Programmatic advertising has changed a lot in the nearly 20 years since the first RTB ad auction.
On the surface, programmatic is just the automated buying and selling of digital ads. Under the hood, there’s a complex, high-speed auction system that decides in milliseconds which ads appear across display, video, CTV, audio, retail and other online media.
And now, generative AI is starting to transform that process.
Programmatic systems need to evaluate impressions in the milliseconds it takes for pages to load and ads to render with numerous players in the middle.
Publishers generate bid requests for their inventory. Sell-side platforms (SSPs) package and route them to buyers. Demand-side platforms (DSPs) decide whether to bid and at what price. Agencies set up and manage campaigns based on what marketers want to achieve, and ad exchanges or SSPs run real-time auctions that match supply and demand across billions of daily transactions.
But, as media buying has become more automated, the ecosystem has gotten more fragmented. Publishers and marketers struggle to maintain visibility into how media is bought, sold and measured across a tangled web of third-party intermediaries, from buy-side and sell-side platforms to curation, data, measurement and verification partners.
All of this would be hard enough to manage on its own. Layer generative AI on top, and the pressure to clean up how programmatic works – and who gets paid what – is suddenly much higher.
Scale created a signal problem
This pressure lands hardest on the sell side.
Many publishers rely on programmatic as their core monetization engine, rather than direct ad sales, and they run the gamut from smaller pubs that rely on third-party systems to run their programmatic business to more sophisticated sellers that have developed their own internal ad tech stacks.
EssentiallySports is an example of the former, sitting atop a tech stack that includes ad servers, header-bidding wrappers and multiple SSP integrations, said Suryansh Tibarewal, the company’s co-founder and growth lead.
Ad servers handle the real-time decisioning and render the winning ads. Header-bidding wrappers set the rules for how different demand sources compete. And many publishers plug into several SSPs, each with their own demand partners and proprietary auction algorithms, to try and maximize revenue from their inventory.
In practice, each page view on a publisher’s site or app generates a bid request with contextual signals, such as content category, device type and user behavior. Requests are simultaneously broadcast to multiple SSPs, which act as supply-side decisioning layers.
SSPs then apply pricing floors – the minimum bid publishers are willing to accept – and set deal priorities and curation rules that determine which supply paths get priority access to certain inventory before routing ad impressions into exchanges or direct demand paths.
SSP OpenX, for example, recently started reducing reselling layers and tightening supply paths so publishers retain more signal quality before impressions reach buyers, said Tyler Romasco, EVP of commercial.When impressions are passed across multiple resellers and intermediaries, Romasco said, it degrades the initial bid-request signal passed by publishers.
“In some cases, we’ve seen up to 72 companies touch a piece of inventory before it reaches a buyer,” Romasco said. “By the time that happens, you’ve lost a lot of the original signal.”
This fragmentation means publishers can no longer solely focus on monetization efficiency. They also need to worry about whether audience and engagement signals survive the journey through multiple hops in the supply chain.
At the same time, buyers are demanding more transparency into the inventory they’re bidding on.
Marketers want outcomes
Speaking of buyers – think brands, ad agencies and the DSPs that help them acquire ad space – their top priority is ad performance. Are the ads driving results?
For agencies, programmatic is the buying infrastructure they use to plan, launch and optimize digital media campaigns for advertisers. Havas Media Group, for example, runs campaigns in multiple DSPs and devises structured buying logic to manage goals across fragmented supply.
But what’s really changed for buyers, thanks to programmatic, is less how media is bought and more how success is defined, said Sarah Karges, SVP of performance investment at Havas Media Group.
“Clients want clarity on performance,” Karges said. “They don’t care how many intermediaries exist. They care whether the media is actually driving outcomes and whether they can trust the data behind it.”
DSPs, therefore, train their algorithms to win auctions for the inventory most likely to drive outcomes for buyers and factor in which supply paths deliver the cleanest data.
Within DSPs, agencies configure campaigns using parameters such as target audience segments and frequency caps that dictate how often to show an ad to the same user. Agencies also use DSPs to control their bid multipliers, auction pacing controls and the rules that dictate which ad creative to serve and when.
Yahoo DSP, for example, processes bid requests using its identity backbone and logged-in user graph. Advertisers can match impressions against authenticated audiences at scale while maintaining more consistent frequency and measurement, said Joshua John, senior director of DSP Strategy at Yahoo DSP.
Essentially, the DSP’s role is to collapse complexity into decisioning, John said. “We’re taking thousands of signals – audience, context, frequency, budget – and reducing it to a single bid decision in milliseconds.”
But agencies increasingly depend on external systems like attribution platforms and identity partners to figure out what happens after an impression is served. The decision to buy impressions happens instantly, but the proof of whether they worked only shows up later in separate systems.
“You can unify reporting layers,” Havas’ Karges said, “but the underlying execution is still happening across multiple systems that don’t fully talk to each other.”
AI is adding speed and new complexity
That’s the system programmatic built: powerful but stitched together. Now, generative AI is reconfiguring how that system runs – and it’s poised to both simplify and complicate the supply chain in new ways.
Across all players in the auction, AI is reshaping how programmatic functions by breaking workflows into modular components that are more interoperable across platforms.
“Programmatic was built to automate advertising, but there are still a lot of manual workflows inside it,” OpenX’s Romasco said. “AI is starting to solve some of those operational inefficiencies.”
SSPs are applying AI to predict revenue yield and aid traffic-shaping, meaning the process of identifying which inventory different buyers are likely to bid on, and sending those impressions to the relevant buy-side platforms.
These SSP improvements are helping publishers like EssentiallySports anticipate inventory value before it enters the auction, said Tibarewal.
Meanwhile, agencies are experimenting with AI-driven planning systems that simulate audience reach across DSP environments, such as Havas Media’s performance stack.
And DSPs are beginning to shift campaign planning, activation and measurement into modular AI-assisted workflows.
“AI allows us to break those functions apart so advertisers can automate pieces of the workflow instead of managing everything manually,” Yahoo’s John said.
But AI also introduces new supply-side risks.
AI-generated content environments and synthetic sites can now enter programmatic pipes at scale, forcing SSPs and DSPs to distinguish between scalable inventory and meaningful inventory in real time.
“AI can create a new dynamic site immediately, analyze an article from The New York Times, generate its own version and enter the ecosystem right away,” said John. “That is something an advertiser does not want to spend money on if there’s no human traffic actually going there.”
Cleaning the pipes
The next phase of programmatic will be to clean up the fragmentation introduced by automation. The industry is restructuring how inventory moves through the auction so that buyers have clearer visibility into where impressions come from and publishers can better preserve performance and audience quality.
The incentives are surprisingly aligned.
Publishers want stronger monetization without losing audience integrity. Agencies want measurable outcomes without losing trust in data. DSPs want cleaner decision inputs, and SSPs want more direct, higher-quality demand paths.
“At the end of the day, buyers spend more when they get performance,” OpenX’s Romasco said. “And publishers make more money when buyers spend more money.”
These goals have not changed; the plumbing has.
After years of layering automation on top of fragmentation, programmatic is now trying to reduce enough complexity to make the system understandable again, without breaking what made it scale in the first place.
