OpenAI Announces ‘Code Red’ To Improve ChatGPT Amid Google Competition

OpenAI announces a ‘Code Red’ initiative to improve ChatGPT amid competition with Google focusing on speed, quality, and new AI capabilities.

OpenAI CEO Sam Altman has reportedly declared a “code red” to focus the company squarely on improving ChatGPT, pausing several monetization and product initiatives as Google’s Gemini 3 family posts strong benchmark results. The directive, detailed in an internal memo covered by The Wall Street Journal and The Information, signals that day-to-day ChatGPT quality is the top internal priority.​

What’s Changing Inside OpenAI

According to the reports, Altman told staff that ChatGPT’s everyday experience needs to improve across core dimensions: personalization, speed and reliability, and the breadth of questions it can handle.

OpenAI uses a color-coded prioritization system, and ChatGPT work has now been escalated from “code orange” to “code red,” indicating the highest internal urgency.​

The memo also references a new reasoning-focused model expected to launch as soon as next week, though OpenAI has not publicly announced details or benchmarks for it yet.

To support the push, teams are being encouraged to temporarily shift onto ChatGPT workstreams, with daily coordination calls for those driving improvements.​

Ads, AI Agents And Other Projects Delayed

To free up resources, several projects are being delayed. Advertising integrations that OpenAI had been testing inside the ChatGPT app are now on hold, according to The Information, putting a pause on early ad experiments tied to assistant usage.​

AI agents aimed at shopping and healthcare workflows are also being pushed back, along with further upgrades to ChatGPT Pulse, the company’s AI-driven trend and monitoring feature.

In effect, OpenAI is trading short-term product expansion and revenue experiments for a concentrated effort on core assistant quality.​

Why Google’s Gemini 3 Is Driving This

The backdrop is Google’s rapid progress with Gemini 3 and related models, which have recorded strong scores on Google’s own Gemini 3 Pro benchmark page.

Google says its Gemini 3 Deep Think variant outperforms earlier Gemini versions on Humanity’s Last Exam, a newer, more difficult benchmark designed by AI safety researchers as a successor to saturated tests like MMLU.​

Those results, published on Google’s own Gemini 3 Pro benchmark page and mirrored on independent leaderboards such as Scale’s Humanity’s Last Exam board and third-party dashboards like Artificial Analysis, show Gemini 3 Pro scoring ahead of current GPT-5-class systems on that specific benchmark.

OpenAI has strong scores on other reasoning suites, but the memo appears to respond to this recent cluster of Gemini 3 results rather than any single test.​

Benchmarks, Context And Limitations

Humanity’s Last Exam is maintained by the Center for AI Safety and Scale AI, with performance tracked across multiple public leaderboards. It is explicitly designed to be harder and less gameable than earlier benchmarks, emphasizing robust reasoning over narrow pattern-matching.​

Still, any “outperforms” claim is benchmark-specific: Gemini 3 Pro leads on certain tests, while GPT-5-series models hold advantages elsewhere.

The challenge for OpenAI is perception as much as raw numbers—Gemini’s recent benchmark wave has created a sense of momentum that Altman’s memo is clearly trying to counter by accelerating visible improvements in ChatGPT’s behavior.​

ChatGPT Still Leads

Despite the technical pressure, OpenAI’s assistant remains far ahead in overall usage. In a recent LinkedIn post, ChatGPT head Nick Turley described the product as:

“#1 AI assistant worldwide,” accounting for “around 70% of assistant usage”

Full comments are available here.

External reporting from outlets such as the Financial Times has estimated more than 800 million weekly users for OpenAI, with the vast majority on the free tier, while Google’s Gemini continues to grow from a smaller base.​

Altman’s memo reportedly acknowledges Google’s recent progress and warns of “temporary economic headwinds,” but also asserts that OpenAI is “catching up fast.”

The combination of cost pressure, competition and user expectations helps explain why leadership is willing to delay new revenue lines in favor of a quality sprint.​

A “Code Red” That Mirrors Google’s Own Past

The framing is also notable because “code red” is the same language Google itself used after ChatGPT’s viral launch. At that time, CEO Sundar Pichai pulled Google’s founders and key teams into an accelerated AI effort, fast-tracking work that led from Bard’s launch in 2023 to the current Gemini lineup.​

Now, the roles are somewhat reversed: Google’s sustained infrastructure and model investments have delivered a competitive edge on several high-profile benchmarks, prompting OpenAI to adopt a similar crisis-style focus for its flagship product. It’s a reminder that the “AI lead” is dynamic, not fixed.​

How OpenAI Is Positioning ChatGPT

Publicly, OpenAI is leaning into competition as a positive force. Turley has used LinkedIn and X to highlight ChatGPT’s scale and to argue that new entrants push OpenAI to move faster, saying the company’s focus is on making ChatGPT “more capable,” “more intuitive” and “more personal,” while broadening access.​

Notably, OpenAI has not commented directly on the leaked memo, which is typical for internal planning documents. Instead, messaging has centered on user value and feature velocity rather than on any explicit response to Gemini 3.​

What To Watch Next

The upcoming reasoning model release will be the first visible test of how quickly OpenAI can translate this “code red” focus into user-facing improvements.

For marketers, SEOs and power users, the key signals will be changes in how ChatGPT handles complex, multi-step queries, research-heavy tasks and extended follow-up conversations, as well as any noticeable shifts in speed and personalization.​

In parallel, Google is likely to keep iterating on Gemini 3 and its broader AI stack, including image systems like Nano Banana and Nano Banana Pro, which bolster its multimodal story.

Final Thought

The net result is an accelerated product cycle on both sides good news for users, but a moving target for anyone building workflows, content strategies or search tactics around AI assistants.