# Stagwell Buys Freestar: Agentic Media Buying Picks the Wrong Agent *A holdco just bought the sell-side OS of 1,000 publishers to feed its buying agents. The agent that actually consumes publisher content, the one writing the AI answer, still pays nothing.* Published: 2026-10-27 | Read time: 5 min read | Author: Smalk AI Research | Source: https://www.smalk.ai/blog/stagwell-freestar-agentic-media-buying-wrong-agent --- Stagwell bought Freestar on October 26, 2026 because agentic media buying needs supply it can see, price and learn from. The deal is rational, and it integrates the wrong agent: the holdco's buying agents now own the pipes to the human pageview, while the answering agents at Google, OpenAI and Perplexity keep consuming the same publisher pages for free. - Stagwell acquired Freestar, the sell-side operating system behind more than 1,000 publishers, to shorten its supply path and feed audience data into The Media Machine, its agentic buying system. - The inventory it bought is the human ad request, the unit Chartbeat says is losing 40.2% of its Google referrals a year at news and media publishers. - The inventory it needs, the publisher page as a source cited inside an AI answer, is not for sale through any sell-side stack, Freestar's included. - Two agent economies now run on the same publisher content: buying agents that pay per impression and answering agents that pay nothing. Only one has a market. - The brands and publishers that price the answering agent's consumption first will own the margin the holdcos are currently buying their way toward. ## What agentic media buying is, and what it needs from publishers Agentic media buying is the delegation of planning, bidding, optimization and reporting to software agents that act inside the buyer's stack instead of a human trader. Stagwell launched The Machine in January 2026 and The Media Machine in June 2026 as exactly that: a layer of agents wired into Google, Meta, The Trade Desk and others. Agents are only as good as the supply they can observe. That is why the second move after building a buying agent is always to own supply: Stagwell Curate in July 2026 to package publisher inventory before the bid, and now Freestar to own the sell-side code that creates the ad request in the first place. ## Vertical integration in the click economy and in the citation economy In the click economy, every buyer who scaled eventually integrated backward into supply. Holdcos bought data companies, then built curation marketplaces, then shortened paths under the banner of supply-path optimization. The logic never changed: control the pipe that produces the human impression and you control price and data. In the citation economy the pipe runs the other way. Publisher pages feed an answering agent, the answer feeds the user, and no ad request is ever generated. A holdco can buy every header-bidding wrapper on the open web and still own nothing on that path, because the path is a crawl, not an auction. Stagwell just vertically integrated into the economy that is shrinking, not the one that is forming. ## What AdExchanger reported, and what we verified AdExchanger's October 26, 2026 report, by Anthony Vargas, frames the deal as media buyers moving closer to publishers to shorten supply paths and access audience data for training AI agents. Stagwell's Owned Media CEO Ben Berentson said the company is buying Freestar's technology and client relationships, not the publishers, and keeping its leadership and staff. Price was not disclosed; Needham advised Freestar. One verification note on scale. AdExchanger puts Freestar's network at nearly a billion ad requests a day; Freestar's own site claims more than 50 billion monthly, roughly 1.6 billion a day, and Adweek's October 26 coverage repeats the 50 billion figure. We use Freestar's number. Freestar's September 2026 pubOS announcement also states it has driven nearly $1 billion in publisher revenue over a decade. > We view ourselves as being an advocate for publishers. > > — Ben Berentson, CEO of Owned Media, Stagwell, to AdExchanger, October 26, 2026 ## Why owning the publisher's pipes is the right move for the wrong agent ### The strongest case for Stagwell's bet The bull case is coherent and Berentson makes it well. Publishers with owned-and-operated ad tech have had mixed results; the agentic era rewards flexible data connections over owned stacks; a managed service like Freestar gives mid-tail and long-tail publishers a path to budget they could not reach alone. Add Stagwell's Future of News initiative, launched in 2024, and a buyer that routes more working media to the Associated Press, Reuters, Fortune and The Boston Globe looks like a publisher ally, not a predator. Digiday's same-day analysis adds the fair caveat: a buyer now owns the seller's sales desk, and no firewall has been disclosed. Publishers will want to see the arrangement pay before they take the advocacy on faith. That is a governance risk, not a strategic flaw. ### Why the inventory it bought is deflating and the inventory it needs is not for sale The strategic flaw is in the unit. Chartbeat's 