ChatGPT's Free Tier Is Now an Advertising Platform — And You Were Opted In by Default
Agentic AIOn April 30, 2026, OpenAI quietly activated marketing cookies by default for all free ChatGPT users — sharing cookie IDs and email addresses with advertising partners on platforms including Instagram, with no user action required to opt in.
The change arrived without a user-facing announcement. It became public when privacy researchers and journalists examined the updated data controls settings and traced the outbound data flows. OpenAI spokesperson Taya Christianson confirmed the change, noting that chat content is not shared with advertisers and that users can disable tracking under Settings > Data Controls > Marketing Privacy.
Context
The rollout is the logical endpoint of a trajectory OpenAI has been on since February 2026, when it began testing in-app ads at the bottom of ChatGPT responses for Free and Go tier users in the United States. That pilot has since expanded to Canada, Australia, and New Zealand. Paid subscribers — Plus, Pro, Business, Enterprise, and Edu accounts — remain exempt from both ads and the new tracking default.
The financial math explains the urgency. Roughly 200 million users access ChatGPT weekly, and the overwhelming majority are on the free tier. Serving those users at scale is not cheap. OpenAI has disclosed targets of approximately $2.5 billion in advertising revenue for 2026, part of a longer-arc plan toward $100 billion annually as it approaches a potential IPO at the end of this year.
The Detail
What OpenAI is sharing is narrower than the phrase “ad tracking” implies. The data flowing to advertising partners is cookie identifiers and email addresses — sufficient for building retargeting audiences on Instagram and similar platforms, but not conversation history, uploaded files, or browsing behavior within ChatGPT. The stated purpose is promoting OpenAI’s own products, not selling targeting data to third-party advertisers.
That framing is technically accurate but practically significant in ways that merit attention. An email address linked to a ChatGPT account tells an advertising system that this user is an AI-tool adopter. That inference is itself valuable audience-segmentation data, regardless of what was said in any particular chat session. The boundary between “data about you” and “data derived from your use patterns” is a meaningful one in privacy law — and it is precisely the line consent defaults have repeatedly been used to redraw.
The opt-out path exists and is accessible, but opt-out defaults produce consistently lower opt-out rates than opt-in defaults. The vast majority of the 200 million free users will never change the setting.
The So What
For individual users: the opt-out is straightforward. Settings > Data Controls > Marketing Privacy, toggle off. It takes thirty seconds.
For organizations that have deployed ChatGPT Free as an informal productivity layer — without formal enterprise licensing — the implication is different. Employees using personal ChatGPT accounts for work purposes are now generating attribution data that flows out of the organization’s control. That is not a theoretical risk; it is a data governance gap that IT and legal teams should address explicitly.
The deeper signal here is structural. OpenAI’s commercial model is undergoing a fundamental shift: from a subscription-first business to a mixed-model platform where free users fund infrastructure through attention and data rather than money. That model is well understood in consumer tech. It is new territory for AI tools, and the governance frameworks most enterprises built around AI usage have not caught up to it yet.
Photo by Sanket Mishra on Pexels
Source: The Decoder
Content created with AI assistance and reviewed for accuracy.
Join the conversation
Stack Insiders is our free community for readers who want to go deeper — share resources, ask questions, and connect with others across every vertical we cover.
Join Stack Insiders →