Recipient Blindness

Every AI can write. None of them know who they're writing to. That gap is Recipient Blindness — and it's why AI communication tools send the same message to everyone.

What it is

Recipient Blindness is the condition of an AI communication system that produces fluent text without any persistent model of the person receiving it. The system optimizes the message for an "average" reader who doesn't exist, so a frustrated parent, a skeptical VP, and an anxious student all get the same words. The output is grammatically perfect and contextually wrong.

Why it's structural, not a model problem

Recipient Blindness is not a capability gap in the underlying model. It is a design gap in the system. No amount of prompt engineering closes it, because the system has no persistent recipient state to draw on — each message starts from zero, with nothing learned about who the recipient is or how they prefer to be communicated with.

Where it shows up

Recipient Blindness isn't a niche failure. It's the default behavior of nearly every tool that writes on a person's behalf:

The cost

The cost compounds quietly. Messages that don't fit the recipient raise the adaptation-failure rate — the share of sends that miss the person they were written for. Over a thread, that erodes trust: each slightly-off message signals "this wasn't really for me." And it shows up in the numbers that matter most — reply rates drift down, and the relationship cools without an obvious culprit.

The fix

Recipient Intelligence closes the gap. It gives the system a durable model of the recipient and adapts each message to them — the same way a thoughtful human reads the room before they speak.

Read what that layer is on Recipient Intelligence, or start building with the developer docs.