For two years the knock on advertising inside AI assistants was simple: nobody could prove it worked. That objection is now being dismantled, and the company doing it is not OpenAI. It is the measurement plumbing arriving around the chatbot. The shift that matters is not another ad unit in ChatGPT. It is that the conversation surface is being wired into the same conversion infrastructure that governs the rest of the open web, which is the moment a novelty becomes a line item.
On June 10, as Digiday reported, LiveRamp became the first independent ad tech company to pipe conversion data into OpenAI’s advertising system, connecting its Conversions API Hub to ChatGPT ads. The mechanics are deliberately mundane, which is the point. A retailer’s transaction record, the product, the price and the date, moves through an encrypted server to server connection, is matched on a hashed email, and is forwarded to OpenAI, which controls the attribution methodology. Support for LiveRamp’s RampID identifier follows in a later phase, and the rollout starts in the United States with select mutual clients before extending to Europe. Travis Clinger, LiveRamp’s chief connectivity officer, said the company expected its first advertiser to go live within the week and framed the logic bluntly: transaction data, he argued, is far more valuable than click based data. The example he reached for was a familiar one, a Nike ad surfaced in a conversation that precedes a purchase made two days later in a physical store, a connection a click never captures.
Why a measurement pipe changes the category
Chatbot ads entered testing in February, and the surface is large. ChatGPT reports roughly a billion users, and OpenAI is preparing for a closely watched public offering where advertising revenue will be scrutinized line by line. None of that scale converts into committed budget without accountability. Performance marketers do not move money toward impressions they cannot tie to outcomes. The LiveRamp connection lets a brand see that an ad surfaced in a conversation preceded a purchase, including a purchase that happens two days later in a physical store, which is precisely the kind of closed loop that justified search and social budgets in the first place.
Advertisement
300 × 250
The provider lineup is filling in around the same need. Criteo became OpenAI’s first ad tech partner in March and has since lowered its minimum spend, and The Trade Desk has reportedly held partnership discussions. OpenAI, meanwhile, is widening the canvas. Per MarTech.org, it is testing multi advertiser placements that put more than one brand inside a single sponsored slot, priced through a second price auction, and it has added campaign tooling that reads like a maturing platform: cloning CPM campaigns into CPC with a click, average daily budgets for weekly pacing, bulk editing and custom CPM bid caps. Targeting has expanded into the United Kingdom, Japan, South Korea, Brazil and Mexico. Each of these is incremental on its own. Together they describe an ad business assembling the parts it needs to be bought programmatically rather than as a test.
It helps to remember how fast this arc has moved. Ads only entered testing inside ChatGPT in February, which means the platform has gone from no advertising to multi advertiser auctions, expanding geographies and an independent measurement partner in roughly four months. That compression is itself the signal. Search advertising took years to build its measurement and attribution scaffolding; the conversational surface is assembling the same apparatus in a single quarter because the playbook already exists and the participants, OpenAI, established ad tech vendors and the holding companies, are simply porting it onto a new interface. The unanswered question is not whether the plumbing can be built. It is whether conversational context, where a user is asking for help rather than browsing, produces the kind of high intent moments that justify premium pricing, or whether ads inside an assistant feel like an intrusion that suppresses the very engagement that makes the surface valuable.
The neutrality question sitting underneath
There is a complication that marketers should not wave away. LiveRamp is being acquired by Publicis Groupe in a 2.2 billion dollar deal expected to close by year end, which means the company verifying advertising outcomes is about to sit inside a holding company that buys media. Clinger said maintaining LiveRamp’s neutrality is an explicit commitment. Independent voices were less relaxed. The ad tech consultant Shirley Marschall warned that a partner with revenue or equity entanglement essentially grades its own homework. That is not a reason to dismiss the integration, but it is a reason to treat OpenAI controlled attribution and an agency owned measurement layer as data to interrogate, not gospel to report upward.
Underneath the commercial questions sits an identity model that marketers should understand before they fund anything. The entire loop depends on matching a hashed email from a retailer’s transaction file to a user OpenAI can recognize, which is the same deterministic matching that powers the rest of the post cookie ecosystem. That has two consequences. First, the quality of the measurement is only as good as the match rate, so brands with strong first party data and clean customer records will see a sharper signal than those relying on thin or stale files. Second, it concentrates a great deal of sensitive linkage inside a small number of intermediaries, which is precisely why the pending move of LiveRamp into an agency holding company is not a footnote. The measurement layer for AI advertising is being built on identity resolution, and identity resolution is where both the accuracy and the governance risk live. Marketers who treat this as a media question and skip the data governance review are evaluating only half of what they are buying.
What it means for the marketing leader
Treat conversational ad surfaces the way you treated retail media three years ago: real, fast growing, and not yet standardized. The practical work is unglamorous. Confirm whether your conversion data can flow through a CAPI style connection without exposing customer records, since hashed matching is the entire trust model here. Insist on incrementality testing rather than accepting platform reported attribution, especially while the measurement vendor and the ad seller share commercial interests. And recognize that this is the same pattern playing out across once organic surfaces, where paid AI search is becoming its own discipline. The assistant is not a separate universe. It is becoming another managed channel, with the same demands for proof.
The strategic read is that measurement, not creative or inventory, is the gate on AI advertising budgets, and that gate is now propped open. The brands that win the early efficiency will be the ones that already run disciplined incrementality programs and can port that rigor onto a new surface, rather than the ones dazzled by a billion users. A channel becomes spendable the day you can argue about its ROI. That argument just started.