2026 Publisher Playbook, published in September, shows Google search referrals to news and media publishers fell 40.2% between July 2025 and July 2026, after a 21.9% drop the year before, and that AI chatbots send 0.01% of publisher traffic. Every one of Freestar's 50 billion monthly ad requests is a human pageview. That is the asset deflating fastest on the open web. Meanwhile the same publisher pages are being read by a different agent. Google's AI Overviews, ChatGPT and Perplexity crawl, extract and cite the AP, Fortune and Sports Reference to build answers that never generate an ad request. That consumption is real, growing and unpriced. No sell-side OS exposes it, because no auction fires. Stagwell bought a seat at the table where the old inventory is sold and gained no seat at the table where the new inventory is consumed. ## What the Stagwell-Freestar deal means for brands and for publishers ### For CMOs, media buyers and agencies: your buying agent cannot reach the answering agent - Audit which of your agency's agentic tools touch the AI answer. Most plan and buy impressions; none place you inside the sources an answering agent cites. - Treat holdco supply integration as a yield story, not an AI Search story. Cheaper paths to a shrinking pageview are not visibility where your category is being decided. - Ask your agency for a line item that buys presence on the publisher pages AI engines read. If it cannot name the inventory, the capability does not exist yet. ### For publishers: the buyer bought your pipes, not your citations - Your monetization partner is now owned by a buyer. Ask for the firewall terms and for an audience-data clause that stops your first-party data from training a system that trades against you. - Separate your pageview inventory from your citation inventory in every forecast. The first is the one Freestar sells; the second is the one AI engines use, and it needs a rate card of its own. - Price being the source. Publishers that package their AI-cited pages as sellable, measurable placements turn the crawl from a cost into supply that buyers will pay to enter. ## Three signals that will decide whether holdcos get into the answer First, the Palantir axis: Stagwell expanded its agentic targeting partnership with Palantir on October 22, 2026, four days before Freestar. Watch whether that targeting ever models AI Search exposure or stays on impressions. Second, Google's October 15, 2026 changes to make sources more visible in AI answers: if source prominence rises, the value of being a cited page rises with it, and that value has no owner yet. Third, the firewall: whether Stagwell discloses one for Freestar within a quarter tells publishers how buyer-side ownership of sell-side tech will behave when the next holdco copies the deal. ## Conclusion Hold on to this: the holdcos are integrating backward into publisher supply because their buying agents need it, and the supply they are buying is the human pageview that AI Search is retiring at 40% a year. The agent that now reads the most publisher content is the one writing the answer, and no sell-side stack reaches it. That gap is the category Smalk AI builds for: Generative Engine Advertising, native ads for AI agents placed on the publisher pages AI engines cite, so brands buy presence where recommendations form and publishers earn from the agent that actually consumes their work. What to watch next: the first holdco that stops buying pipes to the pageview and starts buying placement in the citation. ## FAQ ### What does Stagwell's acquisition of Freestar mean for publishers? A media buyer now owns the technology that runs ad sales for more than 1,000 publishers, including the Associated Press, Fortune and The Boston Globe. Freestar's name, leadership and service continue, but publishers gain a partner with a buyer's interests and should ask for firewall and data-use terms. The deal does not touch how AI engines use publisher content. ### How is agentic media buying different from AI Search advertising? Agentic media buying delegates planning and bidding on existing ad impressions to software agents inside the buyer's stack. AI Search advertising places a brand inside the content an answering agent cites when it builds an answer, and pays the publisher whose page was used. The first optimizes a pageview; the second buys presence in a citation. ### Why does owning sell-side ad tech not help publishers monetize AI Search? Sell-side ad tech creates and sells an ad request when a human loads a page. AI engines consume publisher pages through crawls that never load an ad, so no request exists to sell. Monetizing AI Search requires inventory defined around the citation, not the impression. ### Should brands expect holdco agentic systems to deliver AI Search visibility? Not yet. Systems like The Media Machine integrate with Google, Meta and The Trade Desk to buy impressions, and acquisitions like Freestar add human pageview supply. None of them place a brand on the publisher pages that AI engines read to form recommendations, which is where AI Search visibility is actually decided.