A mortgage campaign has cleared compliance. Its headline is supportable, the representative information is correctly presented and the landing page matches the offer. The advertiser is licensed, the audience is permitted and the campaign settings reflect the approved market.
For most channels, those checks would describe much of the approval decision but advertising on ChatGPT adds another: whether the conversation beside which the advert might appear is suitable for advertising.
OpenAI's advertising policies treat the advertiser, creative and landing page separately from the conversational placement. The advertiser can establish the facts supporting the campaign, while OpenAI decides whether an individual conversation is an appropriate place to show it.
That division has a consequence for regulated marketers. A campaign approval now depends partly on a platform policy and control system that the firm does not operate, so a material change to those controls can change the basis of approval without altering the advert itself.
One campaign, two approval decisions
OpenAI's policy, updated on 31 August 2026, describes review at three levels: the advertiser, the creative and landing page, then the placement. Operationally, the first two establish whether the business and campaign are eligible, while the third determines whether an approved advert can appear beside a particular conversation.
Those decisions can produce different answers. A mortgage provider may be permitted to advertise and its campaign may meet every content requirement, while the circumstances of a particular interaction make advertising inappropriate.
This is not a new principle of regulated marketing, which has always considered audience, medium and context. The development is that an AI platform can interpret a live interaction and use that interpretation during delivery. Context becomes part of the placement control rather than simply a characteristic of the publication or audience segment.
OpenAI's changelog shows how specific that distinction has become. Medical, legal and financial advice contexts are no longer categorically blocked from advertising by default, yet sensitive conversations and other prohibited contexts remain ineligible. A regulated subject may be permitted without every conversation about that subject becoming a permitted placement.
What the advertiser must establish
Much of the review remains familiar. Claims need evidence, required qualifications must be sufficiently prominent and the product must be appropriate for the intended audience and market. The advertiser must also hold any licence or professional registration the campaign requires.
The landing page belongs inside the same decision because OpenAI requires consistency across the complete advertising experience. An approved creative cannot link to a destination that introduces disallowed content, while a modestly worded advert cannot cure a page that makes stronger claims, removes a material limitation or presents a different product.
Regulated categories receive additional scrutiny. OpenAI may approve ads from eligible advertisers in financial services, healthcare and legal services through manual, case-by-case review.
In the US, potentially permitted financial categories include mortgages, insurance, investment services, credit cards and personal loans, and proof of licensing may be required. Financial-services advertising outside the US is generally prohibited at present. Certain US healthcare services may also be permitted, while legal-services advertisers must be licensed to practise in the jurisdiction where an advert is shown.
These are OpenAI's platform policies rather than regulation. Platform approval confirms that OpenAI considers the advertiser and campaign eligible under its current rules; the firm must still reach its own conclusion about the communication, target market and channel.
The placement decision belongs to the conversation
The approved mortgage advert remains unchanged, yet there is a practical difference between showing it beside a general exchange about mortgage types and showing it during an interaction suggesting acute financial distress or emotional vulnerability.
The example does not predict how OpenAI would classify either conversation. It illustrates why suitability cannot always be established by examining the advert alone.
OpenAI's placement policy excludes advertising from inappropriate and sensitive contexts, including personal, high-stakes or emotionally vulnerable situations. Specific exclusions cover emotionally reliant interactions, mental and personal health conversations and other sensitive user journeys.
OpenAI says its advertising system can consider the context and intent of the current conversation alongside the creative, landing page, advertiser-supplied context hints and targeting selections when deciding which advert to show.
This may provide clearer controls than some fragmented programmatic environments because the category rules, destination requirements and contextual exclusions sit within one stated policy. Whether that is sufficient remains a decision for each advertiser.
The advertiser cannot reproduce that decision
The advertiser does not receive the conversation used to inform placement. OpenAI says advertisers have no access to users' chats, chat history, memories or personal details and instead receive aggregated, non-identifying performance information such as views and clicks.
That boundary is part of the product's privacy design. A firm can control the proposition, evidence, audience, destination and campaign configuration but cannot inspect the signal behind every impression or independently determine whether a conversation should have been classified as sensitive.
The answer is not to demand private conversations. The advertiser needs to decide whether the controls at the boundary are sufficient: which contexts the platform excludes, how regulated categories are treated, where campaigns can run and how policy changes or incidents will be communicated.
This recognises the division of responsibility without leaving a gap in governance. The advertiser approves the use it can control and records the external conditions on which that approval depends; the platform remains responsible for operating its conversational placement safeguards.
A platform-policy change can change the approval
The final advert remains important evidence, but it no longer records the complete basis of the decision. A useful campaign record should preserve the creative, claims evidence, qualifications, landing-page version, product scope, audience, geography, licences and campaign configuration.
It should also identify the version of OpenAI's policy considered during review and the placement controls on which the decision relied. If the organisation's conclusion depends on OpenAI excluding sensitive financial conversations, that dependency belongs in the approval record.
A landing-page revision may introduce a new claim, an expanded geography may alter licensing or category eligibility and a different context hint may broaden the intended delivery. A platform-policy revision can be equally material because it may change the exclusions supporting the campaign.
OpenAI's policy already has a version history covering changes to placement rules, regulated categories, advertiser requirements and review processes. For regulated marketers, an update may therefore require campaign reassessment rather than passive circulation as legal news.
Advertising on ChatGPT makes approval conditional on more than a finished asset. The firm is approving this campaign, for this product, audience, geography and destination, while a defined set of platform controls remains in effect.
As advertising becomes more contextual and adaptive, approval systems need to preserve that full proposition: what was approved, which external conditions supported the decision and what would make the reasoning unsafe to reuse.
What teams need to know
Can regulated businesses advertise on ChatGPT?
Some can. OpenAI may approve ads from eligible advertisers in financial services, healthcare and legal services on a manually reviewed, case-by-case basis. The permitted categories, licensing requirements and geographic restrictions differ, so platform eligibility must be checked for the specific product and market.
What does OpenAI review before an advert can run?
OpenAI describes review at three levels: the advertiser, the creative and landing page, then the conversational placement. An advert that passes the advertiser and content reviews remains eligible only in conversations that comply with the placement policy.
Can advertisers see the conversations beside their adverts?
No. OpenAI says advertisers do not receive users' chats, chat history, memories or personal details. Advertisers receive aggregated, non-identifying performance information. If a user chooses to message an advertiser through an advert, the advertiser sees only the messages sent directly to it.
What should marketing compliance approve?
The decision should cover the creative, claims evidence, qualifications, destination page, product and audience eligibility, geography, licences and campaign configuration. The record should also identify the OpenAI policy version considered and any platform controls on which the approval depends.
When should a ChatGPT advertising campaign be reviewed again?
A material change to the creative, landing page, product, audience, geography, context hints, targeting configuration or applicable platform policy should prompt reassessment. Complaints, platform warnings or evidence that the original approval assumptions no longer hold may also justify another review.