ChatGPT's new ad format puts paid placement beside citations

OpenAI has, for the first time, given advertisers a way to buy their way into a ChatGPT answer. The company's blog post describes a "visual ad format" appearing alongside chat responses, paired with new measurement tools, attribution partnerships and brand-suitability controls, the standard furniture of a mature ad platform. Google took thirteen years to build that furniture around search. OpenAI has built the frame in under three.

The significance is not the ad unit itself, which by OpenAI's own description sits beside the answer rather than inside it. The significance is sequencing. OpenAI spent 2023 and 2024 training enterprises to treat ChatGPT citations as a new, ostensibly neutral discovery channel, the organic listing of the LLM era. It has spent 2025 building the monetisation layer underneath that channel. Any company that has watched Google's SERP evolve from ten blue links to a page where the first screen is paid, Reddit threads, and "People Also Ask" boxes knows how this story tends to end. The neutral-seeming layer gets crowded first, then commercialised, then optimised against.

The attribution build-out is the tell

OpenAI's post is heavier on measurement infrastructure than on the ad unit itself: attribution partnerships, brand suitability tooling, reporting dashboards. That is not incidental. An ad product without measurement is a favour to a few launch partners. An ad product with measurement is a sales motion, one that requires proving incrementality to a CFO, which requires controlling the counterfactual, which requires OpenAI to know, with some precision, what would have been surfaced organically versus what was paid for. Building that infrastructure is also, functionally, building the capability to distinguish and rank paid from organic placement inside a single answer. Once that distinction exists as a product feature, it becomes a lever OpenAI can pull in either direction: toward advertisers, or toward publishers it wants to keep onside.

For brands that have spent the past eighteen months chasing "share of model," as the industry has taken to calling citation frequency in LLM answers, this changes the unit economics of the chase. Earned citation and paid placement will now sit on the same screen, competing for the same attention, inside the same visual hierarchy. A bank's carefully placed thought-leadership citation and a rival's purchased unit will appear in the same breath. OpenAI says the ad format is visually distinct from chat responses; Google said the same about search ads in 2007, and about AI Overviews ads more recently, and the line between "distinct" and "adjacent enough to borrow credibility" has eroded every time.

Who actually has to worry

Three sectors should read this announcement closely rather than filing it under "ad tech news." Financial-services brands have spent heavily on earned citation in ChatGPT precisely because the channel felt uncorrupted by media spend, a rare venue where a well-argued research note could outrank a bank with ten times the marketing budget. That asymmetry narrows the moment budget becomes a relevant input again. Multilaterals and UN agencies, which have no advertising budgets to speak of and have relied entirely on citation quality and source authority to appear in ChatGPT's answers on development finance or climate risk, now compete for visual real estate against organisations that can simply buy adjacency. Industrial groups face a subtler problem: procurement and specification queries, the ones that currently reward technical documentation and ISO-grade precision, are exactly the high-intent queries an ad platform will want to monetise first, because commercial intent is what every ad platform sells.

OpenAI will insist, as every platform insists at this stage, that ads remain clearly labelled and organic answers remain untouched by payment. That was also the pitch for Google's first text ads. The pitch holds right up until the quarter revenue growth slows and some beleaguered product manager is asked to find another two points of monetisation from the answer page. The infrastructure described in this announcement, the attribution, the brand-suitability tooling, the measurement partnerships, is precisely the infrastructure that makes that later expansion easy rather than hard.

The practical implication for comms and marketing leaders is immediate, not speculative. Earned visibility in ChatGPT just became a contested commercial asset rather than a free byproduct of good content. Brands that treated LLM citation strategy as a content-quality exercise alone now need to treat it as a competitive-media exercise too, because the organisation sitting next to their citation may simply have bought the seat.