Testing generative AI in programmatic ad buying
At Marketecture Live in New York this week, The Trade Desk confirmed it is quietly testing a way for advertisers to build and run campaigns using generative AI. CEO Jeff Green told the audience the company has opened a closed beta that lets some clients create campaigns through Anthropic’s Claude, offering an early look at how large language models might be wired into programmatic ad buying. Reporting for Adweek by Trishla Ostwal first detailed the trial. (Source: Trishla Ostwal, Adweek). https://www.adweek.com/media/the-trade-desk-claude-campaign-tests/
What is being trialled
The experiment is small and intentionally opaque. According to Green, select advertisers are able to use Claude to generate campaign elements and make programmatic decisions inside The Trade Desk’s stack. The company has previously signalled plans for an agentic AI framework aimed at partners in 2026, and this closed test appears to be a preliminary step toward that broader vision. For background on Anthropic, see Anthropic’s corporate site. https://www.anthropic.com
Why the move matters
Programmatic advertising is highly automated by design – billions of bid opportunities arrive every hour and decisions must be made in milliseconds. Green argued that this environment is especially fertile ground for AI because models can act at scale and speed. If successful, AI-driven campaign creation could compress planning cycles, reduce human overhead and personalise creative and targeting across far larger audiences.
That is attractive to publishers, platforms and ad buyers, and it is precisely why major players are racing to develop agentic systems that can propose, test and optimise campaigns with minimal human supervision. Amazon, for example, is already opening parts of its ad stack to AI-driven workflows, and other ad tech vendors have flagged similar moves. See Amazon Ads for its platform overview. https://advertising.amazon.com
Economic and strategic tensions
Green used the stage to set out a wider economic case. He suggested many AI companies will look to advertising as a sustainable revenue route as they wrestle with capex and monetisation pressures. That framing is strategically useful for The Trade Desk: if AI platforms need advertising inventory and programmatic infrastructure, intermediaries that control buying workflows may gain bargaining leverage.
Green also renewed a pointed critique of Amazon’s long-term role in open web ad tech. He warned that antitrust exposure – the risk regulators attach to firms that combine retail, cloud and advertising power – could shorten Amazon’s willingness to operate an open demand-side platform. That argument reframes competition as not just technical but legal and political, and it is worth watching given regulators in multiple jurisdictions have already scrutinised Big Tech ad practices.
Regulation, privacy and industry friction
Bringing LLMs into the buying process raises thorny questions. Data privacy is front and centre for European advertisers and publishers – generative models trained on broad corpora could change how personalisation is implemented, and GDPR considerations will not be far from any deployment in Ireland or the EU. At the same time, the potential for opaque decision-making by agentic systems increases the need for auditability and accountability.
There is also industry friction to navigate. Agency groups have in recent months re-evaluated some integrations with The Trade Desk, and the platform’s push deeper into supply chains has triggered concern among holding companies. Green declined to comment on specific agency moves at Marketecture Live, but the tensions underline the wider trade-off: operational efficiencies versus control and transparency.
Practical implications for Irish advertisers
For marketers in Ireland and across Europe, the Trade Desk trial is both opportunity and warning. On the opportunity side, automation could lower barriers to sophisticated programmatic buys for mid-market advertisers, improving campaign performance and reducing agency fees. On the warning side, advertisers must insist on contractual clarity about data usage, model provenance and the ability to audit AI-driven decisions. Irish advertisers should weigh potential efficiency gains against compliance obligations under GDPR and any forthcoming AI regulation from Brussels.
What to watch next
– The Trade Desk’s promised agentic AI framework, currently pencilled for partners in 2026, and any public roadmap or launch dates. See The Trade Desk for company updates. https://www.thetradedesk.com
– Further partnerships or discussions with major AI providers – Green would not confirm reported talks with OpenAI at the panel.
– Moves by Amazon around its DSP offering and any regulatory developments that could force platform strategy changes.
– Agency and publisher reactions, especially whether agencies reconfigure integrations or demand new transparency guarantees.
Conclusion
The Trade Desk’s Claude trial is a credible early test of a larger play: to make programmatic buying more autonomous while positioning the company as infrastructure for an AI-first advertising economy. The risks are not trivial – regulatory scrutiny, privacy obligations and industry pushback could slow adoption. Nevertheless, for a sector built on automation, the logic of embedding generative AI is obvious. The balance now will be between speed to market and the governance frameworks that ensure those AI systems serve advertisers, publishers and users without undermining competition or privacy.
Reporting draws on Trishla Ostwal at Adweek (www.adweek.com). https://www.adweek.com/media/the-trade-desk-claude-campaign-tests/
