types.GoogleSearch doesn't return GroundingChunk with actual text snippet

Running generate_content with types.Tool(google_search=types.GoogleSearch()) returns candidates with format:

GroundingChunk(
  web=GroundingChunkWeb(
    domain='somedomain.com',
    title='somedomain.com',
    uri='https://vertexaisearch.cloud.google.com/grounding-api-redirect/...'
  )
)

The grounding chunk should be accompanied by the actual snippet/full text used during generation. Otherwise, results are not possible to validate other than by fetching the original link.

Environment details

Programming language: python
OS: macos
Language runtime version: 3.14.3
Package version: 2.33.0

Steps to reproduce

from google.genai import types, Client

client = Client(vertexai=True, project="...", location="...", credentials="...")
response = await client.aio.models.generate_content(
    model="gemini-3.8-flash",
    contents="why doesn't google genai with GoogleSearch grounding return GroundingChunk with text snippets?",
    config=types.GenerateContentConfig(tools=[types.Tool(google_search=types.GoogleSearch())]),
)
print(response.candidates[-1].grounding_metadata.grounding_chunks[0].web)
>>> GroundingChunk(
>>>   web=GroundingChunkWeb(
>>>     domain='somedomain.com',
>>>     title='somedomain.com',
>>>     uri='https://vertexaisearch.cloud.google.com/grounding-api-redirect/...'
>>>   )
>>> )

Thanks!

@guillermojp

Hi @guillermojp,

The GroundingChunk.web object returned by Google Search grounding provides only uri and title by backend API design; search index text snippets are not returned in the web grounding chunk payload.

To correlate sources with the generated output without fetching the URLs, inspect candidate.grounding_metadata.grounding_supports. This structure maps segments in the generated response to the corresponding chunk indices:

metadata = response.candidates[0].grounding_metadata
if metadata.grounding_supports:
    for support in metadata.grounding_supports:
        print(f"Segment: {support.segment.text}")
        sources = [metadata.grounding_chunks[i].web.uri for i in support.grounding_chunk_indices]
        print(f"Sources: {sources}")

If your workload strictly requires retrieved text snippets inside grounding_chunks, consider using Vertex AI Search data stores (types.VertexAISearch), which return retrieved_context chunks containing the snippet text.

For details on grounding structures and response fields, refer to the Vertex AI Grounding overview.

Because GroundingChunk fields are defined and populated directly by the Vertex AI backend service rather than the Python SDK, adding search snippet content to types.GoogleSearch chunks would require a service-level update. You can submit a feature request directly through Google Cloud Support or the Cloud Console feedback channel..

Thanks

This issue has been marked as stale because it has been open for 7 days with no activity. It will be closed in 2 days if no further activity occurs.

Navigation Menu

Repository navigation

`types.GoogleSearch` doesn't return `GroundingChunk` with actual text snippet #3235

Environment details

Steps to reproduce

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

types.GoogleSearch doesn't return GroundingChunk with actual text snippet #3235

Description

Environment details

Steps to reproduce

Activity

Venkaiahbabuneelam commented on Oct 29, 2026

github-actions commented on Nov 6, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

`types.GoogleSearch` doesn't return `GroundingChunk` with actual text snippet #3235