Should I fill in the audience survey?

Complete all three fields, or skip entirely

There are two paths: fill in role, level, and task, or leave the survey blank. Both produce targeted, useful results. The path that can produce noisier results is filling in only a role, or a role and level without a task.

The survey appears after Phase 1, which is free. Phase 2, which uses your audience input to produce the writing brief and explanation framework, requires an analysis credit.

Why a partial survey can produce more recommendations, not fewer

This is counterintuitive, so the study data is worth stating directly. A generic role and level without a task averaged 8.0 combined toClarify and toMakeExplicit guidance items. No audience at all averaged 6.8. A complete audience description with role, level, and task averaged 4.0.

Fewer, more targeted items is the goal. Partial information appears to create an ambiguous audience model, which the pipeline resolves by broadening its guidance scope to serve more possible reader types. Only a complete audience description reliably narrows the guidance below the blank baseline.

Note: this comparison covers toClarify and toMakeExplicit only. The toAdd count, which reflects concepts missing entirely from the page, follows a different pattern and is not included in these figures.

What "complete" means

All three fields, with specificity. Role: name the actual job and context, not just the function. "B2B SaaS content strategist" produces a different calibration than "marketer." Level: the reader's expertise with this specific subject, not their general seniority. Task: the specific workflow this content needs to support.

"Identify LLM retrieval gaps in technical documentation" is a task. "Improve my content" is a goal, not a task. See the task description guide for the full pattern and examples.

You cannot enter something wrong

Nonsense input produces valid, coherent results. Contradictory input resolves sensibly. There is no input state that produces errors or degraded output. Enter what you know, as specifically as you can, and the analysis will use it. See the resilience guide for full detail on what was tested and how the pipeline responds.