diff --git a/context/code/formula-and-io.md b/context/code/formula-and-io.md
index adf752eb..1d3bd443 100644
--- a/context/code/formula-and-io.md
+++ b/context/code/formula-and-io.md
@@ -638,7 +638,15 @@ ones beside it (`Date_Year`, `Date_Quarter`, `Date_Month`):
   ungrouped item, with a warning naming the column and the row count, rather
   than being dropped. The warning is separate from the split shape's, because
   that one reports an absent coordinate and this one a coordinate that is there
-  and ungrouped.
+  and ungrouped. **Unless its spelling is one the hierarchy generates**
+  (improvise-d0n6): a raw `2024-01` or `2024` is the same name as the generated
+  month or year, so it would join the grouped item and the warning would be
+  false. The build **refuses**, naming the column and the spellings and pointing
+  at the split shape. That follows the collision rule below: where two records
+  would silently be one thing, the user is asked and nothing is defaulted. A
+  fallback to the split shape was rejected as a silent change of the user's
+  choice, and a renamed raw item as altering the data. The check is
+  `Plan::spelling_clash`.
 - **Item order is first-seen, not chronological.** Order is a correctness
   property here (gotcha 18), and a group need not be contiguous, so a CSV that
   is not in date order still nests correctly. It is just drawn in the order
diff --git a/context/test-evidence/formula-and-io.md b/context/test-evidence/formula-and-io.md
index 189a7990..cbd1f4f0 100644
--- a/context/test-evidence/formula-and-io.md
+++ b/context/test-evidence/formula-and-io.md
@@ -1904,3 +1904,18 @@ improvise-8jk6.12 (4). A `region` of `"   "` is kept as a record with no
 predates this PR (improvise-ych), but the sparse-record rule now rests on it:
 without it the value becomes an item named `"   "`. Rival F4 (the reviewer's
 fault) fails it.
+
+### a_raw_date_spelled_like_a_generated_item_is_refused
+
+improvise-d0n6 (Codex P2 on PR #322). A `%Y-%m-%d` column imported Year > Month
+with `2024-01-15` and a raw `2024-01`. Before the refusal the build succeeded,
+put both records on the grouped item `2024-01` (summing them) and warned that
+the raw value was "kept ungrouped", which was false. Fault injected: removing
+the `spelling_clash` call makes this test fail with `unwrap_err()` on an `Ok`.
+
+### a_raw_date_spelled_like_a_generated_group_is_refused
+
+improvise-d0n6. The same with a raw `2024`, a generated group. Removing the
+`spelling_clash` call fails it the same way. The control,
+`a_raw_date_spelled_like_nothing_generated_is_kept_ungrouped`, passes either
+way, so a blanket refusal of every unparsed value fails it.
diff --git a/crates/improvise-io/src/import/wizard.rs b/crates/improvise-io/src/import/wizard.rs
index 622dc3db..a66ea166 100644
--- a/crates/improvise-io/src/import/wizard.rs
+++ b/crates/improvise-io/src/import/wizard.rs
@@ -1,4 +1,4 @@
-use std::collections::HashMap;
+use std::collections::{HashMap, HashSet};
 
 use anyhow::{Result, anyhow};
 use serde_json::Value;
@@ -252,6 +252,9 @@ impl ImportPipeline {
         }
 
         let placements: Vec<Placement> = self.records.iter().map(|r| plan.place(r)).collect();
+        if let Some(clash) = plan.spelling_clash(&placements) {
+            return Err(anyhow!(clash));
+        }
         let collisions = collisions_among(&placements);
         let occurrence = match (&self.on_duplicate, collisions.is_empty()) {
             (_, true) | (DuplicatePolicy::Sum, false) => None,
@@ -617,6 +620,60 @@ impl Plan<'_> {
     }
 }
 
+impl Plan<'_> {
+    /// Why the build stops when a value the date format could not read is spelled
+    /// like an item or group the hierarchy generates (improvise-d0n6). Both would
+    /// be one name in the category, so the raw record would silently join the
+    /// grouped item, and the "ungrouped" warning would be false.
+    fn spelling_clash(&self, placements: &[Placement]) -> Option<String> {
+        let generated = |field: &str| -> HashSet<&str> {
+            placements
+                .iter()
+                .filter_map(|p| match p {
+                    Placement::Placed { chains, .. } => Some(chains),
+                    Placement::NotObject => None,
+                })
+                .flatten()
+                .filter(|(cat, _)| cat == field)
+                .flat_map(|(_, chain)| chain.iter().map(String::as_str))
+                .collect()
+        };
+        for (at, h) in self.date_hierarchies.iter().enumerate() {
+            let tally = self.date_extractions.len() + at;
+            let names = generated(h.field);
+            let mut clashing: Vec<&str> = vec![];
+            for p in placements {
+                let Placement::Placed {
+                    coords, unparsed, ..
+                } = p
+                else {
+                    continue;
+                };
+                let raw = coords.iter().find(|(c, _)| c == h.field).map(|(_, v)| v);
+                match raw {
+                    Some(v) if unparsed.contains(&tally) && names.contains(v.as_str()) => {
+                        if !clashing.contains(&v.as_str()) {
+                            clashing.push(v);
+                        }
+                    }
+                    _ => {}
+                }
+            }
+            if !clashing.is_empty() {
+                return Some(format!(
+                    "'{}' has value(s) that do not match the date format '{}' but are spelled \
+                     like a month, quarter or year it generates: {}; fix them, or import the \
+                     column with the split shape",
+                    h.field,
+                    h.format,
+                    clashing.join(", ")
+                ));
+            }
+        }
+        None
+    }
+}
+
 /// One hierarchical date coordinate: its category and its chain, root first.
 type DateChain = (String, Vec<String>);