diff --git a/datastore/infrastructure/engines/bigquery/backend.py b/datastore/infrastructure/engines/bigquery/backend.py
index 8f31d84..130bb1e 100644
--- a/datastore/infrastructure/engines/bigquery/backend.py
+++ b/datastore/infrastructure/engines/bigquery/backend.py
@@ -42,6 +42,7 @@
     column_defs,
     delete_sql,
     drop_columns_sql,
+    format_select_column,
     insert_sql,
     merge_sql,
     normalize_pk,
@@ -448,8 +449,7 @@ def upsert(
         schema = self._read_schema(resource_id)
         if schema is None:
             raise NotFoundError(
-                f"resource {resource_id!r} is not declared; call "
-                "datastore_create before upsert"
+                f"resource {resource_id!r} not found."
             )
 
         rows = records or []
@@ -521,8 +521,7 @@ def search(
         schema = self._read_schema(resource_id)
         if schema is None:
             raise NotFoundError(
-                f"resource {resource_id!r} is not declared; call "
-                "datastore_create first"
+                f"resource {resource_id!r} not found."
             )
 
         try:
@@ -867,8 +866,7 @@ def info(self, resource_id: str) -> InfoResult:
         schema = self._read_schema(resource_id)
         if schema is None:
             raise NotFoundError(
-                f"resource {resource_id!r} is not declared; call "
-                "datastore_create first"
+                f"resource {resource_id!r} not found."
             )
 
         total = self._count_rows(resource_id)
@@ -1284,30 +1282,15 @@ def _build_export_select(schema: Any, fmt: str) -> str:
     """SELECT column list for EXPORT DATA.
 
     Parquet preserves native logical types → `*` is enough. For CSV /
-    NDJSON, cast TIMESTAMP and DATETIME columns to ISO 8601 (BigQuery's
-    default text format uses a space separator and `UTC` suffix, which
-    most clients reject as non-ISO). DATE and TIME already serialise as
-    ISO and pass through.
+    NDJSON, every column goes through `format_select_column` (in
+    `bigquery/lib.py`) — the same helper `datastore_search` uses — so a
+    given column renders identically in a dump and in a search response.
     """
     if fmt == "parquet":
         return "*"
-    parts: list[str] = []
-    for field in schema:
-        ftype = (field.field_type or "").upper()
-        if ftype == "TIMESTAMP":
-            # `%E*S` keeps all fractional seconds; trailing Z marks UTC.
-            parts.append(
-                f"FORMAT_TIMESTAMP('%Y-%m-%dT%H:%M:%E*SZ', "
-                f"`{field.name}`, 'UTC') AS `{field.name}`"
-            )
-        elif ftype == "DATETIME":
-            parts.append(
-                f"FORMAT_DATETIME('%Y-%m-%dT%H:%M:%E*S', `{field.name}`) "
-                f"AS `{field.name}`"
-            )
-        else:
-            parts.append(f"`{field.name}`")
-    return ", ".join(parts)
+    return ", ".join(
+        format_select_column(f.name, f.field_type) for f in schema
+    )
 
 
 def _is_export_too_large(exc: BaseException) -> bool:
diff --git a/datastore/infrastructure/engines/bigquery/lib.py b/datastore/infrastructure/engines/bigquery/lib.py
index 395fa2e..99d78b9 100644
--- a/datastore/infrastructure/engines/bigquery/lib.py
+++ b/datastore/infrastructure/engines/bigquery/lib.py
@@ -66,6 +66,30 @@ def _system_col_insert_list(include_updated_at: bool) -> str:
     return "`_id`, `_updated_at`" if include_updated_at else "`_id`"
 
 
+def format_select_column(name: str, bq_type: str | None) -> str:
+    """Render a SELECT-list expression for one column, casting TIMESTAMP /
+    DATETIME values to a fixed-shape ISO 8601 string
+    (`YYYY-MM-DDTHH:MM:SS` — UTC implicit, no offset, no fractional).
+    Other types pass through unchanged.
+
+    Shared by `datastore_search` (engine-built projection) and the
+    `datastore_dump` `EXPORT DATA` SELECT, so both read paths emit the
+    same timestamp shape from BigQuery's side. TIMESTAMP is rendered in
+    UTC; the resulting string carries no offset, so consumers must treat
+    any timestamp value as UTC.
+    """
+    bq = (bq_type or "").upper()
+    if bq == "TIMESTAMP":
+        return (
+            f"FORMAT_TIMESTAMP('%Y-%m-%dT%H:%M:%S', `{name}`, 'UTC') AS `{name}`"
+        )
+    if bq == "DATETIME":
+        return (
+            f"FORMAT_DATETIME('%Y-%m-%dT%H:%M:%S', `{name}`) AS `{name}`"
+        )
+    return f"`{name}`"
+
+
 def normalize_pk(schema: dict) -> list[str]:
     """`schema.primaryKey` as a list (str → 1-elem, missing → empty)."""
     pk = schema.get("primaryKey")
diff --git a/datastore/infrastructure/engines/bigquery/search.py b/datastore/infrastructure/engines/bigquery/search.py
index ee1b21b..2458f81 100644
--- a/datastore/infrastructure/engines/bigquery/search.py
+++ b/datastore/infrastructure/engines/bigquery/search.py
@@ -14,7 +14,9 @@
 from datastore.infrastructure.engines.bigquery.lib import (
     JSON_FRICTIONLESS_TYPES,
     SYSTEM_COLUMN_NAMES,
+    format_select_column,
 )
+from datastore.infrastructure.engines.bigquery.types import bigquery_type
 
