, with documented cases reaching 150 days.
The structural contradiction: agencies facing 90-to-120-day client payment terms are simultaneously compelled by an expanding body of state legislation to settle with freelancers in 30 days or face double-damages penalties. California’s Freelance Worker Protection Act, effective January 2025, requires written contracts for freelance work above $250 and mandates payment within 30 days of project completion or the contract date, whichever is earlier. Illinois, New York, and New Jersey have parallel statutes. The Association of National Advertisers data cited in the Kaplan Group report shows that finance and procurement leaders at publicly traded companies have explicitly connected extended supplier terms to working capital improvements that analysts reward. Delaying payment to agencies is treated as a treasury lever, not a commercial risk. The consequence passes directly to freelancers, who bear the gap between when the agency gets paid and when the law says they must be paid.
The Kaplan Group’s analysis finds that late-payment collections for marketing freelancers have increased substantially in 2025 and 2026 as state protections have created a legal pathway for contractors to pursue unpaid invoices through collections rather than informal negotiation. The combination of stronger legal standing and a commercial debt collection infrastructure that specialises in professional services claims has altered the risk calculus. For agencies that have historically managed cash flow by delaying contractor payments, the exposure is now legal rather than merely reputational.
The Kaplan Group report maps the debt cascade in specific sequence. A major brand pays its agency 120 days after invoice. The agency, to preserve the client relationship and avoid formal complaint, absorbs the delay internally. The internal absorption requires the agency to either access a credit line, delay its own supplier payments, or defer freelancer settlements. Freelancers are the last and most vulnerable point in that chain. They are typically individuals without treasury infrastructure, without access to commercial credit at institutional rates, and without the negotiating leverage to push back on a client relationship that the agency controls. The California statute and its equivalents change that dynamic by giving the freelancer a legal claim that is independent of the agency’s client relationship. The freelancer no longer needs the agency’s permission or cooperation to pursue payment. The legal mechanism bypasses the commercial dependency.
That change has not yet produced widespread litigation. The Kaplan Group describes the primary enforcement pathway as collections rather than lawsuits, which keeps the process faster and cheaper for contractors while still creating reputational and financial exposure for agencies. The net effect is that the structural cash flow problem marketing agencies have managed informally for years is now being formalised into a legal and financial risk. The combination of longer client payment terms at one end of the chain and shorter statutory freelancer payment windows at the other is a compression that will only tighten as more states adopt equivalent legislation. The IAS pre-bid data, the Guideline-Mediaocean API, the Cint dashboard, and the WunderKIND benchmarks are all technical stories. The Kaplan Group report is the week’s structural business story, and it sits alongside all of them without fitting neatly into any category. It is about money, and about who waits for it.
Also noted
June 17, 2026: IAS launched episode-level pre-bid optimisation for Spotify podcasts on The Trade Desk, covering 33 avoidance segments across 11 categories, available from July 2026, extending brand safety controls that previously existed only for display and video inventory.
June 17, 2026: Guideline linked MediaTools to Prisma by Mediaocean via API, automating plan-vs-actual reporting and eliminating manual reconciliation that agencies and brands previously managed through spreadsheet exports and re-entry. The integration maintains a live link between planned spend in Guideline’s MediaTools and actual spend tracked in Mediaocean’s Prisma, removing a reconciliation step that has historically required dedicated time at campaign completion.
June 17, 2026: Smartly integrated with Roku Ads Manager via API, allowing performance marketers to launch, optimise, and measure CTV campaigns using social workflow tools, lowering the operational barrier for teams moving CTV budget into the open exchange.
June 17, 2026: PPC Land covered Fox’s $22 billion Roku acquisition alongside Uber’s Offsite Ads launch, IAS Spotify controls, and X Ads Manager updates in a single briefing on the programmatic stack developments of the week.
June 17, 2026: The EU Court of Justice grand chamber ruled in case C-188/24 that EU member states can legally require age verification on pornographic websites established in other EU countries, a jurisdictional clarification with implications for how consent and age-verification obligations apply across borders in digital advertising contexts. The ruling establishes that the country-of-origin principle under the E-Commerce Directive does not prevent destination states from imposing access controls on content targeted at their populations, a principle that could inform future regulatory moves around ad targeting permissions across EU member states.