ChatGPT Ads Reach 89 Countries: How the Ad System Works

OpenAI expanded ChatGPT Ads to 26 additional countries on October 21, taking its announced availability to 89 markets. It also published a technical explanation of how ads are selected. Together, the updates show a product moving beyond its initial rollout, with practical consequences for how advertisers prepare their content.

Where ChatGPT Ads are expanding

The expansion announcement includes 14 additional Latin American countries, with Argentina, Chile, Colombia and Peru among the examples. Other named additions include Nigeria, South Africa, Serbia, Ukraine, Cambodia, Bangladesh and Pakistan. Advertisers can use technology partners or self-service Ads Manager in the new markets.

OpenAI says more than 100,000 advertisers have run campaigns. Ads appear on Free and Go, while Plus, Pro and Enterprise remain ad-free. That advertiser figure describes participation, not campaign profitability.

For a business entering one of these markets, the first question should be which customer decision its product can help with. Geographic availability only removes one barrier. It does not establish that the product, audience and buying journey are a good fit.

How ChatGPT chooses an ad

OpenAI's engineering account describes a custom, post-trained model that interprets conversation context and calls advertising tools. It maintains state as the conversation develops, prepares retrieval queries, retrieves candidates, checks eligibility and evaluates relevance.

The orchestrator nominates candidates for an auction tool. It does not directly set the price or make the final auction decision. Further checks can still prevent delivery, including whether an ad arrives in time for the response.

OpenAI says this system runs separately from ChatGPT's answer generation, using different models and code. If ad selection is slow, the answer does not wait. Returning no ad is a normal outcome.

This makes the clarity of the underlying offer important. A company has to explain enough for its product to make sense in a developing conversation, rather than relying only on a memorable headline.

Imagine an AI documentation tool. Its landing page might clearly explain whether it serves public API references, internal engineering documentation or customer-support articles. Those distinctions help a human buyer decide whether it fits. They also provide more specific material for a system trying to interpret an advertiser's offering.

That is a practical content recommendation, not evidence that adding certain words guarantees more placements.

Creative enhancements need an account-level check

OpenAI's creative-enhancements documentation says the system can adapt headlines and descriptions, translate text, resize images and optimize destination URLs using existing ads and landing-page content. Enabled variations can run without individual approval or a preview of every version.

There is a timing discrepancy to check. The official help page currently says automatic enablement begins October 29, 2026, while reporting on advertiser emails cites October 30. Advertisers should inspect their own notice and Ads Manager settings rather than assume the later date applies.

The help page says existing opt-outs are preserved and campaign-specific settings override the account default. This is a settings decision, not something to leave to a headline about the rollout.

For a technical product, I would review the source material before enabling automated variations. A page that mixes a planned feature with an available feature creates ambiguity. So does a price that excludes a mandatory component without saying so clearly. Fixing those problems benefits buyers even if the advertising system never rewrites the copy.

Measure useful demand, not just clicks

The important test is whether a campaign produces the next meaningful customer action. For a developer tool, that might be a completed integration or an activated trial, rather than a visit to the pricing page. The appropriate measure depends on the product and its sales cycle.

A sensible experiment would keep the offer stable, record the settings used and compare results over a defined period. If several changes happen together, it becomes difficult to attribute a result to creative enhancements, a new market or a different landing page.

Broader availability makes testing easier; it does not make the outcome predictable. The useful advantage will come from understanding which conversations lead to a real product decision and having content that explains that decision honestly.

I’m Paras. I work at the intersection of AI, engineering and distribution.

I help AI founders turn product details and launch updates into clear technical stories that people can understand and share.

Book a 30-minute call