diff --git a/adapters/README.md b/adapters/README.md
index 096a6e4..10b9d07 100644
--- a/adapters/README.md
+++ b/adapters/README.md
@@ -5,6 +5,7 @@
 > **If you just want to use an existing adapter**, jump to the package READMEs:
 > - [`@payloadcms-vectorize/pg`](./pg/README.md) — PostgreSQL + pgvector
 > - [`@payloadcms-vectorize/cf`](./cf/README.md) — Cloudflare Vectorize
+> - [`@payloadcms-vectorize/mongodb`](./mongodb/README.md) — MongoDB Atlas + self-hosted 8.2+
 > - Or see the [main README](../README.md) for end-to-end setup.
 
 ---
@@ -380,7 +381,7 @@ The dev harness in [`dev/`](../dev) runs the integration suite against any adapt
 3. Run the suites that matter:
    - `pnpm test:int` — full integration suite (real Payload, real DB).
    - `pnpm test:e2e` — Playwright E2E.
-   - `pnpm vitest adapters/pg/dev/specs/vectorSearchWhere.spec.ts` — the `where` operator suite (31 tests). **Every adapter should pass this.**
+   - `pnpm vitest adapters/pg/dev/specs/vectorSearchWhere.spec.ts` — the `where` operator suite (31 tests). **Every adapter should pass this** (or an equivalent — CF's equivalent is `where.spec.ts`).
 
 If your store doesn't support an operator, document it in [Adapter feature parity](#adapter-feature-parity) and have your `where` translation throw a clear error rather than silently returning wrong results.
 
@@ -398,14 +399,14 @@ If your store doesn't support an operator, document it in [Adapter feature parit
 
 ## Adapter feature parity
 
-| Feature | PG | CF | Notes |
-|---|---|---|---|
-| Real-time ingest (`storeChunk`) | ✅ | ✅ | |
-| Bulk ingest (`hasEmbeddingVersion` + `storeChunk`) | ✅ | ✅ | |
-| `where` operators | full | full | Both pass `vectorSearchWhere.spec.ts`. |
-| Server-side `like` regex | ✅ | ✅ (with regex escape — see CHANGELOG 0.7.1) | |
-| Migration CLI bin | ✅ (`vectorize:migrate`) | ❌ | CF uses indexes managed via Cloudflare API. |
-| Score range | `[0, 1]` (cosine similarity) | varies by index metric | Document yours. |
+| Feature | PG | CF | MongoDB | Notes |
+|---|---|---|---|---|
+| Real-time ingest (`storeChunk`) | ✅ | ✅ | ✅ | |
+| Bulk ingest (`hasEmbeddingVersion` + `storeChunk`) | ✅ | ✅ | ✅ | |
+| `where` operators | full | full | full | PG & MongoDB pass `vectorSearchWhere.spec.ts` (31 tests each); CF covers `where` in `where.spec.ts` (59 tests). MongoDB has documented pre/post-filter behavior — see [Limitations](./mongodb/README.md#limitations). |
+| Server-side `like` regex | ✅ | ✅ (with regex escape — see CHANGELOG 0.7.1) | ✅ | |
+| Migration CLI bin | ✅ (`vectorize:migrate`) | ❌ | ❌ | CF uses indexes managed via Cloudflare API. MongoDB creates search indexes at runtime via `ensureSearchIndex` (no bin). |
+| Score range | `[0, 1]` (cosine similarity) | varies by index metric | `vectorSearchScore`; range depends on the index similarity metric | Document yours. |
 
 If something here is out of date, please [open an issue](https://github.com/techiejd/payloadcms-vectorize/issues) — adapter parity drift is exactly what this table exists to surface.
 
