diff --git a/docs/assets/capabilities/cat.png b/docs/assets/capabilities/cat.png
new file mode 100644
index 0000000..e14663a
Binary files /dev/null and b/docs/assets/capabilities/cat.png differ
diff --git a/docs/assets/capabilities/kitchen.jpg b/docs/assets/capabilities/kitchen.jpg
new file mode 100644
index 0000000..34ed085
Binary files /dev/null and b/docs/assets/capabilities/kitchen.jpg differ
diff --git a/docs/assets/capabilities/roses-are.wav b/docs/assets/capabilities/roses-are.wav
new file mode 100644
index 0000000..6f3eadc
Binary files /dev/null and b/docs/assets/capabilities/roses-are.wav differ
diff --git a/docs/assets/capabilities/surprise.png b/docs/assets/capabilities/surprise.png
new file mode 100644
index 0000000..9c42ccd
Binary files /dev/null and b/docs/assets/capabilities/surprise.png differ
diff --git a/docs/deployment/ai-edge-gallery.md b/docs/deployment/ai-edge-gallery.md
index 593af51..3f65e30 100644
--- a/docs/deployment/ai-edge-gallery.md
+++ b/docs/deployment/ai-edge-gallery.md
@@ -1,5 +1,5 @@
 ---
-title: AI Edge Gallery
+title: Try Gemma on your phone with AI Edge Gallery
 description: Run Gemma 4 on an Android or iOS phone with no code — chat, vision, audio transcription and benchmarking, fully offline.
 models: [gemma-4]
 frameworks: [litert, android, ios]
diff --git a/docs/deployment/gemini-api.md b/docs/deployment/gemini-api.md
index 2207a82..4816779 100644
--- a/docs/deployment/gemini-api.md
+++ b/docs/deployment/gemini-api.md
@@ -1,5 +1,5 @@
 ---
-title: Gemini API
+title: Run Gemma with the Gemini API
 description: Call hosted Gemma 4 models through the Gemini API instead of running and serving your own instance.
 models: [gemma-4]
 frameworks: [gemini-api]
diff --git a/docs/deployment/google-cloud.md b/docs/deployment/google-cloud.md
index d80ef17..9747bf3 100644
--- a/docs/deployment/google-cloud.md
+++ b/docs/deployment/google-cloud.md
@@ -1,5 +1,5 @@
 ---
-title: Google Cloud
+title: Gemma on Google Cloud
 description: "Deploy, serve, and fine-tune Gemma 4 on Google Cloud: Cloud Run, GKE, the Agent Development Kit, vLLM on TPUs, MaxText, and Sovereign Cloud."
 models: [gemma-4]
 frameworks: [google-cloud, vllm]
diff --git a/docs/deployment/huggingface.md b/docs/deployment/huggingface.md
index 7d9fd4f..d39588f 100644
--- a/docs/deployment/huggingface.md
+++ b/docs/deployment/huggingface.md
@@ -1,6 +1,6 @@
 ---
 title: Run Gemma with Hugging Face Transformers
-description: Generating text, summarizing, and analysing content are just some of the tasks you can accomplish with Gemma open models.
+description: "Run Gemma 4 with Hugging Face Transformers in Python for text, image, audio, and multi-turn chat generation."
 models: [gemma-4, gemma-3, gemma-3n]
 frameworks: [transformers]
 source_url: "https://ai.google.dev/gemma/docs/core/huggingface_inference.md.txt"
@@ -163,7 +163,7 @@ messages = [
         "role": "user",
         "content": [
             {"type": "text", "text": "Transcribe the following speech segment in its original language. Follow these specific instructions for formatting the answer:\n* Only output the transcription, with no newlines.\n* When transcribing numbers, write the digits, i.e. write 1.7 and not one point seven, and write 3 instead of three."},
-            {"type": "audio", "audio": "https://ai.google.dev/gemma/docs/audio/roses-are.wav"},
+            {"type": "audio", "audio": "https://gemma.ai/assets/capabilities/roses-are.wav"},
         ]
     }
 ]
@@ -176,7 +176,7 @@ pipe(text=messages, return_full_text=False, generate_kwargs=gen_kwargs)
     'content': [{'type': 'text',
       'text': 'Transcribe the following speech segment in its original language. Follow these specific instructions for formatting the answer:\n* Only output the transcription, with no newlines.\n* When transcribing numbers, write the digits, i.e. write 1.7 and not one point seven, and write 3 instead of three.'},
      {'type': 'audio',
-      'audio': 'https://ai.google.dev/gemma/docs/audio/roses-are.wav'}]}],
+      'audio': 'https://gemma.ai/assets/capabilities/roses-are.wav'}]}],
   'generated_text': 'Roses are red, violets are blue.<turn|>'}]
 ```
 
diff --git a/docs/deployment/keras.md b/docs/deployment/keras.md
index 8278b98..139fc16 100644
--- a/docs/deployment/keras.md
+++ b/docs/deployment/keras.md
@@ -1,6 +1,6 @@
 ---
 title: Run Gemma with Keras
-description: Generating content, summarizing, and analysing content are just some of the tasks you can accomplish with Gemma open models.
+description: "Run Gemma 4 inference with Keras and KerasHub for text and multimodal image generation in Python."
 models: [gemma-4]
 frameworks: [keras]
 source_url: "https://ai.google.dev/gemma/docs/core/keras_inference.md.txt"
@@ -269,7 +269,7 @@ Load the image and format the data so the model can process it. Use `read_image(
 from matplotlib import pyplot as plt
 
 image = read_image(
-    "https://ai.google.dev/gemma/docs/images/thali-indian-plate.jpg"
+    "https://gemma.ai/assets/deployment/thali-indian-plate.jpg"
 )
 plt.imshow(image)
 ```
diff --git a/docs/deployment/litert.md b/docs/deployment/litert.md
index 95a631b..147e54f 100644
--- a/docs/deployment/litert.md
+++ b/docs/deployment/litert.md
@@ -1,5 +1,5 @@
 ---
-title: LiteRT-LM
+title: Run Gemma on-device with LiteRT-LM
 description: Run Gemma 4 on-device with LiteRT-LM — CLI install, prebuilt .litertlm weights, and exporting your own fine-tune.
 models: [gemma-4]
 frameworks: [litert]
diff --git a/docs/deployment/llamacpp.md b/docs/deployment/llamacpp.md
index 1c93499..658f618 100644
--- a/docs/deployment/llamacpp.md
+++ b/docs/deployment/llamacpp.md
@@ -1,5 +1,5 @@
 ---
-title: llama.cpp
+title: Run Gemma with llama.cpp
 description: Run Gemma 4 with llama.cpp — GGUF repos for every size, image/audio/video input, and the OpenAI-compatible server.
 models: [gemma-4]
 frameworks: [llamacpp]
diff --git a/docs/deployment/lmstudio.md b/docs/deployment/lmstudio.md
index b329157..0873b19 100644
--- a/docs/deployment/lmstudio.md
+++ b/docs/deployment/lmstudio.md
@@ -1,5 +1,5 @@
 ---
-title: LM Studio
+title: Run Gemma with LM Studio
 description: Run Gemma 4 with LM Studio — desktop app, lms CLI, and the OpenAI-compatible local server.
 models: [gemma-4]
 frameworks: [lmstudio]
diff --git a/docs/deployment/mlx.md b/docs/deployment/mlx.md
index ecf49bd..b802fe6 100644
--- a/docs/deployment/mlx.md
+++ b/docs/deployment/mlx.md
@@ -1,5 +1,5 @@
 ---
-title: Apple MLX
+title: Run Gemma with MLX
 description: Run Gemma 4 on Apple silicon with MLX — text and vision generation, model repos, and the OpenAI-compatible server.
 models: [gemma-4]
 frameworks: [mlx]
diff --git a/docs/deployment/ollama.md b/docs/deployment/ollama.md
index d64ed67..ac30581 100644
--- a/docs/deployment/ollama.md
+++ b/docs/deployment/ollama.md
@@ -1,5 +1,5 @@
 ---
-title: Ollama
+title: Run Gemma with Ollama
 description: Run Gemma 4 locally with Ollama — model tags, image input, and the OpenAI-compatible server.
 models: [gemma-4]
 frameworks: [ollama]
diff --git a/docs/deployment/sglang.md b/docs/deployment/sglang.md
index 8443b15..116038d 100644
--- a/docs/deployment/sglang.md
+++ b/docs/deployment/sglang.md
@@ -1,5 +1,5 @@
 ---
-title: SGLang
+title: Serve Gemma with SGLang
 description: Serve Gemma 4 with SGLang — install, launch_server, model repositories, and the OpenAI-compatible API.
 models: [gemma-4]
 frameworks: [sglang]
diff --git a/docs/deployment/unsloth.md b/docs/deployment/unsloth.md
index 65280f7..ac4be19 100644
--- a/docs/deployment/unsloth.md
+++ b/docs/deployment/unsloth.md
@@ -1,5 +1,5 @@
 ---
-title: Unsloth
+title: Run Gemma with Unsloth
 description: Run Gemma 4 with Unsloth — the desktop Studio app, dynamic-quant GGUFs for llama.cpp, and 2x faster fine-tuning.
 models: [gemma-4]
 frameworks: [unsloth, llamacpp]
diff --git a/docs/deployment/vllm.md b/docs/deployment/vllm.md
index 5ec7bb2..8599966 100644
--- a/docs/deployment/vllm.md
+++ b/docs/deployment/vllm.md
@@ -1,5 +1,5 @@
 ---
-title: vLLM
+title: Serve Gemma with vLLM
 description: Serve Gemma 4 with vLLM — install, vllm serve, model repositories, and the OpenAI-compatible API.
 models: [gemma-4]
 frameworks: [vllm]
diff --git a/docs/get-started/capabilities/audio.md b/docs/get-started/capabilities/audio.md
index f15b526..ffb8da5 100644
--- a/docs/get-started/capabilities/audio.md
+++ b/docs/get-started/capabilities/audio.md
@@ -1,6 +1,6 @@
 ---
 title: Audio understanding
-description: Starting with Gemma 3n, you can use audio directly into your prompts and workflows.
+description: "Process and transcribe speech, music, and sound inputs with Gemma 4 and Gemma 3n multimodal audio understanding."
 models: [gemma-4]
 requires: audio
 frameworks: [transformers]
@@ -185,7 +185,7 @@ messages = [
         "role": "user",
         "content": [
             {"type": "text", "text": "Transcribe the following speech segment in English, then translate it into Korean. When formatting the answer, first output the transcription in English, then one newline, then output the string 'Korean: ', then the translation in Korean."},
-            {"type": "audio", "audio": "https://ai.google.dev/gemma/docs/audio/roses-are.wav"},
+            {"type": "audio", "audio": "https://gemma.ai/assets/capabilities/roses-are.wav"},
         ]
     }
 ]
diff --git a/docs/get-started/capabilities/function-calling.md b/docs/get-started/capabilities/function-calling.md
index 4bcedf9..fef94f2 100644
--- a/docs/get-started/capabilities/function-calling.md
+++ b/docs/get-started/capabilities/function-calling.md
@@ -1,6 +1,6 @@
 ---
 title: Tool calling
-description: When using a generative artificial intelligence (AI) model such as Gemma, you may want to use the model to operate programming interfaces in order to complete tasks or answer questions.
+description: "Build agentic workflows with Gemma 4 native tool calling and structured function schemas in Hugging Face Transformers."
 models: [gemma-4]
 frameworks: [transformers]
 colab: https://colab.research.google.com/github/google-gemma/docs/blob/main/docs/notebooks/function-calling-gemma-4-e2b.ipynb
diff --git a/docs/get-started/capabilities/image.md b/docs/get-started/capabilities/image.md
index 8ffc8ab..12c24f8 100644
--- a/docs/get-started/capabilities/image.md
+++ b/docs/get-started/capabilities/image.md
@@ -1,6 +1,6 @@
 ---
 title: Image understanding
-description: You can use Gemma 3 and later models to analyze and understand the content of images.
+description: "Analyze images with Gemma 4 using variable aspect ratios, configurable vision token budgets, and multi-image prompts."
 models: [gemma-4]
 frameworks: [transformers]
 colab: https://colab.research.google.com/github/google-gemma/docs/blob/main/docs/notebooks/image-gemma-4-e2b.ipynb
@@ -103,8 +103,8 @@ config.max_new_tokens = 512
 gen_kwargs = dict(generation_config=config)
 
