Advertisers in the practitioner discussions are comparing early experiences with OpenAI ads, and the dominant reaction is caution. Many participants report weak conversions, poor lead quality, short visits, questionable traffic, or costs they did not consider sustainable. Several of those accounts include campaign-level spend, traffic, or conversion observations, but they remain individual advertiser reports rather than a representative measure of the platform.

The dominant concern is unreliable performance

Across the discussion, advertisers most often describe results as volatile or uneconomic. Some commenters report spending without meaningful conversions, while others say the traffic they received did not retain or convert well enough to justify continued testing. These reports support a cautious early view, but they do not establish that the channel performs the same way across every market or campaign.

Measurement uncertainty makes that verdict less clean. Several participants report gaps between platform clicks and measured sessions, limited conversion reporting, or disagreements between store attribution and the ad platform's numbers. Commenters therefore argue that apparent campaign performance cannot always be separated from the quality of the measurement setup.

The strongest counterexamples are commercially meaningful

A minority of advertisers in the discussions report more promising outcomes. Their accounts include lower click costs than another paid channel in one niche campaign, legitimate leads or conversions across several accounts, sales associated through UTM tracking, and one claimed positive blended return. These less common reports directly challenge the idea that the channel is uniformly uneconomic, although the discussion does not establish how repeatable those outcomes are.

A separate evidence-backed counter-position focuses on attribution mechanics. Commenters describe cases where a click identifier may be lost during redirects or omitted from server-side conversion reporting. They recommend checking whether the identifier survives the landing path, using server-side tracking where appropriate, and allowing a longer attribution window. Under that view, at least some zero-conversion reports could reflect broken measurement rather than bad traffic alone.

Most commenters still treat it as an experiment

Several participants recommend assigning only a small experimental budget while controls, reporting, and inventory remain immature. Their advice is not that the channel can never work; it is that businesses needing dependable weekly leads should not yet treat it as predictable performance media.

Commenters also report that audience fit varies sharply. Some describe irrelevant placements or weak business-to-business volume, while another niche advertiser attributes a small set of engaged visits to detailed contextual instructions. the discussion therefore points to setup and market fit as possible reasons results differ.

Operational friction forms another theme. Participants report unexplained account denials, eligibility changes, inconsistent approval in restricted sectors, and an agency-access workflow they consider less mature than established advertising platforms. Those issues are separate from campaign economics but still affect whether advertisers can test the channel reliably.

What remains unresolved

The discussions do not settle whether OpenAI ads are broadly effective. They show a dominant set of negative or inconclusive early reports, meaningful positive exceptions, and plausible attribution failures that can change how results are interpreted. This report reflects only the discussions reviewed, not the wider practitioner community.