JEFF SU: After analyzing millions of inputs, Google discovered that the most successful prompts are 21 words long on average, yet most of us are inputting fewer than nine words in ChatGPT and Google Gemini. Obviously, as most guys will tell you, length is not everything. It's really about the effort. So in this video, I'm going to share the top-five actionable tips from Google's latest guide to writing better prompts. Let's get started. 

Tip number one, the power of three. One of the least-used features in ChatGPT is the regenerate function. Similarly, very few of us actively choose to show other drafts in Google Gemini, even though seeing different options can actually inspire new ideas and directions we may not have initially considered. 

It's not that surprising, though. We input a prompt, and we expect the chatbot to give us what we're looking for. Who has the time to go through different variations? 

To get the best of both worlds, Google recommends asking for three variations in the initial prompt. For example, it could be something simple like generate three options for a new slogan emphasizing reliability, innovation, and a long history of popularity for a company. Or I'm a marketing manager at Spotify. I need to A/B test new messaging. Here is our current messaging. Generate three different variations of ad copy. 

This works for more complex prompts as well. For example, I'm writing an email to job candidates who just finished the interview process. Create an offer-letter template for the selected candidate, for the position, with a request to schedule a call to discuss benefits, compensation, and start date. Give me three variations. 

Or draft three email templates to check in on my retail customer who just started to spend on Google ads. Use one value proposition-- cost, ease of use, security-- as the main topic for each email, and let them know they can reach out to me with questions. 

In this last example, you might think the customer cares about cost when, in reality, they prioritize security. And if you didn't ask for three different emails on three different topics, you might have missed the opportunity to engage the client. 

Try this out for yourself. It was a game changer, for me and I basically have "give me three variations" saved in my text-expander app. Drop a like if you found this tip helpful, but feel free to click the dislike button twice if you didn't. 

Tip number two, multistep workflows. One of the simplest prompting techniques is called chain-of-thought prompting. Basically when you break down a large task into smaller sequential prompts, you actually get better results by reducing potential errors. For example, instead of asking ChatGPT to generate an entire cover letter, which will end up being super generic, you ask it to first write the hook based on the job description. Then write the body paragraph based on the hook, so on and so forth. 

The way Google recommends we think about this is instead of trying to write one super-long prompt to solve a super-complex issue, can we break the problem down into mini milestones that can each be tackled by a separate prompt? 

For example, if your end goal is to create a presentation to secure additional budget for your project, here are three mini milestones and their corresponding prompts. Step one, outline three key benefits of my project for target audience, and quantify the potential impact for target audience. Step two, identify potential objections target audience may have, and generate counterarguments to address them. Step three, recommend a visual format-- charts, graphs, et cetera-- to best present the financial data in a clear and impactful way. 

I'll be very honest, I used to be a one-shot wonder. Get it, one-shot prompting and one-shot drinking? But now whenever I catch myself trying to type an essay in ChatGPT, I try to think of ways to break it down into smaller, bite-sized pieces. By the way, I've included a link to Google's full guide down below, and if you use Google Workspace at work, you definitely want to grab my free Workspace toolkit as well. 

Tip number three, template time savers. The idea here is super simple. First, we ask ChatGPT or Google Gemini to brainstorm template ideas based on our role and responsibility. Then we generate those templates. 

Diving right into an example, "I'm a project manager responsible for our next product launch event called Apple Unfolded scheduled for September 30. I'm working with three teams, the hardware team, the software team, and the retail team. Your task is to generate five ideas for communication templates I can create now that will save me time over the course of the project." 

ChatGPT then outputs five extremely specific ideas. Some of these I would not have thought of myself. I think this bug-report template could be very useful. And then I follow up with, "Draft a bug report template for the software team to use," and ChatGPT goes ahead and creates that for me as well. 

Some of you might have noticed this is also an example of a multi-step workflow. We first asked ChatGPT to brainstorm ideas. Then we expanded on one of those generated ideas. 

Another example might be, "I'm a product marketing manager on the Spotify ads team. I've been assigned to mentor and onboard a new hire with no marketing experience. Your task is to brainstorm five templates I can use to streamline his onboarding process." And then step two, draft one of those templates. 

What I like most about this tip is how it helps me overcome mental blocks by suggesting a variety of templates I wouldn't have thought of myself, and usually at least one of these templates proved to be super useful. 

Pro tip-- if you already have access to an existing template, you can improve upon it by providing additional relevant details. For example, "I will share a job description template HR sent me. Suggest ways to update the job description based on my requirements." And you input your requirements. Here's the job description, and then you paste the job description. 

I almost forgot. I'll link all the prompts I go over in this video down below so you can try them out yourself. 

Moving on to tip number four, top-down competitive analysis. When conducting research, we can think of the process like an upside-down triangle where we start off with broad concepts to understand the overall landscape, and as we move down the funnel, we use narrower prompts to uncover actionable insights. For example, we start off with something like, "Provide a comprehensive overview of the current drone delivery industry, including major players, regulations, and key market trends." Then, "I'm conducting a competitive landscape analysis on the drone delivery industry. Provide the strengths and weaknesses of the major players we just identified, including their specific strategies, tactics, target audiences, and results." 

Now that we have a high-level overview, let's get a little personal. "I'm a product marketing manager in the Wing division of Alphabet. Identify actionable insights and recommendations for how my company can improve its approach and gain a competitive advantage." 

And finally, let's get super tactical. "Brainstorm, three innovative marketing campaigns and corresponding messaging that lean into our strengths while attacking the weaknesses of our competitors. List the strengths and weaknesses referenced for each message." 

As you can see, starting with broader prompts help both the chatbot and ourselves understand the topic at a very high level. Then by asking progressively more specific questions, we guide the chatbot to give us the exact information we need. 

Pro tip-- Google is not paying me to say this, although they probably should, but I found that Google Gemini is much better at summarizing web pages than ChatGPT when conducting research. For example, I throw in a few URLs, one of which is a YouTube video on drone delivery, and I use the prompt, "Summarize these articles. Provide key insights, and contextualize why these announcements are important." And Gemini provides a very balanced summary that draws from all the links, whereas ChatGPT only pulled information from one of the five links that I shared. 

Speaking of Gemini, though, tip number five is only applicable for Google users, and that's to use Google Docs to supercharge your PDFs. Right now, there's no easy way for Gemini users to upload PDFs directly. But if you upload a PDF onto Google Drive, right click on the PDF to open with Google Docs, you can actually bring up Gemini within Google Docs by typing at and Help me write and then inserting your prompt here. 

For example, this is a McKinsey report on package delivery, and I can prompt Gemini in Google Docs with, "Summarize the key findings and implications of this report. Focus on the different strategies Chinese companies and Western companies are employing when it comes to last-mile delivery." 

The output is actually pretty good, and I can even take this a step further and click Refine, Custom. "Create a presentation outline with five slides. Each slide should have one headline and three supporting bullet points." And Gemini for Google Docs gives me exactly that, which is pretty cool. 

Unfortunately, this Help me write feature is only available for users on the Google One AI Premium plan, but I guess that makes sense because you also have to be on the paid ChatGPT plan to upload PDFs onto ChatGPT. 

If you enjoyed these tips, you definitely want to check out how to use ChatGPT to create amazing presentations. See you on the next video. In the meantime, have a great one.