Coframe Raises $25M Series A for Self-Optimizing Websites

Coframe has raised US$25 million in Series A funding to build websites that generate, test and deploy changes continuously, instead of waiting for the next manual optimization cycle. The round, co-led by Khosla Ventures and Hanabi Capital and announced on November 9, brings the San Francisco company's disclosed total funding to US$35 million. The valuation was not disclosed.

The pitch is straightforward: landing pages, ads and eventually other digital experiences should adapt to visitor behavior faster than a human growth team can manage. Whether that speed translates into a durable edge over existing testing tools remains unproven.

What does the money actually buy?

The company applies AI agents to a part of marketing still dominated by experimentation queues, design reviews and engineering tickets. Coframe says it uses a model trained on user interactions alongside agents assigned roles such as strategy, design, engineering, quality assurance and data analysis. These systems generate alternatives, run them as experiments and feed results back into the next round of changes.

That collapses several stages of a conventional conversion-rate-optimization workflow. Instead of preparing a handful of page versions against a testing calendar, a marketer could manage a larger set of ongoing variations. The core decision shifts from "which variant do we build next" to "which objectives, audience boundaries and safeguards do we set."

Backers include Khosla Ventures, Hanabi Capital, Nat Friedman and Antler. The round follows an earlier seed investment and extends a bet on automated conversion optimization as a category, not just a better testing dashboard.

Founder and CEO Josh Payne framed the ambition in one line: "We're building an internet that improves itself." A useful statement of direction — but a slogan is not evidence that automated decision-making always improves the customer experience.

What are the performance claims?

The headline numbers in the announcement concern customer results:

• More than US$300 million in incremental customer revenue
• Average conversion-rate improvement above 20%
• Named clients including Dropbox, Replit and Cash App

Both revenue and conversion figures are Coframe-reported estimates from the November 9 announcement. The attribution methodology and independent audit were not supplied.

The missing details matter as much as the figures. A conversion lift on one landing page may not extrapolate across an entire site. Results can shift with traffic mix, seasonality, campaign spend, baseline performance and experiment duration. Revenue attribution is especially sensitive to how the counterfactual — what would have happened without the tool — is constructed.

Business Insider's November 9 reporting confirms the investment and Coframe's broader pitch, but the commercial performance numbers remain company-provided.

Who is the competition?

Coframe is not entering empty ground. Adobe Target, Optimizely and VWO already offer experimentation and personalization, and newer specialists are racing to automate more of the creative and operational work. Coframe's differentiation claim is not better testing software — it is that more of the testing workflow runs with less manual intervention.

The company's November announcement also describes work with OpenAI around ChatGPT ads and landing-page optimization. That is Coframe's account of the relationship. It is not an independently verified claim of advertiser performance, and OpenAI did not invest in the company.

What should buyers check first?

The clearest near-term opportunity is operational. A large team can spend weeks coordinating routine web tests across copy, design, development, legal review and analytics. Automating parts of that process could increase the number of ideas tested without a matching increase in production work.

But optimization is not only about maximizing clicks. An AI system can drift toward misleading urgency, inaccessible layouts or aggressive personalization if those tactics lift a narrow metric. Teams need constraints for brand consistency, user consent, accessibility and legal claims before software touches live customer experiences.

Buyers should also separate variation generation from measurement quality. Questions worth asking:

• How are visitors randomized?
• Are segments compared fairly?
• Which changes require human approval?
• How is a version rolled back?
• Who owns the experiment data?

Those controls determine whether automation produces reliable learning or simply more activity.

The funding gives Coframe room to pursue a more automated website stack. For marketers, the relevant test is narrower: can it shorten experiment cycles while preserving evidence quality and customer trust? Until the claimed gains are reproducible under clear measurement conditions, the honest takeaway is a faster way to test — not a guaranteed faster way to grow.