diff --git a/app/models/archive_content_models.py b/app/models/archive_content_models.py
index 3c80183..4fd26bd 100644
--- a/app/models/archive_content_models.py
+++ b/app/models/archive_content_models.py
@@ -1,7 +1,7 @@
 import datetime, pytz
 
 from typing import Dict, List, Optional, Set, Union, ClassVar
-from pydantic import BaseModel, EmailStr, Field, Json, PrivateAttr, ValidationError, validator
+from pydantic import BaseModel, EmailStr, Field, Json, PrivateAttr, ValidationError, validator, root_validator
 
 from app.db_sql import redis_lookup_id_random
 from app.lib_general import log, logging
@@ -18,13 +18,12 @@ class Archive_Content_Base(BaseModel):
     #     log.debug(test_var)
     #     return test_var
 
-    id_random: Optional[str] = Field(
-        # **base_fields['archive_content_id_random'],
-        alias = 'archive_content_id_random',
-    )
-    id: Optional[int] = Field(
-        alias = 'archive_content_id'
-    )
+    # --- Vision IDs (primary public identifiers — always random strings) ---
+    id: Optional[str] = Field(None, **base_fields['archive_content_id_random'])
+    archive_content_id: Optional[str] = Field(None, **base_fields['archive_content_id_random'])
+    # Legacy alias kept for backward compatibility; populated by root_validator
+    id_random: Optional[str] = Field(None, alias='archive_content_id_random')
+
     account_id_random: Optional[str]
     account_id: Optional[int]
 
@@ -94,6 +93,7 @@ class Archive_Content_Base(BaseModel):
 
     # Fields that are part of the model (for reading) but should not be saved to the DB table
     fields_to_exclude_from_db: ClassVar[list] = [
+        'id', 'archive_content_id', 'id_random',
         'account_id', 'account_id_random', 'archive_id_random', 'hosted_file_id_random',
         'hosted_file_path', 'api_hosted_file_path_download', 'api_hosted_file_path_stream',
         'hosted_file_hash_sha256', 'hosted_file_content_type', 'hosted_file_size'
@@ -101,12 +101,21 @@ class Archive_Content_Base(BaseModel):
 
     _processed_at: datetime.datetime = PrivateAttr(default_factory=datetime.datetime.now)
 
-    @validator('id', always=True)
-    def archive_content_id_lookup(cls, v, values, **kwargs):
-        if isinstance(v, int) and v > 0: return v
-        elif id_random := values.get('id_random'):
-            return redis_lookup_id_random(record_id_random=id_random, table_name='archive_content')
-        return None
+    @root_validator(pre=True)
+    def map_v3_ids(cls, values):
+        """
+        Vision Transformer: Map DB random-string keys to clean Vision ID fields.
+        Collision prevention strips any integer that snuck into the string ID fields.
+        """
+        rid = values.get('id_random') or values.get('archive_content_id_random')
+        if rid and isinstance(rid, str):
+            values['id'] = rid
+            values['archive_content_id'] = rid
+        # Collision prevention: reject integer values in Vision string fields
+        for k in ['id', 'archive_content_id']:
+            if k in values and not isinstance(values[k], str):
+                del values[k]
+        return values
 