 img_urls = [
-    "https://ai.google.dev/gemma/docs/capabilities/vision/images/surprise.png",
-    "https://ai.google.dev/gemma/docs/capabilities/vision/images/kitchen.jpg",
+    "https://gemma.ai/assets/capabilities/surprise.png",
+    "https://gemma.ai/assets/capabilities/kitchen.jpg",
 ]
 for img in img_urls:
     display(Image.open(requests.get(img, stream=True).raw))
@@ -196,7 +196,7 @@ config = GenerationConfig.from_pretrained(MODEL_ID)
 config.max_new_tokens = 512
 gen_kwargs = dict(generation_config=config)
 
-img_url = "https://ai.google.dev/gemma/docs/capabilities/vision/images/cat.png"
+img_url = "https://gemma.ai/assets/capabilities/cat.png"
 input_image = Image.open(requests.get(img_url, stream=True).raw)
 display(input_image)
 
diff --git a/docs/get-started/capabilities/video.md b/docs/get-started/capabilities/video.md
index 609072a..4024529 100644
--- a/docs/get-started/capabilities/video.md
+++ b/docs/get-started/capabilities/video.md
@@ -1,6 +1,6 @@
 ---
 title: Video understanding
-description: Video data is a rich source of information that can help you accomplish tasks and understand the world around you.
+description: "Analyze video clips and interleaved frame sequences with Gemma 4 multimodal vision-language models."
 models: [gemma-4]
 frameworks: [transformers]
 colab: https://colab.research.google.com/github/google-gemma/docs/blob/main/docs/notebooks/video-gemma-4-e2b.ipynb
diff --git a/docs/llms.txt b/docs/llms.txt
index 9869569..4d52d97 100644
--- a/docs/llms.txt
+++ b/docs/llms.txt
@@ -12,26 +12,26 @@ Follow the Markdown links below to read canonical page content without cloning t
 - [Best practices](https://gemma.ai/get-started/best-practices.md): Sampling parameters, thinking-mode tokens, modality order and image token budgets that get the best results out of Gemma 4.
 - [Inference](https://gemma.ai/get-started/capabilities/text.md): Basic text inference with Gemma 4 in Hugging Face Transformers: single-turn prompts, multi-turn conversations, and system instructions.
 - [Reasoning](https://gemma.ai/get-started/capabilities/thinking.md): Use Gemma 4's thinking mode to reason before answering, for text and image prompts, and parse the reasoning apart from the answer.
-- [Tool calling](https://gemma.ai/get-started/capabilities/function-calling.md): When using a generative artificial intelligence (AI) model such as Gemma, you may want to use the model to operate programming interfaces in order to complete tasks or answer questions.
+- [Tool calling](https://gemma.ai/get-started/capabilities/function-calling.md): Build agentic workflows with Gemma 4 native tool calling and structured function schemas in Hugging Face Transformers.
 - [Vision overview](https://gemma.ai/get-started/capabilities/vision.md): Prepare and pass visual data to Gemma 4 for object detection, OCR, visual question answering, captioning, and multi-image reasoning.
-- [Image understanding](https://gemma.ai/get-started/capabilities/image.md): You can use Gemma 3 and later models to analyze and understand the content of images.
-- [Video understanding](https://gemma.ai/get-started/capabilities/video.md): Video data is a rich source of information that can help you accomplish tasks and understand the world around you.
-- [Audio understanding](https://gemma.ai/get-started/capabilities/audio.md): Starting with Gemma 3n, you can use audio directly into your prompts and workflows.
+- [Image understanding](https://gemma.ai/get-started/capabilities/image.md): Analyze images with Gemma 4 using variable aspect ratios, configurable vision token budgets, and multi-image prompts.
+- [Video understanding](https://gemma.ai/get-started/capabilities/video.md): Analyze video clips and interleaved frame sequences with Gemma 4 multimodal vision-language models.
+- [Audio understanding](https://gemma.ai/get-started/capabilities/audio.md): Process and transcribe speech, music, and sound inputs with Gemma 4 and Gemma 3n multimodal audio understanding.
 - [Prompt Formatting](https://gemma.ai/get-started/prompt-formatting.md): The control tokens Gemma 4 uses to structure a prompt, and the older start_of_turn format used by Gemma 1 to 3.
 - [Speed-up Gemma 4 with Multi-Token Prediction](https://gemma.ai/get-started/capabilities/mtp.md): In Gemma 4, Multi-Token Prediction (MTP) is the specific architecture used to enable highly efficient Speculative Decoding.
 - [Quantization-Aware Training](https://gemma.ai/get-started/qat.md): Which Gemma 4 quantized weights to download for each engine, and why QAT beats post-training quantization at 4-bit.
 
