The firm, traditionally associated with advertisement retargeting, is trying to reposition itself for an AI-driven business era, and its most current demonstrations suggest the business currently sees large language versions– not just merchants or demand-side systems– as the following major circulation channel for advertising and marketing. Far from dealing with LLMs as an existential risk to its efficiency organization, Criteo is leaning right into them, through early trying outs a (publicly unnamed) LLM to place its commerce dataset as the missing component that can make product discovery inside tools like ChatGPT or Claude in fact function.

The company has actually begun piping organized signals– relevance, trendiness, retailer-level performance– right into LLM environments using its Design Context Protocol server, effectively allowing any kind of representative inside those versions to hit Criteo’s API when recommending a product. The bet is that common web-crawl information is simply not good enough for high-fidelity business suggestions. If LLMs intend to play in retail media or product suggestions, they’ll require something closer to Criteo’s longitudinal, transaction-linked dataset.

“The ability in the engine drives the full funnel … our deep learning capability can make sense of scattered communications to reveal item suggestion chances,” clarified Criteo chief executive officer Michael Komasinski, at a current press event organized in New York City.

That’s the tactical tale the business is currently telling: its dataset, built over almost two decades, is a differentiator in a globe where AI comes to be the front-end, and ad-tech vendors end up being the framework providing importance behind the scenes.

Muscle mass memory

Criteo’s leadership is increasingly forefronting its origins– initially as a DVD referral engine, then as an efficiency advertisement technology company– to stress that the current shift toward LLM-driven suggestions is not foreign territory. The firm emphasizes that deep discovering work, much of it increased throughout the” signal loss age and Privacy Sandbox shift, now powers a wider, more adaptive optimization layer.

That engine underpins Criteo’s argument that it can operate throughout a full-funnel operations: recognizing spread browsing communications, presuming product fondness, and generating the embeddings that allow neural models to factor concerning relevance at range. This is the plumbing the business now wants to link into LLM-based commerce experiences.

Every one of this rests on a dataset it declares is under-appreciated throughout the community: 720 million day-to-day energetic users, more than a trillion bucks in observed on-line transactions, and billions of SKUs sewn together via retailer and brand integrations. To hear Criteo inform it, this is the type of behavioral splendor LLMs basically lack.

AI as process

Internally, the company is pressing AI deeper right into campaign set up and optimization. The “audience agent” tools are made to let a marketing professional define an objective in ordinary language– “help me sell extra camping knapsacks”– and immediately assemble the sections, reach curves, and relevance scoring that formerly called for sifting with taxonomies and manual logic.

Business Go– a fuller workflow tool that’s currently billed for a launch in Q 1, 2026– builds on this reasoning. It automates target market development, network mapping, and imaginative generation (text-to-image, image-to-video, URL-to-video) in something more detailed to 10 minutes than the multi-vendor, multi-day procedure lots of online marketers still withstand. Criteo asserts these automated campaigns supply materially greater return on advertisement spend, and says thousands have actually already run through the system.

Crucially, none of this depends on a user logging right into a Criteo UI. Via MCP, the exact same campaign-creation moves currently run inside ChatGPT or Claude. A project supervisor can rest inside an LLM, recommendation an advertiser ID and brief the system conversationally. The LLM passes directions to Criteo’s APIs and returns a completely assembled project that can then be fine-tuned or introduced inside Criteo’s platform.

Criteo CTO Todd Parsons included, “This is simply an instance of how we’re using AI to relocate our tradition items forward. In the past, you would need to look up our taxonomy, and you would have had to select or deselect a tons of audience qualities.”

This is where Criteo sees the near-term performance play: breaking down media preparing actions by utilizing AI to do the connective tissue job that was traditionally manual, sluggish and fragmented.

Criteo’s retail media consumers, meanwhile, are competing to layer AI chat experiences into their very own sites– a shift driven partially by buyer actions and partially by fear of delivering exploration to LLMs. Criteo is placing itself as the bridge in between retailer-controlled AI conversation and advertisement money making, making sure funded item positionings remain pertinent and do not threaten buyer count on.

One technical emphasis is giving sellers finer control over significance scoring– relocating from binary qualification to graded scoring where a brand may bid against nuanced relevance thresholds. It’s an effort to solve the long-lasting stress in retail media: stabilizing monetization with genuine usefulness to the buyer.

The “agentic” usage instance crosses both sides of your home. Whether the front-end is a merchant’s very own conversation interface or a general-purpose LLM, Criteo desires its business data to power the suggestion reasoning beneath it.

LLMs as the next ad network?

The business is clear that the test-and-learn phase with a significant LLM carrier is still very early. But the framing is informing: Criteo sees agentic buying as an incremental channel, not a cannibalistic one. If the performance holds up, the firm anticipates to designate budget plans the same way it currently reallocates between open-web and social performance– wherever product sales can be determined and validated.

Commercial versions are still liquid: Criteo speaks about every little thing from data licensing and pay-per-query designs to indigenous ad formats inside LLMs. Nevertheless, the thesis remains consistent: high-fidelity commerce signals will be required for high quality product exploration, and this creates a chance for an ad technology company with a robust suggestion stack.

For an organization that has spent years attempting to reshape its identity after the third-party cookie age, the LLM change offers something truly new to lean into– not equally as a defensive posture, but as a path right into an extra varied collection of revenue lines.

In short, Criteo is attempting to make certain that when AI becomes the store window, it is the vendor supplying the racks.