ChatGPT Ads Hit $1B. Your Free Tier Isn't Free Anymore.

OpenAI's ad business crossed $1 billion in annualized revenue in 200 days. If anyone on your team runs real work through ChatGPT's free tier, that number just became your problem, not just a line in OpenAI's IPO deck.

Here's why this matters past the headline. Ads inside a product change what the product optimizes for. Google search got worse for a decade under exactly this pressure, and everyone in marketing knows it happened even if nobody agreed on when. OpenAI just told you, in a press release, that it's running the same play — and it's running it faster than Google ever did. Six months from a standing start to $1 billion is not a pilot anymore. It's the business model.

The number, and how fast it moved

OpenAI announced the $1 billion annualized run-rate milestone on August 31, roughly 200 days after ChatGPT Ads launched [1]. Ads are now live in more than 40 countries, and OpenAI opened self-serve access to marketers in India, Europe, and the Middle East and North Africa the same week [1]. OpenAI has told investors it expects $2.5 billion in ad revenue this year, climbing to $100 billion by 2030 [4]. Keep the scale straight, though: the company's overall annualized run rate topped $40 billion in August — roughly double where it sat at the end of 2025 [5]. Ads are the fastest-moving line, not the biggest one.

The part that should actually change how you use the product: ads run against the free tier and the lower-priced Go plan, while Plus, Pro, Business, Enterprise, and Education stay ad-free [6] — and the free tier is the vast majority of ChatGPT's more than 1 billion weekly active users [1]. If your team logs in without a paid seat, you're not a beta tester anymore. You're inventory.

"Low-tech" targeting isn't a compliment

Back in March, when OpenAI's ad business was still sitting around $100 million, eMarketer surveyed agency buyers actually running campaigns on it. The verdict wasn't flattering. Click-through rates were measured at 0.91%, against a 6.4% benchmark for Google search ads — seven times worse [2]. Advertisers called the ad-serving process "relatively low-tech." DEPT's SVP of media went further, saying the process "lacks transparency" [2].

Read that from the other side of the product. "Low-tech and non-transparent targeting" doesn't mean your ads underperform. It means the system deciding what shows up in a free-tier user's chat, based on what that user typed, is a system nobody outside OpenAI can fully explain yet — including, per the same reporting, some of the marketers paying to use it [2].

Story: OpenAI — A milestone in expanding access to AI. Image via Digiday.

That's not a reason to panic. It's a reason to stop treating the free tier like a neutral tool. A system that's opaque about how it targets ads is, by definition, a system you can't audit for what it does with what you type into it — and "I don't know what happens to this" is exactly the sentence that should keep client data, pricing strategy, and unreleased product plans off it.

The paid side already answers a question the free tier won't

OpenAI has a real answer for the "what happens to my data" question — it's just gated behind a paid relationship. In August, OpenAI rolled out zero data retention for eligible API and enterprise customers: prompts and responses aren't stored after a request completes, OpenAI staff can't review them, and none of it trains future models unless you opt in [3]. It's not automatic — enterprise customers request it through sales, and it applies to specific endpoints and plans, not to ChatGPT's ad-supported consumer tier [3].

That's the actual line. On one side: a product funded by ads, opaque targeting, and 1 billion mostly-free users. On the other: a product funded by your invoice, with a written data-handling commitment you can point to in a client contract. Same company, same models underneath, completely different deal.

Where I draw the line for clients

I run this table with every client who tells me "we already use ChatGPT, why would we pay for anything else." The free tier isn't a lesser version of the same tool. It's a different business relationship wearing the same interface.

Here's the actual decision I walk through before I sign off on a workflow touching any AI product, not just OpenAI's:

flowchart TD
  Start([Task involves an AI tool]) --> Q1{Does client data,<br/>pricing, or unreleased<br/>work touch the prompt?}
  Q1 -->|No — personal/throwaway use| Free[Free tier is fine]
  Q1 -->|Yes| Q2{Is the account<br/>a paid seat or API key?}
  Q2 -->|No, still free tier| Upgrade[Upgrade first —<br/>don't run this through ad-funded inventory]
  Q2 -->|Yes, paid| Q3{Does the vendor publish a<br/>data-retention commitment<br/>for this specific plan?}
  Q3 -->|No| Ask[Get it in writing<br/>before you scale usage]
  Q3 -->|Yes, in writing| Ship[Approved for the workflow]

If you don't pay for the product, its incentives don't work for you. OpenAI just published the receipts.

None of this means ChatGPT is a bad tool — I still recommend it constantly. It means the free tier and the paid tier stopped being the same product the day ads started paying the bills, and treating them as interchangeable is exactly the kind of gap I look for when I audit a client's tool stack before letting it touch anything that matters.

Sources

[1] OpenAI — A milestone in expanding access to AI — openai.com [2] eMarketer — OpenAI's ChatGPT ads hit $100 million in annualized revenues, but doubts remain — emarketer.com [3] OpenAI — Offering Zero Data Retention for frontier models — openai.com [4] Reuters — OpenAI projects $2.5 billion in ad revenue this year, $100 billion by 2030, Axios reports — reuters.com [5] CNBC — OpenAI CFO Friar tells investors that enterprise is bigger than consumer — cnbc.com [6] OpenAI — Testing ads in ChatGPT — openai.com

So what do you do with this

Ask your team one question this week: who's still doing real work — client research, drafts with a client's name in them, anything with a number attached — on a free ChatGPT login. Whatever the answer is, it's higher than you think, because nobody flags "I used the free version" as a decision worth mentioning. That's exactly the kind of quiet gap I close when I audit a client's stack before recommending what to consolidate, pay for, or drop — not because free is always wrong, but because "free" and "no cost" stopped being the same sentence in February. If you want a second set of eyes on where your team's AI use actually sits on that table, book a 15-minute call.

The short version

• OpenAI's ad business hit $1 billion in annualized revenue 200 days after launch, expanding self-serve access to India, Europe, and the Middle East/North Africa the same week.
• Ads run against ChatGPT Go subscribers and the free tier — and free-tier users make up the vast majority of ChatGPT's ~1 billion weekly users.
• Agency buyers surveyed by eMarketer measured a 0.91% click-through rate (versus Google's 6.4% benchmark) and called OpenAI's targeting "low-tech" and lacking transparency.
• OpenAI's zero-data-retention commitment exists — but it's for paid API and enterprise plans on request, not the ad-supported consumer tier.
• My rule: if client data, pricing, or unreleased work touches the prompt, it doesn't touch the free tier. Pay for the seat or use the API with a written retention commitment.