diff --git a/adapters/mongodb/dev/specs/extensionFields.spec.ts b/adapters/mongodb/dev/specs/extensionFields.spec.ts
index ac5d1f2..938fbe1 100644
--- a/adapters/mongodb/dev/specs/extensionFields.spec.ts
+++ b/adapters/mongodb/dev/specs/extensionFields.spec.ts
@@ -106,4 +106,24 @@ describe('Extension fields (mongodb)', () => {
     expect((hit as any).category).toBe('cat-a')
     expect((hit as any).priority).toBe(7)
   }, 90_000)
+
+  test('stored fields outside filterableFields are also returned by search (CF/PG parity)', async () => {
+    const target = Array(DIMS).fill(0.66)
+    await adapter.storeChunk(payload, 'default', {
+      sourceCollection: 'posts',
+      docId: 'doc-nonfilterable',
+      chunkIndex: 0,
+      chunkText: 'parity',
+      embeddingVersion: testEmbeddingVersion,
+      embedding: target,
+      extensionFields: { category: 'cat-a', priority: 9, note: 'not-a-filterable-field' },
+    })
+
+    await new Promise((r) => setTimeout(r, 1500))
+
+    const r = await adapter.search(payload, target, 'default', 5)
+    const hit = r.find((x) => x.docId === 'doc-nonfilterable')
+    expect(hit).toBeDefined()
+    expect((hit as any).note).toBe('not-a-filterable-field')
+  }, 90_000)
 })
diff --git a/adapters/mongodb/src/search.ts b/adapters/mongodb/src/search.ts
index 02c5392..508c944 100644
--- a/adapters/mongodb/src/search.ts
+++ b/adapters/mongodb/src/search.ts
@@ -5,6 +5,14 @@ import { convertWhereToMongo, evaluatePostFilter } from './convertWhere.js'
 import { ensureSearchIndex } from './indexes.js'
 import { RESERVED_FIELDS, type ResolvedPoolConfig } from './types.js'
 
+const RESERVED_AND_META = new Set<string>([
+  ...RESERVED_FIELDS,
+  '_id',
+  'score',
+  'createdAt',
+  'updatedAt',
+])
+
 export interface MongoSearchCtx {
   uri: string
   dbName: string
@@ -53,20 +61,10 @@ export async function searchImpl(
   if (pool.forceExact) vectorSearchStage.exact = true
   if (preFilter) vectorSearchStage.filter = preFilter
 
-  const projection: Record<string, unknown> = {
-    _id: 1,
-    score: { $meta: 'vectorSearchScore' },
-    sourceCollection: 1,
-    docId: 1,
-    chunkIndex: 1,
-    chunkText: 1,
-    embeddingVersion: 1,
-  }
-  for (const f of pool.filterableFields) projection[f] = 1
-
   const pipeline: Record<string, unknown>[] = [
     { $vectorSearch: vectorSearchStage },
-    { $project: projection },
+    { $addFields: { score: { $meta: 'vectorSearchScore' } } },
+    { $project: { embedding: 0 } },
   ]
 
   const collection = client.db(ctx.dbName).collection(pool.collectionName)
@@ -76,19 +74,19 @@ export async function searchImpl(
     ? rawDocs.filter((d) => evaluatePostFilter(d as Record<string, unknown>, postFilter!))
     : rawDocs
 
-  return filtered.map((d) => mapDocToResult(d as Record<string, unknown>, pool.filterableFields))
+  return filtered.map((d) => mapDocToResult(d as Record<string, unknown>))
 }
 
-function mapDocToResult(
-  doc: Record<string, unknown>,
-  filterable: string[],
-): VectorSearchResult {
+function mapDocToResult(doc: Record<string, unknown>): VectorSearchResult {
   if (typeof doc.score !== 'number') {
     throw new Error(
-      `[@payloadcms-vectorize/mongodb] Search result is missing numeric "score" field; ensure $project includes { score: { $meta: 'vectorSearchScore' } }`,
+      `[@payloadcms-vectorize/mongodb] Search result is missing numeric "score" field; ensure the pipeline adds { score: { $meta: 'vectorSearchScore' } }`,
     )
   }
-  const result: Record<string, unknown> = {
+  const extensionFields = Object.fromEntries(
+    Object.entries(doc).filter(([k]) => !RESERVED_AND_META.has(k)),
+  )
+  return {
     id: String(doc._id),
     score: doc.score,
     sourceCollection: String(doc.sourceCollection ?? ''),
@@ -97,11 +95,6 @@ function mapDocToResult(
       typeof doc.chunkIndex === 'number' ? doc.chunkIndex : Number(doc.chunkIndex ?? 0),
     chunkText: String(doc.chunkText ?? ''),
     embeddingVersion: String(doc.embeddingVersion ?? ''),
-  }
-  for (const f of filterable) {
-    if (f in doc && !(RESERVED_FIELDS as readonly string[]).includes(f)) {
-      result[f] = doc[f]
-    }
-  }
-  return result as VectorSearchResult
+    ...extensionFields,
+  } as VectorSearchResult
 }