 ## Models
 - [Gemma Model Family](https://gemma.ai/models.md): What Gemma is, how the generations relate, and which variant was built from which base model.
-- [Gemma 4](https://gemma.ai/models/gemma-4.md): Gemma is a family of generative artificial intelligence models and you can use them in a wide variety of generation tasks, including question answering, summarization, and reasoning.
-- [Gemma 4 — Model card](https://gemma.ai/models/gemma-4/model-card.md): Gemma is a family of open models built by Google DeepMind.
+- [Gemma 4 model overview](https://gemma.ai/models/gemma-4.md): Overview of Google's Gemma 4 open-weight multimodal models (E2B, E4B, 12B, 26B A4B, and 31B): capabilities, context windows, VRAM, and QAT checkpoints.
+- [Gemma 4 model card](https://gemma.ai/models/gemma-4/model-card.md): Model card for Gemma 4 (E2B, E4B, 12B, 26B A4B, and 31B): architecture, multimodal inputs and outputs, training data, and evaluation benchmarks.
 - [Gemma 3 model card](https://gemma.ai/models/gemma-3.md): Model card for the multimodal Gemma 3 models: 128K context, 140+ languages, inputs and outputs, training data, and evaluation.
 - [Gemma 2 model card](https://gemma.ai/models/gemma-2.md): Model card for Gemma 2 text-to-text models: description, inputs and outputs, training data, and evaluation.
 - [Gemma model card](https://gemma.ai/models/gemma-1.md): Model card for the first-generation Gemma text-to-text models: description, inputs and outputs, training data, and evaluation.
 - [Gemma 3n model overview](https://gemma.ai/models/gemma-3n.md): Gemma 3n is a generative AI model optimized for use in everyday devices, such as phones, laptops, and tablets.
 - [Gemma 3n model card](https://gemma.ai/models/gemma-3n/model-card.md): Model card for Gemma 3n: on-device inputs and outputs, training data, evaluation results, and intended use.
 - [DiffusionGemma model overview](https://gemma.ai/models/diffusiongemma.md): DiffusionGemma is an experimental open model that explores text diffusion, an exceptionally fast approach to text generation.
-- [DiffusionGemma model card](https://gemma.ai/models/diffusiongemma/model-card.md): DiffusionGemma is a generative model built by Google DeepMind.
+- [DiffusionGemma model card](https://gemma.ai/models/diffusiongemma/model-card.md): Model card for DiffusionGemma (26B A4B MoE): discrete text diffusion architecture, multimodal inputs, training data, and evaluation benchmarks.
 - [DiffusionGemma — Diffusion in Text Generation Explained](https://gemma.ai/models/diffusiongemma/explained.md): To understand DiffusionGemma, it helps to examine the core limitations of standard language models and how text-based diffusion differs.
 - [DiffusionGemma — Generate output](https://gemma.ai/models/diffusiongemma/inference.md): Generate text with DiffusionGemma using Hugging Face Transformers, including the GPU memory the model requires.
 - [FunctionGemma model overview](https://gemma.ai/models/functiongemma.md): FunctionGemma is a specialized version of our Gemma 3 270M model tuned for function calling.
@@ -67,20 +67,20 @@ Follow the Markdown links below to read canonical page content without cloning t
 
 ## Run
 - [Overview](https://gemma.ai/deployment.md): Choose an execution framework and a Gemma variant for your hardware, then run your first generation request.
-- [Ollama](https://gemma.ai/deployment/ollama.md): Run Gemma 4 locally with Ollama — model tags, image input, and the OpenAI-compatible server.
-- [llama.cpp](https://gemma.ai/deployment/llamacpp.md): Run Gemma 4 with llama.cpp — GGUF repos for every size, image/audio/video input, and the OpenAI-compatible server.
-- [LM Studio](https://gemma.ai/deployment/lmstudio.md): Run Gemma 4 with LM Studio — desktop app, lms CLI, and the OpenAI-compatible local server.
-- [Apple MLX](https://gemma.ai/deployment/mlx.md): Run Gemma 4 on Apple silicon with MLX — text and vision generation, model repos, and the OpenAI-compatible server.
-- [Run Gemma with Hugging Face Transformers](https://gemma.ai/deployment/huggingface.md): Generating text, summarizing, and analysing content are just some of the tasks you can accomplish with Gemma open models.
-- [Run Gemma with Keras](https://gemma.ai/deployment/keras.md): Generating content, summarizing, and analysing content are just some of the tasks you can accomplish with Gemma open models.
-- [LiteRT-LM](https://gemma.ai/deployment/litert.md): Run Gemma 4 on-device with LiteRT-LM — CLI install, prebuilt .litertlm weights, and exporting your own fine-tune.
-- [AI Edge Gallery](https://gemma.ai/deployment/ai-edge-gallery.md): Run Gemma 4 on an Android or iOS phone with no code — chat, vision, audio transcription and benchmarking, fully offline.
-- [vLLM](https://gemma.ai/deployment/vllm.md): Serve Gemma 4 with vLLM — install, vllm serve, model repositories, and the OpenAI-compatible API.
-- [SGLang](https://gemma.ai/deployment/sglang.md): Serve Gemma 4 with SGLang — install, launch_server, model repositories, and the OpenAI-compatible API.
-- [Gemini API](https://gemma.ai/deployment/gemini-api.md): Call hosted Gemma 4 models through the Gemini API instead of running and serving your own instance.
-- [Google Cloud](https://gemma.ai/deployment/google-cloud.md): Deploy, serve, and fine-tune Gemma 4 on Google Cloud: Cloud Run, GKE, the Agent Development Kit, vLLM on TPUs, MaxText, and Sovereign Cloud.
+- [Run Gemma with Ollama](https://gemma.ai/deployment/ollama.md): Run Gemma 4 locally with Ollama — model tags, image input, and the OpenAI-compatible server.
+- [Run Gemma with llama.cpp](https://gemma.ai/deployment/llamacpp.md): Run Gemma 4 with llama.cpp — GGUF repos for every size, image/audio/video input, and the OpenAI-compatible server.
+- [Run Gemma with LM Studio](https://gemma.ai/deployment/lmstudio.md): Run Gemma 4 with LM Studio — desktop app, lms CLI, and the OpenAI-compatible local server.
+- [Run Gemma with MLX](https://gemma.ai/deployment/mlx.md): Run Gemma 4 on Apple silicon with MLX — text and vision generation, model repos, and the OpenAI-compatible server.
+- [Run Gemma with Hugging Face Transformers](https://gemma.ai/deployment/huggingface.md): Run Gemma 4 with Hugging Face Transformers in Python for text, image, audio, and multi-turn chat generation.
+- [Run Gemma with Keras](https://gemma.ai/deployment/keras.md): Run Gemma 4 inference with Keras and KerasHub for text and multimodal image generation in Python.
+- [Run Gemma on-device with LiteRT-LM](https://gemma.ai/deployment/litert.md): Run Gemma 4 on-device with LiteRT-LM — CLI install, prebuilt .litertlm weights, and exporting your own fine-tune.
+- [Try Gemma on your phone with AI Edge Gallery](https://gemma.ai/deployment/ai-edge-gallery.md): Run Gemma 4 on an Android or iOS phone with no code — chat, vision, audio transcription and benchmarking, fully offline.
+- [Serve Gemma with vLLM](https://gemma.ai/deployment/vllm.md): Serve Gemma 4 with vLLM — install, vllm serve, model repositories, and the OpenAI-compatible API.
+- [Serve Gemma with SGLang](https://gemma.ai/deployment/sglang.md): Serve Gemma 4 with SGLang — install, launch_server, model repositories, and the OpenAI-compatible API.
+- [Run Gemma with the Gemini API](https://gemma.ai/deployment/gemini-api.md): Call hosted Gemma 4 models through the Gemini API instead of running and serving your own instance.
+- [Gemma on Google Cloud](https://gemma.ai/deployment/google-cloud.md): Deploy, serve, and fine-tune Gemma 4 on Google Cloud: Cloud Run, GKE, the Agent Development Kit, vLLM on TPUs, MaxText, and Sovereign Cloud.
 - [Get started with Gemma and LangChain](https://gemma.ai/deployment/langchain.md): This tutorial shows you how to get started with Gemma and LangChain, running in Google Cloud or in your Colab environment.
-- [Unsloth](https://gemma.ai/deployment/unsloth.md): Run Gemma 4 with Unsloth — the desktop Studio app, dynamic-quant GGUFs for llama.cpp, and 2x faster fine-tuning.
+- [Run Gemma with Unsloth](https://gemma.ai/deployment/unsloth.md): Run Gemma 4 with Unsloth — the desktop Studio app, dynamic-quant GGUFs for llama.cpp, and 2x faster fine-tuning.
 
