OpenAI hit a $1 billion annualized ad revenue run rate in roughly 200 days, according to Reuters reporting from August 31, 2026. That number gets attention. But the infrastructure updates shipping alongside the global expansion tell you more about where this channel is headed than the revenue milestone does.

What Actually Shipped

ChatGPT Ads started as a U.S.-only test in February 2026, expanded to Canada, Australia, and New Zealand in March, then the U.K., Japan, South Korea, Brazil, and Mexico through May and June. Late August brought 31 European markets, and self-service Ads Manager access is now rolling out across India, the Middle East, and North Africa. Total: over 40 countries.

Geography alone doesn't make a channel useful. The ad stack updates change the operational calculus.

Custom audiences now support add/remove/replace without full rebuilds, mixed identifier types in a single request, and audience sizes above 5 million members. Google Advertising ID (GAID) is a supported identifier. Exclusion audience restrictions loosened, and audience size estimates got more granular.

On measurement: the Measurement Pixel now accepts hashed phone numbers, names, regions, and postal codes for conversion matching. The Conversions API gained the same identifiers plus Android GAID support. Product-feed advertisers get carousel-card-level impressions and clicks in Ads Manager, with product-level fields available through the Insights API. OpenAI says those card impressions are separate from billable impressions.

A conversion optimization model on the roadmap factors in both click-through and view-through conversions, billed by impression while the system optimizes toward a selected conversion goal. That's the automated optimization pattern that underpins most performance advertising on Google and Meta today.

Where This Gets Complicated for B2B

Ads only appear on the Free and Go tiers. Not Plus. Not Pro. Not Enterprise. If your ICP uses ChatGPT through an enterprise license, they won't see your ads. That's a structural constraint worth sizing before you allocate budget.

The measurement story is better than six months ago, but still early. Hashed identifiers and a Conversions API improve match rates, yet the platform doesn't offer the campaign-control granularity or third-party verification that mature channels provide. You can match more conversions back to campaigns now. Whether you can prove incremental lift with confidence is a different question.

Expert guidance across multiple analyst and practitioner sources through 2026 frames ChatGPT advertising as a test-and-learn channel, not a replacement for Google or Meta. Measurement, targeting, and campaign controls remain limited relative to platforms with a decade of optimization infrastructure. That's a maturity assessment, not a dismissal.

The Experiment Worth Running

Hypothesis (falsifiable): If we run ChatGPT Ads against a high-intent use case (comparison/evaluation queries from ICP-adjacent Free-tier users), then qualified demo requests will increase by 10–15% above baseline from that segment, because conversational context aligns with decision-support intent rather than interruption.

Setup: One product line, one geo where self-serve is available, one landing page built for conversational intent. Creative should read like a helpful recommendation, not a display ad. Budget: allocate what you'd spend on a two-week paid social test, not a quarterly commitment.

Success metric: Qualified pipeline from attributed conversions. Guardrails: CPA stays within 1.5x your paid search benchmark for the same offer. Stop-loss: If CPA exceeds 2x after week two with no quality signal, pause and diagnose.

What to measure (and what not to over-interpret): Match-back conversion data will improve with the new pixel identifiers, but treat attribution as directional. Run a holdout if volume allows. Don't claim incrementality from platform dashboards alone.

One thing worth pairing with any paid test: organic answer-engine visibility. Multiple expert sources recommend combining paid pilots with content authority work (structured content, citations, reviews, backlinks) so your brand surfaces in AI-generated responses, not just the ad slot. Paid gets the impression; organic authority gets the unpaid mention. The combination is stronger than either alone.

The Trade-Off You're Accepting

ChatGPT Ads sit at a strange intersection: high-intent context (people asking questions during active research) with limited audience control (Free/Go tier only, no chat-history-based targeting per OpenAI's stated policy). You're buying access to a moment, not a precisely defined audience. For some B2B use cases, the moment is worth more than the audience definition. For others, the tier restriction kills channel fit before the test starts.

The Ads Manager plugin (which lets you create, manage, and analyze campaigns via natural-language instructions inside ChatGPT or Codex) is a workflow convenience, not a strategy. Confirm every recommended update before it applies. Automation without review is a failure mode, not a feature.

Six months ago, the story was "ChatGPT has ads now." Today it's infrastructure: identifiers, match rates, audience management, card-level reporting, conversion optimization models. That's the trajectory of a platform chasing performance budgets, not experimental line items. Whether measurement catches up fast enough to earn those budgets is the open question. The $1 billion run rate says some advertisers aren't waiting for the answer.