Programmatic advertising now accounts for 91% of all digital display spending in the US, according to eMarketer's 2024 Digital Advertising Forecast. AI has fundamentally changed how DSPs optimise delivery, how contextual targeting operates after cookie deprecation, and how brand safety is managed at scale — creating both significant efficiencies and new risks for advertisers who do not adapt their strategies.
Programmatic advertising in 2025 looks fundamentally different from the environment that existed three years ago. Third-party cookie deprecation has reshaped audience targeting. AI has taken over real-time optimisation in ways that remove transparency from the buying process. New deal structures have emerged to compensate for signal loss.
Navigating this environment well requires understanding what the AI is actually doing inside DSPs, where the leverage points are, and how to protect brand safety when automation controls placement decisions.
How Has AI Changed DSP Optimisation in the Post-Cookie Era?
Demand-side platform AI now operates primarily on contextual and first-party signals rather than the third-party behavioural data that defined the previous decade. Google's deprecation of third-party cookies in Chrome through 2024 and 2025 removed the cross-site tracking that enabled user-level audience targeting. According to eMarketer ([eMarketer Digital Advertising Forecast](https://www.emarketer.com), 2024), the transition has shifted DSP optimisation weight toward page-level context, supply-path quality, and first-party data matching — a fundamentally different signal mix.
DSP AI now makes impression decisions based on contextual relevance scoring (what content is on the page), publisher quality signals (engagement rates, viewability, brand safety scores), deal-specific performance history, and any first-party audience data you provide through clean room partnerships or publisher direct integrations. The algorithm's job has not changed — maximise conversion probability per impression — but the data inputs have.
The practical implication: advertisers who invested early in first-party data infrastructure — CRM integrations, data clean rooms, publisher direct relationships — are significantly outperforming those who relied purely on third-party audience segments. The signal advantage has shifted toward data-rich advertisers and away from data-poor ones.
The post-cookie programmatic landscape rewards first-party data investment above all else. Advertisers who built clean rooms, CRM integrations, and publisher direct relationships before cookie deprecation now have a structural performance advantage that competitors cannot quickly replicate.
How Does Contextual Targeting Work in AI-Driven Programmatic?
Modern contextual targeting is far more sophisticated than keyword-based adjacency matching. AI-powered contextual tools — built into DSPs like The Trade Desk, DV360, and Xandr — analyse full page content using natural language processing to understand semantic meaning, topic relevance, and sentiment context. According to a 2024 analysis by Integral Ad Science ([IAS Industry Pulse Report](https://integralads.com/research/), 2024), AI contextual targeting now achieves audience relevance accuracy within 15% of cookie-based behavioural targeting — a significant reduction in the performance gap that existed when contextual was re-emerging in 2022.
The strongest contextual strategies combine topic-level targeting (broad category alignment) with sentiment-level filtering (positive or neutral sentiment only) and brand safety adjacency rules (excluding content categories that conflict with your brand position). This three-layer approach protects against the most common contextual targeting failures: appearing in relevant but brand-unsafe contexts, or appearing in safe but low-relevance content that produces poor engagement.
What Are Deal IDs and How Should You Structure Your Private Marketplace Strategy?
Deal IDs are unique identifiers for programmatic deals negotiated directly between advertisers (or their agencies) and publishers. Private Marketplace deals (PMPs) and Programmatic Guaranteed deals give advertisers access to specific publisher inventory at agreed CPMs or pricing floors, bypassing open exchange competition. According to the Interactive Advertising Bureau ([IAB State of Data Report](https://www.iab.com/resources/), 2024), PMP and Programmatic Guaranteed deals now represent 47% of programmatic spending — up from 28% in 2021 — as advertisers seek inventory quality assurance and first-party data access that open exchange cannot provide.
A strong deal ID strategy addresses three objectives: quality assurance (accessing premium publisher inventory with verified viewability and brand safety standards), data enrichment (publishers with strong first-party data offer deal IDs that include audience matching against their registered user base), and scale efficiency (PMPs reduce auction competition for the most valuable inventory, often producing lower effective CPMs than open exchange for equivalent audience quality).
For most mid-market advertisers, a practical deal strategy starts with 3 to 5 premium publisher PMPs in your category's most relevant content environment, combined with open exchange for reach and frequency management. Layer in first-party data-enriched deals with major publishers (news, streaming platforms) as your data clean room infrastructure matures.
Curated Marketplaces and Supply Path Optimisation
Supply Path Optimisation (SPO) is the practice of reducing the number of SSPs you buy through, prioritising direct or near-direct paths to publisher inventory. Fewer intermediaries means lower fees and better data signal preservation. DSP AI benefits from SPO because clean supply paths carry more reliable impression-level data — viewability signals, engagement history, brand safety classifications — than deeply intermediated paths where data gets stripped or altered in transit.
How Do You Manage Brand Safety in AI-Optimised Programmatic Campaigns?
Brand safety in programmatic requires a layered approach because AI optimisation does not inherently prioritise brand safety — it prioritises conversion probability. An impression may score high on conversion likelihood but appear on content that damages brand perception. According to a 2024 IAS study ([IAS Brand Safety Report](https://integralads.com/research/), 2024), 66% of advertisers experienced at least one significant brand safety incident in programmatic campaigns in the past 12 months, with AI-optimised open exchange buys representing the highest risk category.
Effective brand safety management at scale uses three layers simultaneously. First, pre-bid filtering through your DSP's brand safety toolset — blocking content categories, applying contextual filtering, and setting keyword exclusion lists. Second, independent verification through IAS, DoubleVerify, or MOAT running alongside delivery to audit actual placement contexts post-bid. Third, publisher allowlists for sensitive brand contexts — rather than open exchange, specify the publishers you are willing to appear next to rather than relying on category exclusions alone.
Frequently Asked Questions
Is open exchange programmatic still worth buying in 2025?
Open exchange remains valuable for reach and frequency management at lower CPMs, but quality standards have declined as premium inventory shifted to PMPs and Programmatic Guaranteed. For brand-sensitive advertisers, use open exchange for low-funnel retargeting of known audiences where context matters less than identity matching. Reserve PMP and direct deals for upper-funnel brand-building where adjacency quality directly affects brand perception. ([IAB State of Data Report](https://www.iab.com/resources/), 2024)
How do data clean rooms work for programmatic audience targeting?
Data clean rooms are secure environments where two parties — typically an advertiser and a publisher — can match their respective first-party data sets without either party exposing raw individual-level data to the other. The output is an anonymised audience match that enables targeting against publisher-verified users who also exist in your CRM. Google Ads Data Hub, Amazon Marketing Cloud, and LiveRamp's Clean Room are the most commonly used options. Clean rooms effectively restore some of the targeting capability lost with third-party cookie deprecation. ([IAB State of Data Report](https://www.iab.com/resources/), 2024)
What is the difference between PMPs and Programmatic Guaranteed deals?
Private Marketplace deals (PMPs) are non-guaranteed — you are given priority access to inventory at an agreed floor price, but impression delivery is not committed. Programmatic Guaranteed deals are contractually committed delivery at a fixed CPM, functioning like traditional direct buys executed through programmatic pipes. PMPs offer flexibility with quality assurance. Programmatic Guaranteed offers delivery certainty for campaign flights where consistent impression volume matters — brand launches, time-sensitive promotions. Most advertisers use both in combination. ([eMarketer Digital Advertising Forecast](https://www.emarketer.com), 2024)