 if TYPE_CHECKING:
     from google.cloud import bigquery
@@ -203,6 +205,19 @@ def _build_where(
     return " AND ".join(clauses)
 
 
+def _project_column(col: str, type_map: dict[str, str]) -> str:
+    """Render a projected column for `datastore_search` / `build_count`.
+
+    Translates the Frictionless type to BigQuery's name (via
+    `bigquery_type`) and delegates to `lib.format_select_column` — the
+    same helper `datastore_dump` uses for `EXPORT DATA`. Net effect:
+    TIMESTAMP / DATETIME columns come back as the fixed ISO 8601 string
+    `2026-06-08T00:00:00` (UTC implicit) in both endpoints; other
+    columns pass through as the native type.
+    """
+    return format_select_column(col, bigquery_type(type_map.get(col)))
+
+
 def build_search(
     *,
     table_ref: str,
@@ -265,7 +280,7 @@ def build_search(
     )
 
     parts: list[str] = []
-    projection = ", ".join(f"`{c}`" for c in projected)
+    projection = ", ".join(_project_column(c, type_map) for c in projected)
     parts.append(
         f"SELECT {'DISTINCT ' if distinct else ''}{projection} "
         f"FROM {table_ref} AS t"
@@ -273,8 +288,14 @@ def build_search(
     if where:
         parts.append(f"WHERE {where}")
     if sort_pairs:
+        # Sort against the underlying table column (`t.<col>`) rather than
+        # the projected alias — the datetime projection rewrites
+        # `col` → `FORMAT_TIMESTAMP(...) AS col`, and we want ORDER BY to
+        # operate on the native TIMESTAMP, not the formatted string. (The
+        # ISO format happens to sort lexicographically the same way, but
+        # the explicit reference future-proofs us if the format changes.)
         parts.append(
-            "ORDER BY " + ", ".join(f"`{c}` {d}" for c, d in sort_pairs)
+            "ORDER BY " + ", ".join(f"t.`{c}` {d}" for c, d in sort_pairs)
         )
     parts.append(f"LIMIT {int(limit)} OFFSET {int(offset)}")
 
@@ -322,7 +343,11 @@ def build_count(
         table_alias="t", params=params,
     )
 
-    projection = ", ".join(f"`{c}`" for c in projected)
+    # Same projection rewrite as `build_search` — datetime columns are
+    # FORMAT_TIMESTAMP-wrapped. Matters when `distinct=True`: the COUNT
+    # then dedupes on the formatted (second-precision) string, matching
+    # what the user sees in the data response.
+    projection = ", ".join(_project_column(c, type_map) for c in projected)
     inner_parts = [
         f"SELECT {'DISTINCT ' if distinct else ''}{projection} "
         f"FROM {table_ref} AS t"
diff --git a/tests/conftest.py b/tests/conftest.py
index 54d89ce..6170304 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -18,6 +18,14 @@
 # it a dummy when the env doesn't carry one — tests override the CKAN
 # client via DI, so this URL is never contacted.
 os.environ.setdefault("CKAN_URL", "http://test-ckan.local")
+# Force CKAN auth for the suite: the endpoint tests are built around
+# `CKANAuthProvider` + `FakeCKAN`, and the dict-resource create branch +
+# the read-only force guard are explicitly gated on `AUTH_TYPE == "ckan"`.
+# A developer .env that flips this to `anonymous` would otherwise reroute
+# those code paths and silently break ~7 endpoint tests locally (CI is
+# fine because it has no .env). `os.environ[...] =` forces an override,
+# unlike `setdefault` above which respects a CI-supplied value.
+os.environ["AUTH_TYPE"] = "ckan"
 
 from collections.abc import Iterator  # noqa: E402
 from typing import Any  # noqa: E402
diff --git a/tests/engines/bigquery/test_tables.py b/tests/engines/bigquery/test_tables.py
index f9b06a0..fb517db 100644
--- a/tests/engines/bigquery/test_tables.py
+++ b/tests/engines/bigquery/test_tables.py
@@ -571,7 +571,7 @@ def test_upsert_undeclared_resource_raises_not_found(
     backend = _backend(mock_client)
     backend._read_schema = MagicMock(return_value=None)
 
