diff --git a/apps/sim/lib/knowledge/application/search-source-overview.test.ts b/apps/sim/lib/knowledge/application/search-source-overview.test.ts
index 7b4fbfff729..08b28960654 100644
--- a/apps/sim/lib/knowledge/application/search-source-overview.test.ts
+++ b/apps/sim/lib/knowledge/application/search-source-overview.test.ts
@@ -24,6 +24,7 @@ vi.mock('@/lib/knowledge/read-access', () => ({
 }))
 
 import { readSearchSourceOverview } from '@/lib/knowledge/application/search-source-overview'
+import { MAX_SEARCH_SOURCE_PROVIDER_TYPES } from '@/lib/knowledge/constants'
 
 const principal = { kind: 'session', userId: 'reader', sessionId: 'session' } as const
 const input = { organizationId: 'org-1', workspaceId: null }
@@ -36,6 +37,14 @@ const searchableProbeCount = () =>
   dbChainMockFns.limit.mock.calls.filter(([rows]) => rows === 1).length -
   AUTHORIZATION_SINGLE_ROW_READS
 
+/** The configured-provider list is read once, before the batches, under the same bound. */
+const CONFIGURED_PROVIDER_READS = 1
+
+/** Every other provider-bounded read in this use case is the indexing probe. */
+const indexingProbeCount = () =>
+  dbChainMockFns.limit.mock.calls.filter(([rows]) => rows === MAX_SEARCH_SOURCE_PROVIDER_TYPES)
+    .length - CONFIGURED_PROVIDER_READS
+
 function yieldBatches(count: number) {
   mocks.batches.mockImplementation(async function* () {
     for (let index = 0; index < count; index += 1) yield sql`batch-${sql.raw(String(index))}`
@@ -73,4 +82,31 @@ describe('readSearchSourceOverview', () => {
     expect(result.hasSearchableDocuments).toBe(false)
     expect(searchableProbeCount()).toBe(3)
   })
+
+  it('stops probing for indexing once every configured provider type is known', async () => {
+    yieldBatches(3)
+    queueTableRows(member, [{ role: 'owner' }])
+    queueTableRows(knowledgeConnector, [{ connectorType: 'gmail' }])
+    queueTableRows(knowledgeConnector, [{ connectorType: 'gmail' }])
+
+    const result = await readSearchSourceOverview.execute({ principal, input })
+
+    expect(result.providers).toEqual([{ connectorType: 'gmail', isSyncing: true }])
+    expect(indexingProbeCount()).toBe(1)
+  })
+
+  it('keeps probing every batch while a configured provider type is still unaccounted for', async () => {
+    yieldBatches(3)
+    queueTableRows(member, [{ role: 'owner' }])
+    queueTableRows(knowledgeConnector, [{ connectorType: 'gmail' }, { connectorType: 'notion' }])
+    queueTableRows(knowledgeConnector, [{ connectorType: 'gmail' }])
+
+    const result = await readSearchSourceOverview.execute({ principal, input })
+
+    expect(result.providers).toEqual([
+      { connectorType: 'gmail', isSyncing: true },
+      { connectorType: 'notion', isSyncing: false },
+    ])
+    expect(indexingProbeCount()).toBe(3)
+  })
 })
diff --git a/apps/sim/lib/knowledge/application/search-source-overview.ts b/apps/sim/lib/knowledge/application/search-source-overview.ts
index 0cfb6795313..c0892fcb856 100644
--- a/apps/sim/lib/knowledge/application/search-source-overview.ts
+++ b/apps/sim/lib/knowledge/application/search-source-overview.ts
@@ -117,16 +117,30 @@ export const readSearchSourceOverview = instrumentSourceOverviewUseCase(
         /** One searchable document is the whole answer, so later batches skip the probe entirely. */
         const probesSearchable: boolean = probesSources && !hasSearchableDocuments
         if (probesSearchable) searchableProbes += 1
+        /**
+         * A provider type is only read back as membership of `indexingTypes`, so once every
+         * configured type is in the set no later batch can change the answer.
+         */
+        const probesIndexing: boolean =
+          probesSources && providers.some(({ connectorType }) => !indexingTypes.has(connectorType))
         /** Annotated so the searchable probe's guard does not infer through its own result. */
         const [indexing, searchable]: [{ connectorType: string }[], { id: string }[]] =
           await Promise.all([
-            probesSources
+            probesIndexing
               ? measureSearchStage('source_overview.indexing', () =>
                   configuredProvidersQuery()
                     .where(
                       and(
                         configured,
                         syncingEnabled,
+                        /**
+                         * The probe narrows the configured set the provider list came from, so a
+                         * type already found stays found; excluding it only drops repeated work.
+                         * An empty set adds no predicate rather than a no-op one.
+                         */
+                        indexingTypes.size > 0
+                          ? notInArray(knowledgeConnector.connectorType, [...indexingTypes])
+                          : undefined,
                         or(
                           inArray(knowledgeConnector.status, ['pending', 'syncing']),
                           and(