Ads Fund Free Access Without Compromising Core Value

OpenAI introduces ads in ChatGPT's Free and Go tiers for logged-in adult US users to cover infrastructure costs for hundreds of millions of users, enabling faster, reliable free access and ongoing AI improvements. Paid tiers (Plus, Pro, Business, Enterprise, Education) remain ad-free. Users can upgrade to eliminate ads or opt out of Free tier ads by accepting fewer daily messages. Ads appear post-response, clearly labeled as "sponsored," and visually separated—e.g., recipe queries trigger meal kit ads like Heirloom Groceries' enchilada kit. Selection uses conversation topics, past chats, and ad interactions for relevance, without advertiser access to specifics.

This sustains mission-aligned growth: free/low-cost tiers stay viable, funding intelligence upgrades while offering paid escapes from ads.

Strict Independence, Privacy, and Safety Guardrails

ChatGPT answers prioritize helpfulness, unaffected by ads—optimization ignores advertiser input. Privacy blocks advertiser access to chats, history, memories, or details; they get only aggregate views/clicks. No ads for under-18 accounts (detected or predicted), or near sensitive topics like health, mental health, politics. Future scaling adds scam prevention, advertiser vetting, and anti-narrow-targeting guardrails.

Ad matching factors: current discussion topics, hidden/engaged ads, past chats/memory—but shares nothing identifiable. Builds trust via separation and transparency, with early pilots showing no trust metric drops, low dismissals, and improving relevance from feedback.

User Controls Drive Experience Ownership

Users dismiss ads, give feedback, view "why this ad" explanations, delete ad data one-tap, and toggle personalization or past chats/memory usage. Ads History logs views by timestamp/brand for clearing. Mobile/web settings centralize: History, Interests, data deletion, personalization toggles.

Long-term, conversational context makes ads more relevant during decision-making (e.g., option exploration), aiding discovery from small businesses too. Pilots refine before broader rollout.

Phased Global Expansion for Learning

US test (Feb 2026) yields positive signals, expanding March 2026 to Canada/Australia/New Zealand, then May 2026 to UK/Mexico/Brazil/Japan/South Korea. Thoughtful per-market rollout gathers real usage data for adjustments. Businesses sign up at openai.com/advertisers for updates; future evolves formats, objectives, buying models. Core invariants: unbiased answers, private conversations, user control.