 ## Fine-tune
 - [Gemma model fine-tuning](https://gemma.ai/tuning.md): Fine-tuning a generative artificial intelligence (AI) model such as Gemma modifies the behavior of the model.
@@ -92,10 +92,10 @@ Follow the Markdown links below to read canonical page content without cloning t
 
 ## Learn
 - [Learn](https://gemma.ai/learn.md): Illustrated walkthroughs of how the Gemma models work, from the architecture down to individual mechanisms.
-- [A Visual Guide to Gemma 4](https://gemma.ai/tutorials/visual-guides/gemma-4.md): A Visual Guide to Gemma 4 — a visual walkthrough.
-- [A Visual Guide to Gemma 4 12B](https://gemma.ai/tutorials/visual-guides/gemma-4-12b.md): A Visual Guide to Gemma 4 12B — a visual walkthrough.
+- [A Visual Guide to Gemma 4](https://gemma.ai/tutorials/visual-guides/gemma-4.md): Illustrated guide to the Gemma 4 architecture: dense and 26B A4B MoE models, Per-Layer Embeddings (E2B/E4B), vision and audio encoders, and KV cache.
+- [A Visual Guide to Gemma 4 12B](https://gemma.ai/tutorials/visual-guides/gemma-4-12b.md): How Gemma 4 12B achieves fast, unified multimodal understanding of text, images, and audio using an encoder-free architecture with linear projections.
 - [A Visual Guide to MTP](https://gemma.ai/tutorials/visual-guides/mtp.md): How multi-token prediction and speculative decoding make Gemma 4 faster — drafter models, shared KV cache, and the clustered LM head.
-- [A Visual Guide to DiffusionGemma](https://gemma.ai/tutorials/visual-guides/diffusiongemma.md): A Visual Guide to DiffusionGemma — a visual walkthrough.
+- [A Visual Guide to DiffusionGemma](https://gemma.ai/tutorials/visual-guides/diffusiongemma.md): How DiffusionGemma uses discrete text diffusion, block-autoregressive multi-canvas sampling, and a 26B A4B MoE architecture to generate tokens in parallel.
 - [Overview of Gemma Architectures](https://gemma.ai/tutorials/gemma-explained/overview.md): Text-to-text models and specialization for coding.
 - [What’s New in Gemma 2](https://gemma.ai/tutorials/gemma-explained/gemma-2.md): Updated architecture with knowledge distillation for 2B and 9B models.
 - [RecurrentGemma Architecture](https://gemma.ai/tutorials/gemma-explained/recurrentgemma.md): The Griffin architecture mixing local attention and linear recurrences for fast long-sequence inference.
diff --git a/docs/models/diffusiongemma/model-card.md b/docs/models/diffusiongemma/model-card.md
index fed31be..4aa8d8e 100644
--- a/docs/models/diffusiongemma/model-card.md
+++ b/docs/models/diffusiongemma/model-card.md
@@ -1,6 +1,6 @@
 ---
 title: DiffusionGemma model card
-description: DiffusionGemma is a generative model built by Google DeepMind.
+description: "Model card for DiffusionGemma (26B A4B MoE): discrete text diffusion architecture, multimodal inputs, training data, and evaluation benchmarks."
 models: [diffusiongemma]
 source_url: "https://ai.google.dev/gemma/docs/diffusiongemma/model_card.md.txt"
 ---
diff --git a/docs/models/embeddinggemma/index.md b/docs/models/embeddinggemma/index.md
index 5ce7bd2..9708271 100644
--- a/docs/models/embeddinggemma/index.md
+++ b/docs/models/embeddinggemma/index.md
@@ -4,7 +4,7 @@ description: EmbeddingGemma 2 is a lightweight, open-source 740M parameter model
 models: [embeddinggemma-2, embeddinggemma]
 source_url: "https://ai.google.dev/gemma/docs/embeddinggemma.md.txt"
 ---
-# EmbeddingGemma model overview {: #hidden-title style="display: none;" }
+# EmbeddingGemma model overview
 
 <div class="video-embed">
   <iframe src="https://www.youtube.com/embed/anPsS6huQk0"
diff --git a/docs/models/embeddinggemma/multimodal.md b/docs/models/embeddinggemma/multimodal.md
index 3418986..b7cb89d 100644
--- a/docs/models/embeddinggemma/multimodal.md
+++ b/docs/models/embeddinggemma/multimodal.md
@@ -214,7 +214,7 @@ Pass audio inputs as `{"audio": <source>}`. Supported audio sources include loca
 ```python { .notebook-cell }
 from IPython.display import display, Audio
 
-aud_url = "https://ai.google.dev/gemma/docs/audio/roses-are.wav"
+aud_url = "https://gemma.ai/assets/capabilities/roses-are.wav"
 
 print("Input Audio:")
 display(Audio(aud_url))
@@ -363,7 +363,7 @@ EmbeddingGemma 2 supports rich, mixed-modality inputs within a single query pass
 ```python { .notebook-cell }
 img_url = "https://raw.githubusercontent.com/google-gemma/cookbook/refs/heads/main/apps/sample-data/GoldenGate.png"
 example_img_url = "https://raw.githubusercontent.com/google-gemma/cookbook/refs/heads/main/apps/sample-data/kitchen_painting.jpg"
-aud_url = "https://ai.google.dev/gemma/docs/audio/roses-are.wav"
+aud_url = "https://gemma.ai/assets/capabilities/roses-are.wav"
 vid_url = "https://github.com/bebechien/gemma/raw/refs/heads/main/videos/cat_dozing.mp4"
 
 # 1. Multimodal query combining text, multiple images, and audio
diff --git a/docs/models/functiongemma/index.md b/docs/models/functiongemma/index.md
index ebedd39..a99a5f8 100644
--- a/docs/models/functiongemma/index.md
+++ b/docs/models/functiongemma/index.md
@@ -4,7 +4,7 @@ description: FunctionGemma is a specialized version of our Gemma 3 270M model tu
 models: [functiongemma]
 source_url: "https://ai.google.dev/gemma/docs/functiongemma.md.txt"
 ---
-# FunctionGemma model overview {: #hidden-title style="display: none;" }
+# FunctionGemma model overview
 
 <div class="video-embed">
   <iframe src="https://www.youtube.com/embed/-Tgc_9uYJLI"
diff --git a/docs/models/gemma-3n/index.md b/docs/models/gemma-3n/index.md
index 49a42c6..39e503c 100644
--- a/docs/models/gemma-3n/index.md
+++ b/docs/models/gemma-3n/index.md
@@ -4,7 +4,7 @@ description: Gemma 3n is a generative AI model optimized for use in everyday dev
 models: [gemma-3n]
 source_url: "https://ai.google.dev/gemma/docs/gemma-3n.md.txt"
 ---
-# Gemma 3n model overview {: #hidden-title style="display: none;" }
+# Gemma 3n model overview
 
 <div class="video-embed">
   <iframe src="https://www.youtube.com/embed/eJFJRyXEHZ0"
diff --git a/docs/models/gemma-4/index.md b/docs/models/gemma-4/index.md
index b28757b..801985f 100644
--- a/docs/models/gemma-4/index.md
+++ b/docs/models/gemma-4/index.md
@@ -1,10 +1,10 @@
 ---
-title: Gemma 4
-description: Gemma is a family of generative artificial intelligence models and you can use them in a wide variety of generation tasks, including question answering, summarization, and reasoning.
+title: Gemma 4 model overview
+description: "Overview of Google's Gemma 4 open-weight multimodal models (E2B, E4B, 12B, 26B A4B, and 31B): capabilities, context windows, VRAM, and QAT checkpoints."
 models: [gemma-4]
 source_url: "https://ai.google.dev/gemma/docs/core.md.txt"
 ---
-# Gemma 4 {: #hidden-title style="display: none;" }
+# Gemma 4 model overview
 