-    with pytest.raises(NotFoundError, match="not declared"):
+    with pytest.raises(NotFoundError, match="not found"):
         backend.upsert(
             "ghost", [{"a": 1}], method="upsert", include_total=False
         )
@@ -963,7 +963,7 @@ def test_info_raises_not_found_for_undeclared_resource(
     backend = _backend(mock_client)
     backend._read_schema = MagicMock(return_value=None)
 
-    with pytest.raises(NotFoundError, match="not declared"):
+    with pytest.raises(NotFoundError, match="not found"):
         backend.info("ghost")
     # No COUNT runs when the resource isn't declared.
     mock_client.query.assert_not_called()
@@ -1050,7 +1050,10 @@ def test_build_search_renders_full_param_set() -> None:
     assert sql.startswith("SELECT `auction_id`, `product_code` FROM `p.d.r` AS t")
     assert "WHERE `product_code` = @f0 AND `accepted` = @f1" in sql
     assert "SEARCH(t, @f2)" in sql
-    assert "ORDER BY `auction_id` DESC" in sql
+    # ORDER BY references the underlying table column (`t.<col>`) so the
+    # FORMAT_TIMESTAMP projection alias for datetime columns doesn't shadow
+    # the native value at sort time.
+    assert "ORDER BY t.`auction_id` DESC" in sql
     assert sql.rstrip().endswith("LIMIT 100 OFFSET 25")
     # Parameter types track the schema (STRING / BOOL / STRING for q).
     by_name = {p.name: p for p in params}
@@ -1106,7 +1109,7 @@ def test_build_search_default_sort_is_id_asc() -> None:
         limit=10,
         offset=0,
     )
-    assert "ORDER BY `_id` ASC" in sql
+    assert "ORDER BY t.`_id` ASC" in sql
 
 
 def test_build_search_rejects_unknown_columns() -> None:
@@ -1257,8 +1260,13 @@ def test_search_unfiltered_uses_cheap_row_count(
     sqls = [call.args[0] for call in mock_client.query.call_args_list]
     # `_updated_at` rides along in default projection because the
     # MagicMock config returns truthy for `INCLUDE_UPDATED_AT`.
+    # `_updated_at` is a `datetime` column → projection now wraps it in
+    # FORMAT_TIMESTAMP (UTC, no fractional, no offset) so the search
+    # response matches the dump endpoint's timestamp shape.
     assert sqls[0].startswith(
-        "SELECT `_id`, `a`, `_updated_at` FROM `proj-1.ds-1.res-1` AS t"
+        "SELECT `_id`, `a`, "
+        "FORMAT_TIMESTAMP('%Y-%m-%dT%H:%M:%S', `_updated_at`, 'UTC') "
+        "AS `_updated_at` FROM `proj-1.ds-1.res-1` AS t"
     )
     assert sqls[1] == (
         "SELECT COUNT(*) AS n FROM `proj-1.ds-1.res-1`"
@@ -1274,7 +1282,7 @@ def test_search_raises_not_found_for_undeclared_resource(
     backend = _backend(mock_client)
     backend._read_schema = MagicMock(return_value=None)
 
-    with pytest.raises(NotFoundError, match="not declared"):
+    with pytest.raises(NotFoundError, match="not found"):
         backend.search(
             resource_id="ghost",
             filters=None, q=None, distinct=False, plain=True,
diff --git a/tests/test_datastore_dump.py b/tests/test_datastore_dump.py
index 19085c9..2180cd4 100644
--- a/tests/test_datastore_dump.py
+++ b/tests/test_datastore_dump.py
@@ -157,6 +157,10 @@ def test_dump_with_denied_key_returns_403(
 
 
 def test_build_export_select_iso_casts_timestamp_and_datetime() -> None:
+    """TIMESTAMP / DATETIME columns render as `YYYY-MM-DDTHH:MM:SS` —
+    no timezone suffix, no fractional seconds. TIMESTAMP is formatted
+    in UTC (clients should assume UTC even though the string carries
+    no offset)."""
     schema = [
         _bq_field("auction_id", "INT64"),
         _bq_field("delivery_start", "TIMESTAMP"),
@@ -165,13 +169,16 @@ def test_build_export_select_iso_casts_timestamp_and_datetime() -> None:
     ]
     select = _build_export_select(schema, fmt="csv")
     assert (
-        "FORMAT_TIMESTAMP('%Y-%m-%dT%H:%M:%E*SZ', `delivery_start`, 'UTC')"
+        "FORMAT_TIMESTAMP('%Y-%m-%dT%H:%M:%S', `delivery_start`, 'UTC')"
         in select
     )
     assert (
-        "FORMAT_DATETIME('%Y-%m-%dT%H:%M:%E*S', `delivery_local`)"
+        "FORMAT_DATETIME('%Y-%m-%dT%H:%M:%S', `delivery_local`)"
         in select
     )
+    # No `Z` suffix and no `%E*S` (which would re-introduce fractional seconds).
+    assert "Z'," not in select
+    assert "%E*S" not in select
     assert "`auction_id`" in select
     assert "`delivery_day`" in select