diff --git a/docs/2-handbook/swarm/ui.md b/docs/2-handbook/swarm/ui.md
index b78b03ebd..e8cf5b505 100644
--- a/docs/2-handbook/swarm/ui.md
+++ b/docs/2-handbook/swarm/ui.md
@@ -299,7 +299,7 @@ The **classifier settings** section shows the runtime, model and
 effort the [review classifier](../../4-implementation/swarm/review-classifier.md) runs with and how many
 requested reviews run at once, and lets you change them. The **PR reviews list
 only failed rules** checkbox makes the review the classifier posts on a PR leave
-out the rules that passed or didn't apply; try-outs always show every rule. The
+out the rules that passed, didn't apply or were below their threshold; try-outs always show every rule. The
 runtime is **API
 key** (claude on the Anthropic key) or **ACP** (opencode on a provider). On the
 API key runtime the model is free text, one word as claude's `--model` takes
diff --git a/docs/4-implementation/swarm/review-classifier.md b/docs/4-implementation/swarm/review-classifier.md
index 182718cf0..95f6470f7 100644
--- a/docs/4-implementation/swarm/review-classifier.md
+++ b/docs/4-implementation/swarm/review-classifier.md
@@ -241,7 +241,8 @@ replaces it.
   answer for as long as the biggest run the limit allows
 - `failures_only`: a boolean, `false` by default (also for a document stored
   before the field existed). When `true`, a posted PR review lists only the
-  rows that failed their rule, the ones that make it ask for changes, and an
+  rows that failed their rule, the ones that make it ask for changes, so the ❔
+  rows (below their threshold, or unsure) are left out too, and an
   approval says `No check failed.` in place of the table. A review with no
   verdict, and a try-out, always show every row
 - `runtime`: `api-key` or `acp` (see [Runtimes](#runtimes))