Amazon Bedrock Enables One‑Stop SEO: From Keyword Mining to AI‑Generated Content

In today's digital marketing landscape, organic search traffic is a cost‑effective acquisition channel, yet many enterprises struggle with talent shortages, high upfront investment, and long ROI cycles.

The article details a full, low‑cost SEO solution built on Amazon Bedrock’s generative AI services, adhering to core search‑engine principles from keyword mining to AI‑driven content creation.

SEO workflow overview

Build a precise keyword pool

Analyze user search intent

Match relevant products

Generate SEO‑friendly content

Traffic mining with Bedrock

Instead of third‑party keyword tools, the solution extracts keywords directly from product titles using Bedrock’s large model. Example product title:

Ekouaer Pajamas for Women Set Cute Pjs Soft Sleepwear Summer Sleeveless Tops and Shorts Travel Lounge Sets with Pockets

The model returns a JSON list of keywords such as "Pajamas for Women", "Sleepwear", "Cute Pajamas", etc.

{
    "keywords": [
        "Pajamas for Women",
        "Women Pajama Set",
        "Sleepwear",
        "Pjs",
        "Cute Pajamas",
        "Soft Sleepwear",
        "Summer Pajamas",
        "Sleeveless Pajamas",
        "Pajama Tops",
        "Pajama Shorts",
        "Lounge Sets",
        "Pajamas with Pockets",
        "Travel Pajamas",
        "Ekouaer Pajamas"
    ]
}

Calling the search‑engine API with the keyword "Pajamas for Women" returns 2,741 related keywords with search volume and competition data, enabling precise demand targeting.

Search intent analysis

High‑value keywords are filtered and their intent is analyzed using Bedrock’s LLM. For the keyword "pajama sets", the model suggests the topic title:

{
    "title": "Seasonal Comfort: The Ultimate Guide to Women's Pajama Sets for Every Occasion"
}

Product data vectorization

Product information is transformed into embeddings with Amazon Titan Embeddings. Example Python function:

def get_titan_embedding(text):
    """使用Amazon Titan Embeddings模型获取文本嵌入向量"""
    response = bedrock_runtime.invoke_model(
        modelId='amazon.titan-embed-text-v1',
        contentType='application/json',
        accept='application/json',
        body=json.dumps({"inputText": text})
    )
    response_body = json.loads(response['body'].read().decode())
    embedding = response_body['embedding']
    return embedding

All embeddings and product metadata are persisted in Amazon S3, with an ID‑to‑index mapping saved as JSON for later retrieval.

# Store embeddings
embeddings_array = np.array(all_embeddings, dtype=np.float32)
with io.BytesIO() as embeddings_buffer:
    np.save(embeddings_buffer, embeddings_array)
    embeddings_buffer.seek(0)
    s3_client.upload_fileobj(embeddings_buffer, S3_BUCKET, f"{S3_PREFIX}product_embeddings.npy")
# Store metadata
product_data = {"product_info": all_product_info, "id_to_index": id_to_index}
s3_client.put_object(Body=json.dumps(product_data), Bucket=S3_BUCKET,
                     Key=f"{S3_PREFIX}product_data.json")

FAISS similarity search

FAISS builds a flat L2 index for fast nearest‑neighbor queries over the product embeddings.

def build_faiss_index(embeddings):
    """构建FAISS索引"""
    dimension = embeddings.shape[1]
    index = faiss.IndexFlatL2(dimension)
    index.add(embeddings)
    return index

def search_similar_products(query_text, index, product_data, k=5):
    """搜索与查询文本相似的商品"""
    query_embedding = get_titan_embedding(query_text)
    query_embedding_array = np.array([query_embedding], dtype=np.float32)
    distances, indices = index.search(query_embedding_array, k)
    similar_products = [product_data["product_info"][int(idx)] for idx in indices[0]]
    return similar_products, distances[0]

AI‑driven content creation

Combining the topic, matched products, and search intent, Bedrock’s LLM generates a structured SEO article in JSON format, which is later rendered by a React component.

You are an expert SEO content writer. Create a concise, SEO‑optimized article based on the following inputs:
- Main topic: {topic}
- Primary keyword: {primary_keyword}
- Related products: {products_list}
...

Serverless, event‑driven architecture

Amazon EventBridge triggers Lambda functions for keyword mining, product extraction, and content generation. Data flows through SQS queues, Amazon Batch handles large‑scale parallel processing, and results are persisted in S3 and DynamoDB. This design provides zero‑ops, on‑demand scaling and cost efficiency.

Benefits and outlook

High efficiency: automated pipeline dramatically reduces manual workload.

Low cost: far cheaper than traditional SEO teams.

Data‑driven: real search data guides content creation.

Quality assurance: AI‑generated content follows SEO best practices while remaining readable.

Scalability: serverless architecture supports massive content volumes.

Future enhancements may incorporate user‑behavior analytics and competitor analysis to further refine the SEO ecosystem and sustain organic traffic growth.