Turn Bing News into an AI Briefing with Claude and SerpApi

Most news apps and search engines already present headlines well, so simply listing them is becoming less valuable. The real value has shifted to connecting the dots: grouping related news stories, identifying patterns across sources, and explaining what actually matters and why.

In this post, we’ll build a simple pipeline that turns raw Bing News results into a structured daily briefing which does exactly that using:

• SerpApi's Bing News API for real-time news data
• Claude for summarization and insight generation

Setting Up

LLMs work best when you give them clean, formatted context instead of raw scraped HTML, so we'll do exactly that.

Get results using SerpApi (Bing News API) → Normalize → Send to Claude with a relevant prompt → Get structured briefing

For this tutorial, we're going to use SerpApi's new Python library to get results from search.

1. Install SerpApi's new Python serpapi library, the python-dotenv library (for using environment variables) and anthropic library (for Claude) in your environment:

pip install serpapi python-dotenv anthropic

serpapi is our new Python library. You can use this library to scrape search results from any of SerpApi's APIs.

2. To begin scraping data, create a free account on serpapi.com. You'll receive 250 free search credits each month to explore the API. Get your SerpApi API key from this page.
3. [Optional but Recommended] Set your API keys in an environment variable, instead of directly pasting them in the code. Refer here to understand more about using environment variables.

SERPAPI_API_KEY=<YOUR PRIVATE SERPAPI API KEY>
CLAUDE_API_KEY=<YOUR PRIVATE CLAUDE API KEY>

Get your SerpApi API key here: https://serpapi.com/manage-api-key
Get your Claude API key here: https://platform.claude.com/settings/keys

4. Import classes needed for this project at the beginning of your code file and set up some constants we'll use later on:

import os
import serpapi
import json
import anthropic
from dotenv import load_dotenv
load_dotenv()

SERPAPI_API_KEY = os.environ["SERPAPI_API_KEY"]
CLAUDE_API_KEY = os.environ["CLAUDE_API_KEY"]

Fetch News with Bing News API

We’ll write a function to use SerpApi’s Bing News API (bing_news engine) and get all the news results.

def fetch_news(query):
    client = serpapi.Client(api_key=SERPAPI_API_KEY)
    results = client.search({
        "engine": "bing_news",
        "q": query,
        "device": "desktop",
        "mkt": "en-au"
    })
    return results.get("organic_results", [])

We can call this function with any query such as "Self driving software" or "AI" and scrape results from Bing News.

You can test the Bing News API out in our playground with parameters of your choice:

SerpApi Playground - SerpApi

Test SerpApi’s Google Search, Google Maps, YouTube, Bing, Walmart, eBay, Baidu, Yandex and more APIs for free in the interactive playground!

SerpApiSerpApi

Prompting Claude

Good prompts can make all the difference when building a AI briefing based on the results.

A bad prompt would look like:

“Summarize these articles”

Better prompts often define structure, force grouping, and require reasoning.

def build_prompt(articles):
    return f"""
You are a technical news analyst.

Task: Convert raw news articles into a structured daily briefing.

Requirements:
- Group related articles into 3-5 themes
- Each theme must have:
  - A short title
  - A 2-3 sentence summary
  - A "Why it matters" explanation
- Be concise and non-generic
- Avoid repeating the same point across themes

Articles:
{articles}
"""

Generate the Briefing (Claude API)

Using Claude's API, we can turn the data we got from SerpApi into a briefing. Here is some simple code to use Anthropic API -

def generate_briefing(prompt):
    client = anthropic.Anthropic(api_key=CLAUDE_API_KEY)
    response = client.messages.create(
        model="claude-opus-4-7",
        max_tokens=1500,
        messages=[{"role": "user", "content": prompt}]
    )
    return response.content[0].text

Putting It Together

Now that we have all the pieces we'll use, let's write the code to call the different functions and get our news briefing:

def run():
    query = "AI"
    articles = fetch_news(query)
    prompt = build_prompt(articles)
    briefing = generate_briefing(prompt)
    print(briefing)

if __name__ == "__main__":
    run()

Now, we're ready to run this file. We can do so by typing python <filename> in the terminal.

Automated Delivery

A briefing is only useful if it reaches you regularly automatically. Add a delivery layer like automated daily email or a slack webhook.

Automated Daily Email

• Simple automation run daily via cron or other scheduling methods
• Email formatted briefing to yourself or a team

Slack Webhook

• Post structured sections into a channel
• Good for teams or shared awareness feeds
• Enables discussion around themes

Read this blog post for more info on setting up a Slack webhook (The use case in the post is different, but the process remains the same). It also covers how to schedule a job every day using the Python schedule library:

Build a Slack Bot That Alerts You When a New Company Ranks for Keywords

Let’s build a system that detects new companies ranking for a keyword and sends a Slack alert.

SerpApiSonika Arora

Generated AI Briefing Example Output

Here's an example run. In my SerpApi call, I set the query to "AI" to keep things simple. I used these parameters for Anthropic API:

model="claude-opus-4-7",
max_tokens=1500,
messages=[{"role": "user", "content": prompt}]

Here's the Claude generated briefing:

Building On Top Of This

Changing The Model

As of May 2026, here are the options Claude provides and what each is best for:

• Opus (Most Intelligent): Best for complex tasks, deep reasoning, coding, and strategic analysis. It is the most powerful but slower and more expensive.
• Sonnet (Balanced): It provides the best balance of speed, intelligence, and cost, making it ideal for daily coding, data analysis, and content generation.
• Haiku (Fastest): Optimized for speed and cost-efficiency. It is perfect for high-volume tasks, quick responses, and simple, repetitive jobs.

You can test the script with different models and figure out the best one for your use case. I used Claude Opus 4.7 for this demo.

Persisting History

Instead of generating standalone briefings, store each run with a timestamp. This turns your system into a time-series of news summaries. Once you have history, you can:

• Compare today vs yesterday (what’s new, what disappeared)
• Track theme momentum (what’s increasing or fading)
• Ask Claude for delta summaries instead of fresh-only summaries

Example prompt changes:

“Compare today’s briefing with yesterday’s. Highlight new, removed, and shifting themes.”

Conclusion

The key takeaway here is that LLMs work best when they’re grounded in high-quality external data. In this project, SerpApi provides structured, real-time Bing News data, while Claude turns that raw information into something readable and useful: a concise daily briefing with themes, summaries, and context.

Instead of asking a model to know everything, we can give it fresh information and let it focus on synthesis and reasoning.

I hope you found this useful. If you have any questions, feel free to reach out at contact@serpapi.com

Relevant Links

• SerpApi Bing News API Documentation
• Claude Python SDK

Related Blog Posts

Connecting Claude AI to the Internet (using Function Calling)

Learn how to connect Claude AI by Anthropic to Internet to get rid of knowledge cutoff and enable access to custom functions using function calling or tool use.

SerpApiHilman Ramadhan

How to Extract Bing News Data with SerpApi and Python

This blog post is a step-by-step tutorial for scraping Bing News using SerpApi and Python.

SerpApiArtur Chukhrai

The Web Search API for AI Apps, Agents, and LLMs in 2026

Build AI apps without scraping headaches. Learn how a Web Search API extends model knowledge, delivers real-time data, and gives you full control and flexibility.

SerpApiHilman Ramadhan