How we test our answers

No one can promise the city will approve your plans. What we can do is check our work before you see it, and count what happens after you resubmit. This page explains exactly what we count.

Checked by automated review before it reaches you

Every response in your report is checked by automated review before the report is sent. A response is marked “covered” only when it:

Anything that needs a design decision, a calculation or information we can’t see becomes a question for you instead, and is not covered until you answer it.

Tested on comments cities really rejected

We took 140 real comments from Seattle’s public plan review records where the city rejected the applicant’s first answer and explained why. We drafted our own answer to the original comment, then asked an AI grader whether our answer already did what the reviewer asked for the second time. It did 44% of the time (October 28, 2026). Then, as we do for customers when the city sends a comment back, we revised each answer that fell short using the reviewer’s follow-up. With that revision, 92% were handled. That second round is easier than the first, because the revision sees the reviewer’s own words, so read it as how often our free revision fixes what the city asked for, not as a city approval. Comments that depended on drawings no one could see were left out rather than counted as passes.

Our record with real customers

You pay for a permit package (every desk’s letter for one review cycle) only after the city clears the comments we covered, so we have to know what the city did. After you resubmit, we look at the city’s next review of your project. Seattle publishes every plan review comment, so there we check the public record ourselves; anywhere else, your next letter tells us. For each comment we marked as covered, we count whether the city sent it back.

We’ll post our record with customers here once 20 covered comments have had their next city review.

What these numbers are not