🤯 I Was Drowning in Google Searches—Then I Built an AI Search Engine

Introduction

I used to spend hours second-guessing myself whenever I searched for answers. I’d worry about missing out on something important or scrolling past the one link that might help me. That fear always left me feeling small and insecure, like I was fighting the whole internet on my own.

But then I found a different way: building an AI search engine right inside Slack. It feels like having a quiet friend who’s always ready to help—no judgment, no fuss. With this AI search engine, I can ask a question in Slack and get a straightforward answer pulled from the web in real time.

It doesn’t solve all my insecurities, but it helps. Instead of feeling overwhelmed, I finally have a system that does most of the searching for me. And I want to share that here, in case it makes someone else’s life a bit easier too.

I. Prerequisites for Building an AI Search Engine in Slack

I never thought I’d be the kind of person who could build anything tech-related. There’s always that voice in my head saying, “You can’t do this; you’re not cut out for it.” But sometimes, having the right support system—tools, really—makes all the difference. When it comes to setting up an AI search engine in Slack, these are the essentials I lean on:

• Slack Channel: A safe space, like #perplexity, where I can drop my questions without feeling judged.

• Zapier (Paid): This is the friend that does the heavy lifting for me. I rely on it to handle all the background tasks I can’t manage on my own.

• Brave Search API: It’s the engine behind my searches. I know it has a limit (2,000 free searches a month), but I trust it to get me what I need before my nerves kick in.

• PhantomJSCloud: This one parses the content from each site. I’m thankful it doesn’t ask too much of me, with 500 free parses a day—enough for me to try, fail, and try again.

• OpenAI Platform: This is where the summarizing happens. It must has at least $10 of credit.

These tools act like a little cheer squad—each one stepping in at the right moment to make sure I can actually build my AI search engine without second-guessing myself too much.

II. Create a Slack Channel for Your AI Search Engine

I used to drop my questions anywhere in Slack and end up losing them. That got frustrating. If you’re building an AI search engine, it’s better to have one channel for all your queries and updates.

Here’s how:

1. Open Slack.

2. Click “Add channels,” then select “Create a new channel.”

   

3. Name it something like #perplexity or #ai-search-engine.

   

This way, you keep everything in one place. No more digging through random channels or worrying if you’re spamming the wrong thread. A single channel keeps your AI search engine work organized and easier to manage.

III. Set Up the Zap (Overview for Your AI Search Engine)

Here's the big picture:

1. Trigger
   Every new message in your Slack channel sets things in motion. You don’t have to do anything extra—just say what you need, and the process starts.

2. Code Step
   Your query goes to the Brave Search API, which grabs the top results. Then PhantomJSCloud fetches the site content, so you get the actual text to work with.

3. AI Summarization
   OpenAI combines everything into a short, clear answer. It doesn’t matter if the pages have extra fluff; the AI focuses on what’s important.

4. Slack Response
   Finally, the answer, along with source links, goes back to Slack. You’ll see it threaded under your original query, keeping everything neat.

Having this system makes me feel less insecure about messing up. Each step is like a little support beam, holding up the rest of my AI search engine. I can trust that when I drop a question in Slack, the process will take care of the details for me.

IV. Step-by-Step Zap Configuration for Your AI Search Engine (Detailed)

Step 1: Configure the Slack Trigger

1. App & Event: In Zapier, pick Slack as the trigger app. Choose “New Message Posted to Channel.”

   

2. Select Channel: Choose #ai-search-engine (or whatever you named it).

3. Ignore Bot Messages: Set “Trigger for Bot Messages?” to “No.” This prevents an endless cycle of AI messages triggering the zap again.

   

4. Example Query: Go to Slack, type a question like “What’s new in AI search engine technology?” You’ll need this to test your zap.

   

5. Test: Click “Test Trigger” in Zapier. You’ll see that Slack message pop up. If you don’t see it, try sending the message again.

   

I remember being worried that I’d pick the wrong channel or accidentally trigger the zap with my own replies. But ignoring bot messages keeps things safe.

Step 2: Add the Code (Python) Step

1. Choose ‘Code by Zapier’: In your zap, add a new action and pick “Run Python.”

   

2. Input: Create one input field called searchterm. Map it to the Slack text from the previous step.

3. Write Your Code:

   I use ChatGPT o1 to create the code follow this prompt:

Create a Python code step to use in Zapier
The code step should do the following:

- Take a single input value in the / query = input_data['searchterm'] / format.
- Search the Brave Search API and retrieve the URLs for the top 3 web results.
- Use the PhantomJSCloud API to parse the content from these 3 URLs in plain text.
- Return a list of the URLs and their associated plain text content.
- Use placeholders for both API keys and I'll add them in the live code.
- Use techniques like threading to make the code run as quick as possible.

• Brave Search: It looks up to 3 top URLs based on searchterm.

• PhantomJSCloud: It opens each URL and extracts plain text.

• Data Returned: A list of URLs plus one combined text block.