     @validator('archive_id', always=True)
     def archive_id_lookup(cls, v, values, **kwargs):
diff --git a/app/routers/api_crud_v3.py b/app/routers/api_crud_v3.py
index 9d6e181..c5f67bd 100644
--- a/app/routers/api_crud_v3.py
+++ b/app/routers/api_crud_v3.py
@@ -464,8 +464,13 @@ async def post_obj(
         if return_obj:
             if sql_select_result := sql_select(table_name=table_name_select, record_id=new_obj_id):
                 resp_data = output_model(**sql_select_result).dict(by_alias=serialization.by_alias, exclude_unset=serialization.exclude_unset)
+                # V3 contract: {obj_type}_id in responses must be the random string, never the integer.
+                _id_key = f'{obj_name}_id' if serialization.by_alias else 'id'
+                _rand_key = f'{obj_name}_id_random' if serialization.by_alias else 'id_random'
+                if isinstance(resp_data.get(_id_key), int) and resp_data.get(_rand_key):
+                    resp_data[_id_key] = resp_data[_rand_key]
                 return mk_resp(data=resp_data, response=response)
-        return mk_resp(data={"obj_id": new_obj_id, "obj_id_random": new_obj_id_random}, response=response)
+        return mk_resp(data={"obj_id": new_obj_id_random, "obj_id_random": new_obj_id_random}, response=response)
     else:
         # Standardized rich error bubbling
         db_err = format_db_error(get_last_sql_error())
diff --git a/app/routers/api_crud_v3_nested.py b/app/routers/api_crud_v3_nested.py
index ed94015..d3218b8 100644
--- a/app/routers/api_crud_v3_nested.py
+++ b/app/routers/api_crud_v3_nested.py
@@ -306,8 +306,13 @@ async def post_child_obj(
         if return_obj:
             if sql_select_result := sql_select(table_name=table_name_select, record_id=new_obj_id):
                 resp_data = output_model(**sql_select_result).dict(by_alias=serialization.by_alias, exclude_unset=serialization.exclude_unset)
+                # V3 contract: {obj_type}_id in responses must be the random string, never the integer.
+                _id_key = f'{child_obj_type}_id' if serialization.by_alias else 'id'
+                _rand_key = f'{child_obj_type}_id_random' if serialization.by_alias else 'id_random'
+                if isinstance(resp_data.get(_id_key), int) and resp_data.get(_rand_key):
+                    resp_data[_id_key] = resp_data[_rand_key]
                 return mk_resp(data=resp_data, response=response)
-        return mk_resp(data={"obj_id": new_obj_id, "obj_id_random": new_obj_id_random}, response=response)
+        return mk_resp(data={"obj_id": new_obj_id_random, "obj_id_random": new_obj_id_random}, response=response)
     else:
         # Standardized rich error bubbling
         db_err = format_db_error(get_last_sql_error())
diff --git a/documentation/GUIDE__AE_API_V3_for_Frontend.md b/documentation/GUIDE__AE_API_V3_for_Frontend.md
index d0fcbbc..8bcec8c 100644
--- a/documentation/GUIDE__AE_API_V3_for_Frontend.md
+++ b/documentation/GUIDE__AE_API_V3_for_Frontend.md
@@ -68,6 +68,30 @@ Modify data in the system.
     *   **Header:** `x-ae-ignore-extra-fields: true`
     *   **Behavior:** When set to `true`, the backend will automatically strip any fields from the payload that are not defined in the object's model before attempting to save to the database.
 
+### D. ID Fields in Responses (Vision ID Convention)
+
+> [!IMPORTANT]
+> **V3 responses always use random string IDs — never database integers.**
+
+After a successful `POST` create or any `GET`, the response contains:
+
+| Field | Type | Use |
+| :--- | :--- | :--- |
+| `{obj_type}_id` | `string` | **Primary public ID.** Use this for subsequent `PATCH` calls and UI routing. |
+| `{obj_type}_id_random` | `string` | Legacy alias. Same value as `{obj_type}_id`. Present for backward compat only. |
+
+**Example — create then immediately PATCH:**
+```ts
+const created = await postArchiveContent(archiveId, payload);
+const newId = created.data.archive_content_id; // random string e.g. "xK9mP3qRtL2"
+
+// Use it directly in the PATCH URL — no lookup needed
+await patchArchiveContent(newId, { name: 'Updated Name' });
+// PATCH /v3/crud/archive/{archive_id}/archive_content/{newId}
+```
+
+> **Note on `_id_random` suffix:** The `{obj_type}_id_random` field is a legacy artifact from the pre-Vision model. Once you confirm `{obj_type}_id` is a random string (length 11–22), you do not need `_id_random` as a fallback. New code should only read `{obj_type}_id`.
+
 ---
 
 ## 4. V3 Uniform Lookup System