 <div class="video-embed">
   <iframe src="https://www.youtube.com/embed/jZVBoFOJK-Q"
diff --git a/docs/models/gemma-4/model-card.md b/docs/models/gemma-4/model-card.md
index d6a3f1a..b1ee60c 100644
--- a/docs/models/gemma-4/model-card.md
+++ b/docs/models/gemma-4/model-card.md
@@ -1,6 +1,6 @@
 ---
-title: Model card
-description: Gemma is a family of open models built by Google DeepMind.
+title: Gemma 4 model card
+description: "Model card for Gemma 4 (E2B, E4B, 12B, 26B A4B, and 31B): architecture, multimodal inputs and outputs, training data, and evaluation benchmarks."
 models: [gemma-4]
 source_url: "https://ai.google.dev/gemma/docs/core/model_card_4.md.txt"
 ---
diff --git a/docs/models/shieldgemma/index.md b/docs/models/shieldgemma/index.md
index 1fcc3ed..5e90094 100644
--- a/docs/models/shieldgemma/index.md
+++ b/docs/models/shieldgemma/index.md
@@ -4,7 +4,7 @@ description: ShieldGemma is a set of instruction tuned models for evaluating the
 models: [shieldgemma-2, shieldgemma]
 source_url: "https://ai.google.dev/gemma/docs/shieldgemma.md.txt"
 ---
-# ShieldGemma {: #hidden-title style="display: none;" }
+# ShieldGemma
 
