Critiquing AI Startup Websites

Tan says the surge in AI capability gives startups a timely reason to exist, but can also mean they are explaining a new category. Reviewing Rosebud, he finds its mix of game assets, game development, and several calls to action unclear. He recommends choosing the primary product and guiding visitors to its key action and aha moment quickly, with a demo that shows what the AI can do.

For MagicFlow, Tan says people look for software to solve a specific problem, not generic workflow tools, so common use cases can help visitors understand where to start. He likes the early “How it works” section and familiar model names, but finds that a promised drag-and-drop interaction leads to login without demonstrating the product. He suggests letting visitors try an initial workflow before asking them to create an account.

Tan says Reality Defender’s site takes too long to explain what it does and who it serves, despite social proof such as government use. He notes that spacing and line-height issues can weaken the sense of enterprise quality. He recommends writing for a specific ideal customer, using language learned from existing users, and showing a concrete AI capability.

Reviewing Pump, Tan finds the animated video visually busy and the group-buying model, which appears to help companies save on AWS, buried in the site. He recommends making that distinction and the savings mechanism explicit, adding a recognizable customer example, and using motion selectively to draw attention to the main action.

Tan praises Voiceflow’s clear product demo, calm motion, visible customer logos, and customer quotes with real people. He says examples help visitors understand what a product is for and whether it fits them. He also points to established products as useful references for design techniques refined through user feedback.