OpenAI quietly shipped another update to its ChatGPT advertising policy this month, and it landed in the same week that trade press reporting exposed how thin the platform's brand safety controls still are. The v1.6 policy, dated September 2026, was updated to clarify OpenAI's right to decline ads where they conflict with its advertising principles, business interests, or competitive position. That follows the bigger shift a few weeks earlier: the v1.5 update in August confirmed that legal services ads are now permitted in the US. For law firm owners and practice managers, this is the moment the newest major ad platform opened its doors to you while simultaneously reminding you how unfinished its guardrails are. Both halves of that story matter before any budget moves.
OpenAI's September Ad Policy Update: What v1.6 Changes
The September revision is short but consequential. OpenAI clarified that it can decline any ad that conflicts with its advertising principles, business interests, or competitive position. Read that carefully. It means an otherwise compliant campaign can be rejected for commercial reasons that have nothing to do with your creative or your category. That is a broader reservation of discretion than you see on mature ad platforms, and it should shape how much you rely on this channel.
The pace of change is the other signal. The legal services policy has changed twice in seven months and will change again. One legal marketing publication noted that its own analysis of the rules was current for less than a day because OpenAI revised the policy the same day the article published. Treat every summary you read, including this one, as a snapshot. Before committing budget, read the live policy at openai.com/policies/ad-policies yourself and confirm the current rules for legal services in your states.
Legal Services Ads Are Now Permitted in the US
The August 31 change is the reason this platform is suddenly on every legal marketer's radar. Effective August 31, 2026, OpenAI officially removed legal services from its disallowed list, and advertisements promoting legal advice, representation, or services are now allowed in the US as long as the advertiser holds a valid license to practice law. OpenAI specifically mentions personal injury, immigration, legal claims and document preparation in its policy.
The licensing rule turns on where ads are shown
The wording matters. The pivotal term in this policy update is "shown," rather than "based." Accountability rests on where your ad is delivered, not where your physical practice operates. A firm licensed in one state cannot let delivery spill into states where it holds no license. And the permission stops at the border: ads for legal services outside the US are still prohibited.
Approval is manual and gradual
Eligibility is not the same as access. The approval process is not necessarily automatic. OpenAI still considers legal services a restricted advertising category, and the company says it is rolling out these advertisers gradually, reviewing approvals on a case-by-case basis. OpenAI reviews the advertiser, ad copy, image, targeting, destination, and placement context, and approval is specific to the submitted setup, not a blanket endorsement of a brand or industry.
Early results from the channel are worth watching, with appropriate skepticism about sample size. One legal advertising agency reported that a client generated 12 leads from approximately $2,000 in ChatGPT advertising spend in a single month, roughly $166 per lead. That is one vendor's account of one client, not a benchmark. But it suggests the auction is not yet crowded, which is exactly when disciplined testing is cheapest.
Where Medical Providers Stand
Healthcare remains on the other side of the gate. Advertisers may promote only products and services permitted under OpenAI's Ads Content Policies, and advertisers whose primary business model or offerings fall within restricted or prohibited categories, such as healthcare, may be ineligible to advertise. There is movement underneath that surface: OpenAI may approve ads from approved advertisers in the financial services, healthcare and medicine, and legal services categories, with those categories being rolled out gradually and approvals reviewed manually. The legal category just graduated from that waiting room. Medical providers should expect their turn to come with similar conditions and should prepare compliance review workflows now rather than after the policy flips.
The Brand Safety Gap the Trade Press Just Exposed
The same week the platform's capabilities jumped, its advertising infrastructure came under scrutiny. On September 3, 2026, OpenAI launched GPT-6 Astra, which the company describes as "the world's most intelligent and aligned model." Days later, reporting from AdExchanger and Adweek revealed a far less flattering story about the same company's advertising business: brand-safety controls that remain thin, opaque, and reactive.
The specifics matter for regulated advertisers. ChatGPT currently offers baseline protections including exclusion of sensitive user contexts and stricter rules for regulated ad categories, with OpenAI testing new negative targeting options for a small group of advertisers. However, much of the placement adjacency decision still sits with OpenAI rather than advertiser-configurable blocklists, limiting granular control. Compare that to the exclusion lists, placement reports and third-party verification you take for granted on Google and Meta, and the gap is obvious.
The practical constraints are real but navigable. Ads still will not appear in sensitive conversations, still only reach Free and Go tier users, and firms must still comply with their state bar's attorney advertising rules regardless of what OpenAI now permits. The trajectory is also encouraging: OpenAI has added exclusion tooling almost monthly since the February launch, and policy updates have consistently moved toward clarifying rather than simply restricting eligibility for previously excluded verticals.
What Law Firms and Practices Should Do Now
This is a channel worth testing with strict guardrails, not a channel worth trusting with core budget yet. Work through this list:
- Read the live OpenAI ad policy yourself and record which version was current on the date you reviewed it. It has changed monthly.
- Confirm your licensing footprint matches your delivery footprint. The rule keys on where the ad is shown, so lock geographic targeting to states where attorneys in your firm are licensed.
- Route all creative through your compliance or ethics reviewer before submission. Platform approval does not satisfy state bar advertising rules.
- Start with a small, capped test budget and judge it on signed matters, not clicks. Tag destination URLs so ChatGPT traffic is visible in your analytics and intake system.
- Document a monitoring cadence. Review available placement and performance data weekly, and check for new OpenAI policy announcements at the same time.
- If you are a medical provider, do not run around the restriction with educational framing without a documented compliance opinion. Prepare your approval package for when the healthcare category opens further.
- Update any agency agreements to require the agency to cite the current policy version. Guidance written in June says the opposite of what the policy says today.
The Bottom Line
OpenAI opened ChatGPT advertising to US law firms in late August, then spent September reminding everyone who controls the platform. The v1.6 update reserves broad discretion to decline ads, and independent reporting confirms the brand safety stack is still years behind Google and Meta. The opportunity is genuine, early and cheap by the standards of legal advertising. The risk is a fast-moving policy environment and limited placement control. Firms that test small, verify the live policy before every campaign change, and keep compliance in the loop will be positioned to scale when the controls mature. Firms that wait for perfect tooling will meet more expensive auctions when they arrive.