 <div class="video-embed">
   <iframe src="https://www.youtube.com/embed/BgWO2iVPlyw"
diff --git a/docs/notebooks/audio-gemma-4-12b.ipynb b/docs/notebooks/audio-gemma-4-12b.ipynb
index d2488c0..db5d200 100644
--- a/docs/notebooks/audio-gemma-4-12b.ipynb
+++ b/docs/notebooks/audio-gemma-4-12b.ipynb
@@ -246,7 +246,7 @@
     "        \"role\": \"user\",\n",
     "        \"content\": [\n",
     "            {\"type\": \"text\", \"text\": \"Transcribe the following speech segment in English, then translate it into Korean. When formatting the answer, first output the transcription in English, then one newline, then output the string 'Korean: ', then the translation in Korean.\"},\n",
-    "            {\"type\": \"audio\", \"audio\": \"https://ai.google.dev/gemma/docs/audio/roses-are.wav\"},\n",
+    "            {\"type\": \"audio\", \"audio\": \"https://gemma.ai/assets/capabilities/roses-are.wav\"},\n",
     "        ]\n",
     "    }\n",
     "]\n",
diff --git a/docs/notebooks/audio-gemma-4-e2b.ipynb b/docs/notebooks/audio-gemma-4-e2b.ipynb
index 7f7615e..e8b9508 100644
--- a/docs/notebooks/audio-gemma-4-e2b.ipynb
+++ b/docs/notebooks/audio-gemma-4-e2b.ipynb
@@ -246,7 +246,7 @@
     "        \"role\": \"user\",\n",
     "        \"content\": [\n",
     "            {\"type\": \"text\", \"text\": \"Transcribe the following speech segment in English, then translate it into Korean. When formatting the answer, first output the transcription in English, then one newline, then output the string 'Korean: ', then the translation in Korean.\"},\n",
-    "            {\"type\": \"audio\", \"audio\": \"https://ai.google.dev/gemma/docs/audio/roses-are.wav\"},\n",
+    "            {\"type\": \"audio\", \"audio\": \"https://gemma.ai/assets/capabilities/roses-are.wav\"},\n",
     "        ]\n",
     "    }\n",
     "]\n",
diff --git a/docs/notebooks/audio-gemma-4-e4b.ipynb b/docs/notebooks/audio-gemma-4-e4b.ipynb
index bf44fa8..0dbbf10 100644
--- a/docs/notebooks/audio-gemma-4-e4b.ipynb
+++ b/docs/notebooks/audio-gemma-4-e4b.ipynb
@@ -246,7 +246,7 @@
     "        \"role\": \"user\",\n",
     "        \"content\": [\n",
     "            {\"type\": \"text\", \"text\": \"Transcribe the following speech segment in English, then translate it into Korean. When formatting the answer, first output the transcription in English, then one newline, then output the string 'Korean: ', then the translation in Korean.\"},\n",
-    "            {\"type\": \"audio\", \"audio\": \"https://ai.google.dev/gemma/docs/audio/roses-are.wav\"},\n",
+    "            {\"type\": \"audio\", \"audio\": \"https://gemma.ai/assets/capabilities/roses-are.wav\"},\n",
     "        ]\n",
     "    }\n",
     "]\n",
diff --git a/docs/notebooks/embeddinggemma-multimodal.ipynb b/docs/notebooks/embeddinggemma-multimodal.ipynb
index f2203b8..7258901 100644
--- a/docs/notebooks/embeddinggemma-multimodal.ipynb
+++ b/docs/notebooks/embeddinggemma-multimodal.ipynb
@@ -237,7 +237,7 @@
    "source": [
     "from IPython.display import display, Audio\n",
     "\n",
-    "aud_url = \"https://ai.google.dev/gemma/docs/audio/roses-are.wav\"\n",
+    "aud_url = \"https://gemma.ai/assets/capabilities/roses-are.wav\"\n",
     "\n",
     "print(\"Input Audio:\")\n",
     "display(Audio(aud_url))\n",
@@ -364,7 +364,7 @@
    "source": [
     "img_url = \"https://raw.githubusercontent.com/google-gemma/cookbook/refs/heads/main/apps/sample-data/GoldenGate.png\"\n",
     "example_img_url = \"https://raw.githubusercontent.com/google-gemma/cookbook/refs/heads/main/apps/sample-data/kitchen_painting.jpg\"\n",
-    "aud_url = \"https://ai.google.dev/gemma/docs/audio/roses-are.wav\"\n",
+    "aud_url = \"https://gemma.ai/assets/capabilities/roses-are.wav\"\n",
     "vid_url = \"https://github.com/bebechien/gemma/raw/refs/heads/main/videos/cat_dozing.mp4\"\n",
     "\n",
     "# 1. Multimodal query combining text, multiple images, and audio\n",
diff --git a/docs/notebooks/image-gemma-4-12b.ipynb b/docs/notebooks/image-gemma-4-12b.ipynb
index 4a13dd6..43b56fb 100644
--- a/docs/notebooks/image-gemma-4-12b.ipynb
+++ b/docs/notebooks/image-gemma-4-12b.ipynb
@@ -159,8 +159,8 @@
     "gen_kwargs = dict(generation_config=config)\n",
     "\n",
     "img_urls = [\n",
-    "    \"https://ai.google.dev/gemma/docs/capabilities/vision/images/surprise.png\",\n",
-    "    \"https://ai.google.dev/gemma/docs/capabilities/vision/images/kitchen.jpg\",\n",
+    "    \"https://gemma.ai/assets/capabilities/surprise.png\",\n",
+    "    \"https://gemma.ai/assets/capabilities/kitchen.jpg\",\n",
     "]\n",
     "for img in img_urls:\n",
     "    display(Image.open(requests.get(img, stream=True).raw))\n",
@@ -205,7 +205,7 @@
     "config.max_new_tokens = 512\n",
     "gen_kwargs = dict(generation_config=config)\n",
     "\n",
-    "img_url = \"https://ai.google.dev/gemma/docs/capabilities/vision/images/cat.png\"\n",
+    "img_url = \"https://gemma.ai/assets/capabilities/cat.png\"\n",
     "input_image = Image.open(requests.get(img_url, stream=True).raw)\n",
     "display(input_image)\n",
     "\n",
diff --git a/docs/notebooks/image-gemma-4-26b-a4b.ipynb b/docs/notebooks/image-gemma-4-26b-a4b.ipynb
index 8f480e9..347e15b 100644
--- a/docs/notebooks/image-gemma-4-26b-a4b.ipynb
+++ b/docs/notebooks/image-gemma-4-26b-a4b.ipynb
@@ -159,8 +159,8 @@
     "gen_kwargs = dict(generation_config=config)\n",
     "\n",
     "img_urls = [\n",
-    "    \"https://ai.google.dev/gemma/docs/capabilities/vision/images/surprise.png\",\n",
-    "    \"https://ai.google.dev/gemma/docs/capabilities/vision/images/kitchen.jpg\",\n",
+    "    \"https://gemma.ai/assets/capabilities/surprise.png\",\n",
+    "    \"https://gemma.ai/assets/capabilities/kitchen.jpg\",\n",
     "]\n",
     "for img in img_urls:\n",
     "    display(Image.open(requests.get(img, stream=True).raw))\n",
@@ -205,7 +205,7 @@
     "config.max_new_tokens = 512\n",
     "gen_kwargs = dict(generation_config=config)\n",
     "\n",
-    "img_url = \"https://ai.google.dev/gemma/docs/capabilities/vision/images/cat.png\"\n",
+    "img_url = \"https://gemma.ai/assets/capabilities/cat.png\"\n",
     "input_image = Image.open(requests.get(img_url, stream=True).raw)\n",
     "display(input_image)\n",
     "\n",
diff --git a/docs/notebooks/image-gemma-4-31b.ipynb b/docs/notebooks/image-gemma-4-31b.ipynb
index 4b629ed..060dcec 100644
--- a/docs/notebooks/image-gemma-4-31b.ipynb
+++ b/docs/notebooks/image-gemma-4-31b.ipynb
@@ -159,8 +159,8 @@
     "gen_kwargs = dict(generation_config=config)\n",
     "\n",
     "img_urls = [\n",
-    "    \"https://ai.google.dev/gemma/docs/capabilities/vision/images/surprise.png\",\n",
-    "    \"https://ai.google.dev/gemma/docs/capabilities/vision/images/kitchen.jpg\",\n",
+    "    \"https://gemma.ai/assets/capabilities/surprise.png\",\n",
+    "    \"https://gemma.ai/assets/capabilities/kitchen.jpg\",\n",
     "]\n",
     "for img in img_urls:\n",
     "    display(Image.open(requests.get(img, stream=True).raw))\n",
@@ -205,7 +205,7 @@
     "config.max_new_tokens = 512\n",
     "gen_kwargs = dict(generation_config=config)\n",
     "\n",
-    "img_url = \"https://ai.google.dev/gemma/docs/capabilities/vision/images/cat.png\"\n",
+    "img_url = \"https://gemma.ai/assets/capabilities/cat.png\"\n",
     "input_image = Image.open(requests.get(img_url, stream=True).raw)\n",
     "display(input_image)\n",
     "\n",
diff --git a/docs/notebooks/image-gemma-4-e2b.ipynb b/docs/notebooks/image-gemma-4-e2b.ipynb
index e578cc3..923e7f2 100644
--- a/docs/notebooks/image-gemma-4-e2b.ipynb
+++ b/docs/notebooks/image-gemma-4-e2b.ipynb
@@ -159,8 +159,8 @@
     "gen_kwargs = dict(generation_config=config)\n",
     "\n",
     "img_urls = [\n",
-    "    \"https://ai.google.dev/gemma/docs/capabilities/vision/images/surprise.png\",\n",
-    "    \"https://ai.google.dev/gemma/docs/capabilities/vision/images/kitchen.jpg\",\n",
+    "    \"https://gemma.ai/assets/capabilities/surprise.png\",\n",
+    "    \"https://gemma.ai/assets/capabilities/kitchen.jpg\",\n",
     "]\n",
     "for img in img_urls:\n",
     "    display(Image.open(requests.get(img, stream=True).raw))\n",
@@ -205,7 +205,7 @@
     "config.max_new_tokens = 512\n",
     "gen_kwargs = dict(generation_config=config)\n",
     "\n",
-    "img_url = \"https://ai.google.dev/gemma/docs/capabilities/vision/images/cat.png\"\n",
+    "img_url = \"https://gemma.ai/assets/capabilities/cat.png\"\n",
     "input_image = Image.open(requests.get(img_url, stream=True).raw)\n",
     "display(input_image)\n",
     "\n",
diff --git a/docs/notebooks/image-gemma-4-e4b.ipynb b/docs/notebooks/image-gemma-4-e4b.ipynb
index 0a1b15b..e9c3ae6 100644
--- a/docs/notebooks/image-gemma-4-e4b.ipynb
+++ b/docs/notebooks/image-gemma-4-e4b.ipynb
@@ -159,8 +159,8 @@
     "gen_kwargs = dict(generation_config=config)\n",
     "\n",
     "img_urls = [\n",
-    "    \"https://ai.google.dev/gemma/docs/capabilities/vision/images/surprise.png\",\n",
-    "    \"https://ai.google.dev/gemma/docs/capabilities/vision/images/kitchen.jpg\",\n",
+    "    \"https://gemma.ai/assets/capabilities/surprise.png\",\n",
+    "    \"https://gemma.ai/assets/capabilities/kitchen.jpg\",\n",
     "]\n",
     "for img in img_urls:\n",
     "    display(Image.open(requests.get(img, stream=True).raw))\n",
@@ -205,7 +205,7 @@
     "config.max_new_tokens = 512\n",
     "gen_kwargs = dict(generation_config=config)\n",
     "\n",
-    "img_url = \"https://ai.google.dev/gemma/docs/capabilities/vision/images/cat.png\"\n",
+    "img_url = \"https://gemma.ai/assets/capabilities/cat.png\"\n",
     "input_image = Image.open(requests.get(img_url, stream=True).raw)\n",
     "display(input_image)\n",
     "\n",
diff --git a/docs/notebooks/run-gemma-with-hf-gemma-4-12b.ipynb b/docs/notebooks/run-gemma-with-hf-gemma-4-12b.ipynb
index d44474c..cb7510b 100644
--- a/docs/notebooks/run-gemma-with-hf-gemma-4-12b.ipynb
+++ b/docs/notebooks/run-gemma-with-hf-gemma-4-12b.ipynb
@@ -230,7 +230,7 @@
     "        \"role\": \"user\",\n",
     "        \"content\": [\n",
     "            {\"type\": \"text\", \"text\": \"Transcribe the following speech segment in its original language. Follow these specific instructions for formatting the answer:\\n* Only output the transcription, with no newlines.\\n* When transcribing numbers, write the digits, i.e. write 1.7 and not one point seven, and write 3 instead of three.\"},\n",
-    "            {\"type\": \"audio\", \"audio\": \"https://ai.google.dev/gemma/docs/audio/roses-are.wav\"},\n",
+    "            {\"type\": \"audio\", \"audio\": \"https://gemma.ai/assets/capabilities/roses-are.wav\"},\n",
     "        ]\n",
     "    }\n",
     "]\n",
diff --git a/docs/notebooks/run-gemma-with-hf-gemma-4-e2b.ipynb b/docs/notebooks/run-gemma-with-hf-gemma-4-e2b.ipynb
index e911f28..c50e1b6 100644
--- a/docs/notebooks/run-gemma-with-hf-gemma-4-e2b.ipynb
+++ b/docs/notebooks/run-gemma-with-hf-gemma-4-e2b.ipynb
@@ -230,7 +230,7 @@
     "        \"role\": \"user\",\n",
     "        \"content\": [\n",
     "            {\"type\": \"text\", \"text\": \"Transcribe the following speech segment in its original language. Follow these specific instructions for formatting the answer:\\n* Only output the transcription, with no newlines.\\n* When transcribing numbers, write the digits, i.e. write 1.7 and not one point seven, and write 3 instead of three.\"},\n",
-    "            {\"type\": \"audio\", \"audio\": \"https://ai.google.dev/gemma/docs/audio/roses-are.wav\"},\n",
+    "            {\"type\": \"audio\", \"audio\": \"https://gemma.ai/assets/capabilities/roses-are.wav\"},\n",
     "        ]\n",
     "    }\n",
     "]\n",
diff --git a/docs/notebooks/run-gemma-with-hf-gemma-4-e4b.ipynb b/docs/notebooks/run-gemma-with-hf-gemma-4-e4b.ipynb
index 6ec1bcd..21a8c37 100644
--- a/docs/notebooks/run-gemma-with-hf-gemma-4-e4b.ipynb
+++ b/docs/notebooks/run-gemma-with-hf-gemma-4-e4b.ipynb
@@ -230,7 +230,7 @@
     "        \"role\": \"user\",\n",
     "        \"content\": [\n",
     "            {\"type\": \"text\", \"text\": \"Transcribe the following speech segment in its original language. Follow these specific instructions for formatting the answer:\\n* Only output the transcription, with no newlines.\\n* When transcribing numbers, write the digits, i.e. write 1.7 and not one point seven, and write 3 instead of three.\"},\n",
-    "            {\"type\": \"audio\", \"audio\": \"https://ai.google.dev/gemma/docs/audio/roses-are.wav\"},\n",
+    "            {\"type\": \"audio\", \"audio\": \"https://gemma.ai/assets/capabilities/roses-are.wav\"},\n",
     "        ]\n",
     "    }\n",
     "]\n",
diff --git a/docs/notebooks/run-gemma-with-keras-gemma-4-26b-a4b.ipynb b/docs/notebooks/run-gemma-with-keras-gemma-4-26b-a4b.ipynb
index 1947409..0ad693d 100644
--- a/docs/notebooks/run-gemma-with-keras-gemma-4-26b-a4b.ipynb
+++ b/docs/notebooks/run-gemma-with-keras-gemma-4-26b-a4b.ipynb
@@ -325,7 +325,7 @@
     "from matplotlib import pyplot as plt\n",
     "\n",
     "image = read_image(\n",
-    "    \"https://ai.google.dev/gemma/docs/images/thali-indian-plate.jpg\"\n",
+    "    \"https://gemma.ai/assets/deployment/thali-indian-plate.jpg\"\n",
     ")\n",
     "plt.imshow(image)\n"
    ]
diff --git a/docs/notebooks/run-gemma-with-keras-gemma-4-31b.ipynb b/docs/notebooks/run-gemma-with-keras-gemma-4-31b.ipynb
index 59b592d..65c4b9c 100644
--- a/docs/notebooks/run-gemma-with-keras-gemma-4-31b.ipynb
+++ b/docs/notebooks/run-gemma-with-keras-gemma-4-31b.ipynb
@@ -325,7 +325,7 @@
     "from matplotlib import pyplot as plt\n",
     "\n",
     "image = read_image(\n",
-    "    \"https://ai.google.dev/gemma/docs/images/thali-indian-plate.jpg\"\n",
+    "    \"https://gemma.ai/assets/deployment/thali-indian-plate.jpg\"\n",
     ")\n",
     "plt.imshow(image)\n"
    ]
diff --git a/docs/notebooks/run-gemma-with-keras-gemma-4-e2b.ipynb b/docs/notebooks/run-gemma-with-keras-gemma-4-e2b.ipynb
index 448098c..b00ed00 100644
--- a/docs/notebooks/run-gemma-with-keras-gemma-4-e2b.ipynb
+++ b/docs/notebooks/run-gemma-with-keras-gemma-4-e2b.ipynb
@@ -325,7 +325,7 @@
     "from matplotlib import pyplot as plt\n",
     "\n",
     "image = read_image(\n",
-    "    \"https://ai.google.dev/gemma/docs/images/thali-indian-plate.jpg\"\n",
+    "    \"https://gemma.ai/assets/deployment/thali-indian-plate.jpg\"\n",
     ")\n",
     "plt.imshow(image)\n"
    ]
diff --git a/docs/notebooks/run-gemma-with-keras-gemma-4-e4b.ipynb b/docs/notebooks/run-gemma-with-keras-gemma-4-e4b.ipynb
index 2cc75bf..1cb8d9a 100644
--- a/docs/notebooks/run-gemma-with-keras-gemma-4-e4b.ipynb
+++ b/docs/notebooks/run-gemma-with-keras-gemma-4-e4b.ipynb
@@ -325,7 +325,7 @@
     "from matplotlib import pyplot as plt\n",
     "\n",
     "image = read_image(\n",
-    "    \"https://ai.google.dev/gemma/docs/images/thali-indian-plate.jpg\"\n",
+    "    \"https://gemma.ai/assets/deployment/thali-indian-plate.jpg\"\n",
     ")\n",
     "plt.imshow(image)\n"
    ]
diff --git a/docs/robots.txt b/docs/robots.txt
new file mode 100644
index 0000000..728c09e
--- /dev/null
+++ b/docs/robots.txt
@@ -0,0 +1,4 @@
+User-agent: *
+Allow: /
+
+Sitemap: https://gemma.ai/sitemap.xml
diff --git a/docs/tutorials/visual-guides/diffusiongemma.md b/docs/tutorials/visual-guides/diffusiongemma.md
index 76f4396..85274c1 100644
--- a/docs/tutorials/visual-guides/diffusiongemma.md
+++ b/docs/tutorials/visual-guides/diffusiongemma.md
@@ -1,6 +1,6 @@
 ---
 title: A Visual Guide to DiffusionGemma
-description: A Visual Guide to DiffusionGemma — a visual walkthrough.
+description: "How DiffusionGemma uses discrete text diffusion, block-autoregressive multi-canvas sampling, and a 26B A4B MoE architecture to generate tokens in parallel."
 models: [diffusiongemma]
 source_url: https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-diffusiongemma
 ---
diff --git a/docs/tutorials/visual-guides/gemma-4-12b.md b/docs/tutorials/visual-guides/gemma-4-12b.md
index 58eff78..2fb0bc9 100644
--- a/docs/tutorials/visual-guides/gemma-4-12b.md
+++ b/docs/tutorials/visual-guides/gemma-4-12b.md
@@ -1,6 +1,6 @@
 ---
 title: A Visual Guide to Gemma 4 12B
-description: A Visual Guide to Gemma 4 12B — a visual walkthrough.
+description: "How Gemma 4 12B achieves fast, unified multimodal understanding of text, images, and audio using an encoder-free architecture with linear projections."
 models: [gemma-4/12b]
 source_url: https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-gemma-4-12b
 ---
diff --git a/docs/tutorials/visual-guides/gemma-4.md b/docs/tutorials/visual-guides/gemma-4.md
index c2a91b0..f20553f 100644
--- a/docs/tutorials/visual-guides/gemma-4.md
+++ b/docs/tutorials/visual-guides/gemma-4.md
@@ -1,6 +1,6 @@
 ---
 title: A Visual Guide to Gemma 4
-description: A Visual Guide to Gemma 4 — a visual walkthrough.
+description: "Illustrated guide to the Gemma 4 architecture: dense and 26B A4B MoE models, Per-Layer Embeddings (E2B/E4B), vision and audio encoders, and KV cache."
 models: [gemma-4]
 source_url: https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-gemma-4
 ---
diff --git a/overrides/main.html b/overrides/main.html
index 916f715..4799cdb 100644
--- a/overrides/main.html
+++ b/overrides/main.html
@@ -18,12 +18,32 @@
 {% extends "base.html" %}
 
