🦾How You Can Leverage:

Is your ChatGPT output actually …. Wrong? 

Hallucinations everywhere? 

Or giving you old info. 

(Prolly.) 

We’ve taught thousands of business professionals the basics of prompt engineering and ChatGPT. 

And a common problem we see over and over? 

The misuse of Browse with Bing. 

Either peeps aren’t using it, when they should! 

Or.....

They’re not using it correctly and missing out on what we think is one of the most underutilized features of any LLM. 

Don’t worry y’all. 

We tackled this today and showed you the correct way to connect ChatGPT to the internet. 

Apparently, this show was a banger. 

The key to getting better outputs out of ChatGPT? 

You need to start with higher quality, fresher and more relevant data than what models provide by default. 

Trust us. This will completely change your outputs. 

To make it easy, we broke today’s show down into 5 steps on how to do this the right way. 

Let’s delve shorties. 

(Lolz)

ChatGPT isn't just a fancy chatbot – it's got a secret weapon called "Browse with Bing".

TBH, it used to really stink. 

But OpenAI updated this bad boy and now it’s a legit answers engine and Perplexity contender. 

But here's the kicker: Browse with Bing is sometimes like a shy superhero that sometimes needs coaxing to show off its powers.

Let’s be honest — your default ChatGPT queries are working with data that's AT BEST from October 2023. 

That's almost a year old, y’all! So, you’ve gotta dive into the Bing deep end. 

Try this: 

Start your prompts with "Using Browse with Bing, tell me about...".

But don't just fire and forget! 

Keep your eyes peeled for that little Bing icon between your query and ChatGPT's response. If you don't see it, ChatGPT might be trying to pull a fast one on you with its outdated knowledge.

Pro tip: Sometimes ChatGPT will use Browse with Bing on its own, but don't count on it. Be the puppet master and MAKE IT HAPPEN.

Train your eyes to always check if Bing was used, even when you didn't explicitly ask for it.

You wouldn't trust a weather forecast from last year, so why settle for ancient AI info?

It's time to turn ChatGPT into your personal time machine.

But like instead of the typical Time Machine where you go back in time, this Time Machine brings ChatGPT up to…. Current day. Lolz. 

The problem: Even when you use Browse with Bing, ChatGPT might still pull outdated info.

We've seen it query things like "LLMs recent developments 2023" when we're well into 2024 and ask it for recent developments.

That's a recipe for stale data, and even worse outputs. 

Yuck. 

Try this: 

Get specific with your time travel. 

Instead of vague "recent" requests, hit ChatGPT with the one-two punch: 

"Using Browse with Bing, tell me about [YOUR TOPIC] from [CURRENT MONTH] [CURRENT YEAR] to [NEXT MONTH] [CURRENT YEAR]". 

BAM! You've just teleported your AI to the present (or even the near future).

Example: "Using Browse with Bing, tell me recent developments about large language models from August 2024 to September 2024 only."

Warning: SEO wizards love to rehash old content with new dates. Don't let ChatGPT fall for their tricks.

By specifying a tight date range, you're forcing it to dig for the freshest intel.

Forget standalone AI search tools. 

We're about to turn ChatGPT into a research beast that would make Google blush.

Here’s our secret sauce: It's all about targeted Browse with Bing calls.

We're talking precision strikes on the info you need.

Try this: 

Here’s the magic formula -

"Using Browse with Bing, search 8 or more different sources about [YOUR TOPIC], making sure those sources are from different domains. Please go step by step, researching deeply, taking your time, and report back only once you've researched the topic thoroughly."

Why it works: We've done the legwork and found that 8 is the Goldilocks number.

Not too few, not too many.

Just right for comprehensive, varied results. Plus, by specifying different domains, you're avoiding the echo chamber effect.

Pro tip: Add "Please take your time" to your prompts. Some studies have shown this can improve ChatGPT's performance by up to 40% on complex tasks like solving math problems. 

It's like giving your AI a deep breath before the deep dive.

Worst case, you’re nice to AI models before they take over? Hahaha. 

Not all sources are created equal. 

You wouldn't trust your health to a random blog, so why let ChatGPT serve you info from Questionable Quotes Quarterly?

Even with targeted Browse with Bing calls, you might end up with a bunch of listicles and fluff pieces. 

Not a fan? 

Samesies. 

That's not gonna cut it for serious research or business decisions.

Try this:

Flex your authority muscle by adding this to your prompts: 

"Please only use authoritative sources for information such as [LIST SPECIFIC REPUTABLE SOURCES IN YOUR FIELD]."

Example: "Please only use authoritative sources such as arXiv, Microsoft, OpenAI, Google DeepMind, AI Kaggle, and other peer-reviewed or industry-leading publications."

Why it matters: This turns ChatGPT from a trivia night contestant to a panel of Nobel laureates.

You're getting the cream of the crop, not just whatever ranked highest in SEO. 

This trick works wonders when you need to impress the boss or shut down that know-it-all coworker. You're welcome.

Another great LLM for research? Perplexity.

We gave you the rundown on how to use it here.

Don’t wanna leave your starting point up to Bing? 

We get it.

Sometimes, a well-rounded Answers Engine-esque approach is exactly what you need. 

Sometimes, it’s the worst possible way to get started and you need precision control. 

(Pew pew laser been sounds pew pew) 

Sometimes, you need to dive deep into specific articles.

But who has time to read six different tabs? Not you.

The problem: Dropping URLs directly into ChatGPT doesn't work like you think it does. 

It doesn't actually visit the page; Browse with Bing just searches for keywords in the URL.

That's a recipe for hallucinations and misinterpretations.

The solution: Level up to an internet-connected GPT like WebPilot, VoxScript, or WebReader.

Try this:

1. Install an internet-connected GPT from the GPT store (it's free, even with a basic account).

2. Use an @mention to call the GPT (just hit @ and start typing its name).

3. Feed it up to 6 URLs at once. (That’s kinda the Goldilocks number we’ve found) 

4. Ask it to summarize or analyze the content.

Example prompt: "@WebPilot, please analyze and bullet point the main ideas from these articles: [URL1] [URL2] [URL3]"

 This method actually visits and reads the full content of each URL. 

You can analyze multiple in-depth articles or reports simultaneously, creating a solid foundation of knowledge before diving into your main task.

For real, you haven’t taken our free and live Prompt Engineering course? 

Change that! 

Reply PPP to this email, and we’ll send you the hidden registration page. 

Refer Friends to Win a Year of ChatGPT!

(Or your fave LLM like Claude, Gemini, Copilot, etc)