AdRoll’s route into ChatGPT arrived on 11 August, at a point when the channel is no longer hard to enter. The San Francisco advertising platform said a selected group of AdRoll and AdRoll ABM customers can test sponsored placements inside OpenAI’s assistant, with AdRoll managing those campaigns alongside their existing display, social and connected TV buys. The company describes the participants as small and mid-sized businesses, mid-market advertisers and business-to-business marketers.

The scarce commodity is no longer access. OpenAI opened its self-serve Ads Manager to all United States businesses on 5 May and removed minimum spend requirements, after a first pilot phase that trade reports said carried six-figure entry commitments. ChatGPT ads went live for UK users on 6 June, the first European market, and OpenAI’s own availability table, updated on 13 August, lists Ads Manager as available in the United Kingdom alongside eight other countries. AdRoll’s offer sits on top of that auction rather than opening it: campaign support, retargeting and attribution for teams with no spare specialist to run a channel test.

What does AdRoll’s ChatGPT pilot actually give advertisers?

AdRoll is offering managed access and measurement plumbing rather than a new ad format. The placements themselves are OpenAI’s: sponsored messages that appear below the end of a response when contextually appropriate, shown separately from the answer and, according to OpenAI, without influencing what the assistant says. AdRoll’s product page sets out what it adds around them. Advertisers can build AdRoll retargeting audiences from visitors who arrive through ChatGPT ad landing pages, track conversions using UTM parameters and existing AdRoll conversion events, and see results attributed to paid OpenAI campaigns inside the same reporting as the rest of their mix.

The measurement route matters more than it sounds, because ChatGPT sits early in the buying cycle and rarely takes credit under last-click rules. Attribution through UTM parameters and AdRoll’s own conversion events puts ChatGPT-sourced sessions in the same reporting view as paid search and display, which is what makes an assisted contribution visible at all. Studies of AI-referred traffic have found it converts at several times the rate of standard search referrals, so the volume a test produces is likely to matter less than what that traffic does afterwards.

Vibhor Kapoor, AdRoll’s chief executive, framed the pitch around the middle of the buying journey, arguing that conversational platforms let marketers take part in the moments when options are discovered and preferences form, rather than at the awareness or capture ends where existing channels already sit. Hans Fischmann, the company’s vice president of product, said the early signal was encouraging and that ChatGPT was showing up across both awareness and performance goals. AdRoll published no impression, click or conversion data alongside that claim, and has not disclosed which markets the pilot covers.

Can ChatGPT ads do account-based marketing?

ChatGPT’s advertising system carries one first-party targeting control, and it is built for lists far larger than a B2B target-account file. Custom Audiences, documented by OpenAI in July, let advertisers upload email addresses or phone numbers, raw or SHA-256 hashed, then include that audience in a campaign, exclude it, or apply a bid multiplier between 0.1x and 10x when a viewer matches. OpenAI’s documentation sets a minimum of 25,000 matched users per audience before it can be used, and recommends at least 100,000. The figure counts matched users, not uploaded rows, so invalid records, duplicates and non-matches are stripped before an audience reaches usable size.

That floor rules out the work most B2B teams mean by account-based marketing. A named-account list of 300 companies, or the 4,000 contacts inside them, cannot be used as an inclusion audience. Nor can a customer-suppression list from a company with a few thousand accounts. Everything below that threshold falls back to OpenAI’s default, which is contextual: ads are matched to the topic and intent of the live conversation using signals such as landing pages, titles and ad copy, with no job title, seniority or firmographic filter. The ABM in AdRoll ABM describes which customers may join the pilot. It does not describe account-level targeting inside ChatGPT, and the two are easy to conflate in a briefing note.

How much of ChatGPT’s usage is actually buying research?

AdRoll’s commercial case rests on third-party usage research it cites rather than produced. Its ChatGPT Ads page attributes to Presenc AI a finding that product research and shopping now account for 14% of ChatGPT sessions, up from 9%, with query volume growing 89% year on year. No sample size or methodology is linked from the page, and the split between consumer shopping and business software evaluation is not given, which is the split that decides whether the number means anything to a B2B advertiser.

The tier structure sets a harder limit. OpenAI shows ads only to logged-in adult users on the free tier and the lower-cost Go plan. Plus, Pro, Business, Enterprise and Education accounts see none, and accounts identified as belonging to under-18s are excluded. In practice that removes a meaningful share of the senior B2B buyers a vendor most wants to reach, because their employers pay for a business tier. The reachable audience skews towards individual practitioners and smaller firms, which fits the SMB and mid-market advertisers AdRoll has selected for the pilot more neatly than it fits enterprise pipeline work.

AdRoll has not said how long the pilot will run, what participants pay, or when access widens beyond the selected group. Existing customers who want in are told to contact their customer success manager.