 {% block site_meta %}
+  {% include "partials/model-names.html" %}
   <meta charset="utf-8">
   <meta name="viewport" content="width=device-width,initial-scale=1">
   {% if page and page.meta and page.meta.title %}
-    {% set social_title = page.meta.title ~ " - " ~ config.site_name %}
+    {% set raw_title = page.meta.title %}
   {% elif page and page.title and not page.is_homepage %}
-    {% set social_title = (page.title | striptags) ~ " - " ~ config.site_name %}
+    {% set raw_title = page.title | striptags %}
+  {% else %}
+    {% set raw_title = "" %}
+  {% endif %}
+  {% set seo_title = raw_title %}
+  {% if page and not page.is_homepage and raw_title and "gemma" not in (raw_title | lower) %}
+    {% if page.meta and page.meta.models %}
+      {% set first_family = (page.meta.models | first) | split("/") | first %}
+      {% set family_name = gemma_model_names[first_family] | default("") %}
+      {% if family_name and (family_name | lower) not in (raw_title | lower) %}
+        {% if page.url and (page.url | trim("/") | split("/") | first) == "models" %}
+          {% set seo_title = family_name ~ " — " ~ raw_title %}
+        {% else %}
+          {% set seo_title = raw_title ~ " — " ~ family_name %}
+        {% endif %}
+      {% endif %}
+    {% endif %}
+  {% endif %}
+  {% if seo_title %}
+    {% set social_title = seo_title ~ " - " ~ config.site_name %}
   {% else %}
     {% set social_title = config.site_name %}
   {% endif %}
@@ -45,6 +65,11 @@
     {% endif %}
     <meta name="description" content="{{ meta_description | e }}">
   {% endif %}
+  {% if page and page.meta and page.meta.search and page.meta.search.exclude %}
+    <meta name="robots" content="noindex, nofollow">
+  {% else %}
+    <meta name="robots" content="index, follow, max-image-preview:large, max-snippet:-1, max-video-preview:-1">
+  {% endif %}
   {% if page and page.meta and page.meta.author %}
     <meta name="author" content="{{ page.meta.author | e }}">
   {% elif config.site_author %}
@@ -78,7 +103,7 @@
   {% else %}
     {% set social_image_url = social_image_path | url %}
   {% endif %}
-  <meta property="og:type" content="website">
+  <meta property="og:type" content="{{ 'website' if (not page or page.is_homepage) else 'article' }}">
   <meta property="og:site_name" content="{{ config.site_name | e }}">
   <meta property="og:title" content="{{ social_title | e }}">
   {% if social_description %}
@@ -100,4 +125,40 @@
   {% endif %}
   <meta name="twitter:image" content="{{ social_image_url }}">
   <meta name="twitter:image:alt" content="{{ social_title | e }}">
+  {% set site_root = ((config.site_url | trim("/")) ~ "/") if config.site_url else "https://gemma.ai/" %}
+  {% if page and page.is_homepage %}
+    <script type="application/ld+json">{"@context":"https://schema.org","@type":"WebSite","name":{{ config.site_name | tojson }},"alternateName":["Gemma Docs","Google Gemma Documentation","Gemma AI","gemma.ai"],"url":{{ (page.canonical_url or site_root) | tojson }},"description":{{ (meta_description or config.site_description) | tojson }},"publisher":{"@type":"Organization","name":{{ (config.site_author or "Google DeepMind") | tojson }},"url":"https://deepmind.google/models/gemma/"}}</script>
+  {% elif page and page.canonical_url %}
+    <script type="application/ld+json">{"@context":"https://schema.org","@graph":[{"@type":"TechArticle","headline":{{ (seo_title or config.site_name) | tojson }},"description":{{ (meta_description or config.site_description) | tojson }},"url":{{ page.canonical_url | tojson }},"image":{{ social_image_url | tojson }},"author":{"@type":"Organization","name":{{ (config.site_author or "Google DeepMind") | tojson }},"url":"https://deepmind.google/models/gemma/"},"publisher":{"@type":"Organization","name":{{ (config.site_author or "Google DeepMind") | tojson }},"url":"https://deepmind.google/models/gemma/"},"isPartOf":{"@type":"WebSite","name":{{ config.site_name | tojson }},"url":{{ site_root | tojson }}}},{"@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":{{ site_root | tojson }}},{"@type":"ListItem","position":2,"name":{{ (seo_title or config.site_name) | tojson }},"item":{{ page.canonical_url | tojson }}}]}]}</script>
+  {% endif %}
+{% endblock %}
+
+{% block htmltitle %}
+  {% include "partials/model-names.html" %}
+  {% if page and page.meta and page.meta.title %}
+    {% set raw_title = page.meta.title %}
+  {% elif page and page.title and not page.is_homepage %}
+    {% set raw_title = page.title | striptags %}
+  {% else %}
+    {% set raw_title = "" %}
+  {% endif %}
+  {% set seo_title = raw_title %}
+  {% if page and not page.is_homepage and raw_title and "gemma" not in (raw_title | lower) %}
+    {% if page.meta and page.meta.models %}
+      {% set first_family = (page.meta.models | first) | split("/") | first %}
+      {% set family_name = gemma_model_names[first_family] | default("") %}
+      {% if family_name and (family_name | lower) not in (raw_title | lower) %}
+        {% if page.url and (page.url | trim("/") | split("/") | first) == "models" %}
+          {% set seo_title = family_name ~ " — " ~ raw_title %}
+        {% else %}
+          {% set seo_title = raw_title ~ " — " ~ family_name %}
+        {% endif %}
+      {% endif %}
+    {% endif %}
+  {% endif %}
+  {% if seo_title %}
+    <title>{{ (seo_title ~ " - " ~ config.site_name) | e }}</title>
+  {% else %}
+    <title>{{ config.site_name | e }}</title>
+  {% endif %}
 {% endblock %}
diff --git a/scripts/frontmatter_schema.py b/scripts/frontmatter_schema.py
index 8153cf3..43b505b 100644
--- a/scripts/frontmatter_schema.py
+++ b/scripts/frontmatter_schema.py
@@ -188,6 +188,13 @@ def validate(frontmatter: dict, *, page_paths: set[str] | None = None) -> list[P
             )
         if "](" in description or description.startswith("#"):
             problems.append(Problem("description", "description contains markdown syntax"))
+        if "visual walkthrough" in description.lower():
+            problems.append(
+                Problem(
+                    "description",
+                    "description uses placeholder 'visual walkthrough' -- summarize what the page covers",
+                )
+            )
     elif description is not None:
         problems.append(Problem("description", "description must be a string"))
 
