OpenAI Allows Some Health & Finance Ads In ChatGPT via @sejournal, @MattGSouthern

A new SE Ranking report found ChatGPT showed ads on 28.69% of the commercial prompts in its healthcare category, more than 10 times the 2.64% the company recorded in Google’s AI Mode. OpenAI’s placement rules no longer keep ads away from medical advice the way its launch announcement describes.

In April, OpenAI made its ad placement rules more flexible. Its changelog describes this update as a “more precise approach,” and notes that medical, legal, and financial advice contexts are no longer blocked by default. Sensitive conversations are still off limits.

What OpenAI’s Ad Policies Say Now

OpenAI’s ad policies page, last updated on August 10, sets separate rules for where ads can be placed and who is permitted to buy them.

Placement rules keep ads out of chats OpenAI treats as unsuitable, a list that includes political content. They also keep ads out of what it calls sensitive user contexts, which cover conversations about mental and personal health.

The second rule is about who can buy, which involves approving advertisers in finance, health care, and legal services, one at a time through manual review.

On the health side, it lists health insurance, dental services, supplements, vision products, medical testing, and hospitals and urgent care, among others. On legal services, it says the opposite. Ads for legal advice, court-related work, and similar services are not allowed, except for general legal education. Political ads are completely banned.

In July, the update to the same page clarified this by reworking the health and finance sections, showing exactly which categories and markets qualify.

OpenAI’s February announcement has not been changed to match. The company has updated that post three times, most recently on August 11 to include the UK, Mexico, Brazil, Japan, and South Korea in the pilot program. The original statement still notes that ads “are not eligible to appear near sensitive or regulated topics.”

What The Report Measured

SE Ranking analyzed 50,006 commercial prompts across 20 niches in the United States, collected on July 23. Out of these, 12,974 returned an ad, a rate of 25.94%.

The report organizes prompts by niche and doesn’t include the prompt list itself. This means the healthcare figure counts commercial queries that the company categorized under healthcare, rather than personal health conversations, which are governed by OpenAI’s placement rules. Additionally, the data cannot indicate whether ads appeared alongside conversations that these rules are meant to protect.

The same gap applies to the News & Politics rate of 28.76%. SE Ranking says many prompts in that niche asked about newspaper and newsletter subscriptions, and it names a direct mail company as the niche’s leading advertiser. Those are commercial media buys, not political ads.

The report also found that 14.35% of ad impressions sat no closer to their own prompt than to a random one, measured by comparing each ad with a shuffled set of prompts. This rate varied from just 2.6% in Pets to a higher 54.2% in News & Politics.

SE Ranking sells ad tracking for ChatGPT and AI Mode within its Competitive Research product, and the report concludes by guiding readers towards it.

Why The AI Mode Comparison Is Looser Than It Looks

The 2.64% AI Mode figure comes from a separate SE Ranking report published July 14, which SEJ covered last month. That analysis used 50,032 commercial keywords picked to trigger a text ad in Google, collected on June 30.

The keyword list was created for Google’s ad system, so it compares ChatGPT to a set designed specifically for another platform. Additionally, these two collections are 23 days apart, as platforms frequently update their ad rules each month.

The healthcare comparison is a bit limited. AI Mode’s healthcare niche had 91 ads from 36 advertisers, making it the smallest among the 20 niches in that report.

Why This Matters

The current policy opens specific health and finance categories to approval one at a time, and its list names real products, which makes them worth checking a campaign against.

Legal services run the other way. The summary line indicates approval might be possible, but a detailed rule says they cannot be advertised. The launch announcement doesn’t mention these details. Instead, it only points out a general restriction on where ads can appear, which makes it the wrong page to plan against.

The same analysis suggests the report’s headline numbers shouldn’t be taken as an indication of a safeguard failure. Just because ads are landing on healthcare-category commercial prompts doesn’t necessarily mean that OpenAI’s placement rules were broken.

Looking Ahead

OpenAI has updated these rules four times in five months, with the latest update on August 10 to include housing and job listings. These updates often change who or what qualifies, so a category that might be closed when a media plan is written could be open by the time it runs.

Featured Image: Nwz/Shutterstock