How many findings does an AI plan review produce?

Across 521 published case-study records, the median is 88 potential findings and the average is 174. Review scope and publishing selection affect these figures.

By InspectMind·Updated September 16, 2026·About 4 min read

Dataset refreshed 2026-09-16. Source report and downloads

The distribution matters more than one average

Published totals range from 1 to 5,199 findings. Larger sets and broader review scopes can produce many more findings, so the median and full distribution give useful context alongside the average.

Potential findings per public case-study record
Findings per record	Case-study records	Share
1–10	26	5.0%
11–25	45	8.6%
26–50	76	14.6%
51–100	138	26.5%
101–200	115	22.1%
201–500	94	18.0%
501–1,000	16	3.1%
1,001+	11	2.1%

Keep the denominators separate

The public library contains 90,768 project-level findings, while its selected structured excerpts contain 40,401 records. These are different populations. Multiple case-study records can describe related projects or review versions; neither total should be presented as a count of unique buildings or independently confirmed defects.

The production snapshot is separate: 1,756 completed checks across 1,688 customer project records as of 2026-10-21. Read the definitions and exclusions.

How to use the benchmark on your next review

Compare a similar document scope and design stage. A three-discipline permit set is not comparable to one structural detail.
Inspect individual findings and their supporting references. Prioritize confirmed discrepancies by consequence and actionability.
Ask the design team to reconcile conflicting values, specifications, and missing information before measuring corrections or outcomes.
Use actual customer outcomes to evaluate value. Do not multiply a raw finding count by an assumed dollar amount and describe the result as realized savings.

Common questions

How many findings does an AI plan check produce?

Across 521 published case-study records, the median is 88 potential findings and the average is 174. Review scope and publishing selection affect these figures.

Do more findings mean better accuracy?

No. Counts alone do not establish accuracy, design quality, or defects prevented. Findings need professional review against the complete documents.

Are these construction-industry defect rates?

No. This is a selected sample of InspectMind case studies, not a random sample of construction projects. Document scope, design stage, and review settings vary.

Explore project studies →Open the sample report →See pricing →