diff --git a/scripts/lint_docs.py b/scripts/lint_docs.py
index 4a4a410..aac79f8 100644
--- a/scripts/lint_docs.py
+++ b/scripts/lint_docs.py
@@ -491,13 +491,38 @@ def check_frontmatter(ctx: Context) -> Iterable[Finding]:
         # `<h1 id="__skip">`). On those pages the title is visible content, not just metadata,
         # so it has to say what the page already said. Where the page *has* an H1, the title
         # only reaches the browser tab and llms.txt and may be more specific than the heading.
-        if first_heading(page) is None and "title" not in page.frontmatter:
+        h1 = first_heading(page)
+        title = str(page.frontmatter.get("title", ""))
+        if h1 is None and "title" not in page.frontmatter:
             yield Finding(
                 "frontmatter",
                 page.repo_rel,
                 0,
                 "page has no H1, so `title` is rendered as the heading -- it must be set",
             )
+        if h1 and "gemma" in h1.lower() and "gemma" not in title.lower():
+            yield Finding(
+                "frontmatter",
+                page.repo_rel,
+                0,
+                "frontmatter `title` drops 'Gemma' from the page's H1 -- keep the full title in "
+                "frontmatter and shorten the sidebar label in zensical.toml instead",
+            )
+        if "model-card" in page.rel and "gemma" not in title.lower():
+            yield Finding(
+                "frontmatter",
+                page.repo_rel,
+                0,
+                "model card `title` must include the model name (e.g. 'Gemma 4 model card')",
+            )
+        for lineno, line in outside_code(page):
+            if line.startswith("# ") and ("hidden-title" in line or re.search(r"display\s*:\s*none", line)):
+                yield Finding(
+                    "frontmatter",
+                    page.repo_rel,
+                    lineno,
+                    "H1 is hidden with display:none -- keep H1 visible for search indexing",
+                )
 
 
 HIDDEN_PREFIXES = ("contributing/",)
diff --git a/tests/test_agent_docs.py b/tests/test_agent_docs.py
index 10a0d4f..765f650 100644
--- a/tests/test_agent_docs.py
+++ b/tests/test_agent_docs.py
@@ -168,6 +168,23 @@ def test_variant_generation_ids_accepted_in_frontmatter(self):
             )
             self.assertEqual(problems, [])
 
+    def test_placeholder_visual_walkthrough_description_rejected(self):
+        import frontmatter_schema
+
+        problems = frontmatter_schema.validate(
+            {
+                "title": "A Visual Guide to Gemma 4",
+                "description": "A Visual Guide to Gemma 4 — a visual walkthrough.",
+                "models": ["gemma-4"],
+            }
+        )
+        self.assertTrue(any("visual walkthrough" in p.message for p in problems))
+
+    def test_robots_txt_references_sitemap(self):
+        robots = (agent_docs.DOCS_DIR / "robots.txt").read_text(encoding="utf-8")
+        self.assertIn("User-agent: *", robots)
+        self.assertIn("Sitemap: https://gemma.ai/sitemap.xml", robots)
+
 
 if __name__ == "__main__":
     unittest.main()