  

1. API Keys: Insert your Brave and PhantomJSCloud keys into the code. Don’t share them publicly.

2. Test: Run it. Wait a few seconds. Zapier will show you the parsed text in the “Data Out” section.

I used to think coding was too scary. But copying a template and adding my keys was enough to get me started. If something breaks, Zapier will show an error so you can fix it.

Step 3: Summarize with OpenAI

1. App: Add a new action for “ChatGPT” (part of Zapier’s OpenAI integration).

2. Model: Pick a smaller model like gpt-4o-mini so you don’t burn through credits too fast.

3. User Message:

Query: {{Slack Message}}
Research: {{Parsed Text from Python Code}}
Answer:

This is how you tell the AI what you asked for and what data it has to work with.

4. Assistant Instructions: Ask it to summarize briefly, skip markdown, and keep it clear.

You are a helpful assistant.

Using the query from the user and the raw research provided, summarise an answer.

Don't use markdown in the response.

5. Max Tokens: Around 1,000 is enough. More can work, but I try not to overshoot.

Sometimes the AI might add extra formatting. Remind it to keep things plain. It’s not being rude; it’s just following instructions.

Step 4: Format the Source URLs

1. App: “Formatter by Zapier.”

   

2. Transform: “Line-item to Text.”

3. Input: Select the “list of URLs” from your Python step.

4. Separator: Type [:newline:]. This forces each URL onto its own line.

   

5. Test: Now they’re neatly separated in the “Data Out.”

   

I like seeing each link on its own line. It looks cleaner when Slack shows the sources.

Step 5: Send Final Answer to Slack

1. App & Event: Choose “Slack” and “Send Channel Message.”

2. Channel: Same one, #perplexity.

3. Message Text:

Answer:
{{ChatGPT Summary}}

Sources:
{{Formatted URLs}}

I like to label them clearly so I know which part is the summary and which part is the source.

5. Send as a bot?: Yes. That way, the bot’s reply won’t trigger the zap again.

   

6. Bot Name: Something short like “Perplexity” or “AI Search Engine Bot.”

   

7. Thread: Map the “Ts” (timestamp) from the Slack trigger. This keeps the answer under your original question.

8. Test: Send a query in Slack, wait 10 seconds, and watch the answer thread appear.

It might feel like a lot, but each step acts like a little safety net. By the time you finish, you’ll have an AI search engine that listens when you talk to it in Slack, searches the web, and returns a concise summary. That sense of support can be really comforting when you’re not sure where else to start.

V. Test and Review for Your AI Search Engine

It’s important to confirm that everything runs smoothly, especially after all the setup work. Here’s what to do:

1. Post a Query
   Go to your #perplexity channel and type a question, such as “What’s new in AI search engine technology right now?”

2. Wait a Few Seconds
   Give Zapier about 10 seconds to process the request. You should see an AI-generated reply in a Slack thread under your question.

   

3. Check the Answer
   Look for a concise summary and a list of sources. If anything seems off—like missing info or strange formatting—there could be an issue with one of the steps.

4. Adjust as Needed
   If the content is incomplete, try limiting the URLs Brave fetches or modify your AI model settings. For slow responses, increase any timeouts set in Zapier.

5. Repeat and Refine
   Test various queries. Each one helps you see where the AI search engine does well and where it might need a little more attention.

Conclusion

I used to feel lost whenever I searched for answers. Too many links, too much noise, and always that nagging fear of missing the one thing that might actually help. It made me second-guess myself, over and over, until I felt small and incapable.

But now, my AI search engine in Slack is my quiet support system. I don’t have to fight the internet on my own anymore. I ask, it listens, and in a few seconds, I get exactly what I need—clean, simple, and straight to the point. No endless scrolling. No self-doubt. Just a response waiting for me, like a steady friend who’s always there.

It wasn’t easy setting this up. There were moments when I thought, This is too much. I can’t do this. But step by step, I built something I could rely on. The Slack channel keeps everything organized. The Zapier automation handles the messy parts I don’t understand. Brave Search and PhantomJSCloud dig through the web so I don’t have to. And OpenAI summarizes it all, saving me from overthinking.

Now, when I drop a question in Slack, I don’t feel overwhelmed. I feel a little less alone. I know my AI search engine will do the heavy lifting, so I don’t have to carry all the stress by myself. And if you ever feel the same way—drowning in information, unsure where to look—I hope this system helps you too.

Because sometimes, you don’t need the whole internet. You just need something that listens and brings you back exactly what matters.

If you are interested in other topics and how AI is transforming different aspects of our lives, or even in making money using AI with more detailed, step-by-step guidance, you can find our other articles here:

• 20 AI Startups Raise Over $460M (Jan 18, 2024 - Jan 24, 2025)*

• Deepseek-R1 Update: The Free AI Model That’s Redefining Open Source*

• How to Use ChatGPT Scheduled Tasks to Earn $1,000 a Week

• How to Build an Automated Financial Assistant Using AI: A Step-by-Step Guide*

• How to Build Your First AI Employee Using Lindy AI

*indicates a premium content, if any