Articos tests messaging, positioning and landing pages against simulated personas matched to a team's ideal customer profile. Users can question the personas about their reasoning and get a report back in under 30 minutes. The MCP release moves that workflow into Claude and ChatGPT, so a study can start from the chat window where copy is already being drafted. The 86% figure is the company's own claim of agreement with human findings from Baymard and Nielsen Norman. It is not an independent audit. The company says its users include PMMs, growth teams, agencies and founders at Instacart, HiJiffy and DealHub. It recently added live calls, which let users talk to the simulated audience in real time. The company also claims roughly 7x better results than general tools such as ChatGPT or Claude. A commenter on Product Hunt asked how that comparison was measured, and the thread does not show a clear answer. For AI founders, the product shows MCP becoming a distribution channel for vertical research tools, not just a way to connect data. Synthetic-persona testing is cheap and fast, but its value rests on validation claims buyers will need to check. Teams considering it should run a few past campaigns with known results before trusting it with live launches.