Coframe’s $25M Series A Puts AI-Driven Website Experiments at the Center of the ‘Living In

Coframe, the AI-powered website experimentation platform, has raised $25 million in a Series A round co-led by Khosla Ventures and Hanabi Capital, bringing its total funding to $35 million. The round includes participation from leaders at OpenAI, Anthropic, and Google DeepMind, as well as Nat Friedman and Antler, according to Coframe’s announcement on November 9, 2026. Founded in 2023 by Josh Payne, the company is betting on a future where websites don’t just serve content—they *learn* from every interaction and adapt in real time.

This isn’t just another testing solution. Coframe’s vision, which it calls the *“living internet,”* frames digital experiences as dynamic systems that evolve based on user behavior, conversion data, and AI-driven experimentation. Unlike traditional web design—where teams spend months on redesigns—Coframe’s platform automates the creation, testing, and optimization of hundreds of website and ad variations in minutes. Customers like Dropbox, BCG, Block, Backcountry, and Sur La Table report $300 million in incremental revenue from Coframe’s experiments, with conversion lifts as high as 158% over six months.

The company’s proprietary user-behavior model, trained on hundreds of millions of monthly sessions, powers this approach. AI agents handle strategy, design, engineering, quality assurance, and data analysis, ranking experiments by predicted conversion and revenue impact before deploying them. A Dropbox case study highlights the speed: an experiment that once took four weeks of manual work was built and deployed in 30 minutes.

Yet the bigger ambition goes beyond websites. Coframe is collaborating with OpenAI on ChatGPT ads, aiming to merge ad optimization with landing-page experiments into a single, continuously learning system. Earlier work fine-tuned GPT-4o for generating on-brand code and creative assets, improving website consistency by 26% compared to base models. Now, the focus is on a user-behavior model that learns from interactions—not just from static data. The company is also preparing for AI agent-readable pages, designing interfaces that agents can parse, understand, and cite, though these capabilities remain in development.

The challenge? Balancing speed with oversight. Coframe’s system requires customer approval before deploying variations, a safeguard against uncontrolled experiments. But as AI agents increasingly act on behalf of users—researching, comparing, and purchasing—Coframe’s roadmap suggests it’s building guardrails into its own product. Whether these controls will scale alongside the “living internet” remains uncertain.

For investors, Coframe’s backing from Khosla Ventures (known for bets on AI infrastructure) and Hanabi Capital (a leader in enterprise SaaS) reflects confidence in the shift from static to adaptive digital experiences. The involvement of OpenAI, Anthropic, and DeepMind leadership suggests a strategic alignment: if AI agents are the future of the web, Coframe is positioning itself as the infrastructure layer that enables them.

The next step will be seeing how Coframe extends its conversion gains to ads and agent-facing pages—while keeping human oversight in place. If it succeeds, the “living internet” could redefine not just web design, but the entire economy of digital experimentation.

Sources: unite.ai