 Hello and welcome to another edition of Whiteboard Friday. My name is Elisa Sharf. I'm the director of SEO here at Sierra Interactive and today I'm gonna be your host. So my talk today is about harnessing generative AI for content and SEO success. So AI is here. How do we feel about it? We're talking to a lot of people at Sierra and people are feeling different ways about it. Some people are feeling really skeptical about it and they're really worried that the outputs from generative AI aren't going to comply with EEAT principles. That's experience, expertise, authority and trust. So how can a machine capture all of those principles? We're also talking to some people who are kind of annoyed or dismissive about the tool and they say it's a plagiarism engine. It has no role in good quality content production. But at Sierra we're feeling the love, we're really excited about it and what we're gonna talk about today is a framework that offers a way to leverage generative AI in a way that's safe but also effective. So let's just plan ourselves in the task at hand. Say we want to create a content brief and for the sake of the example today we're gonna be talking about GPT-4 and so there's two different ways that we can do this. One way is to say look, GPT-4 is really, really smart. It can pass the bar, it can earn an MBA. So surely it's smart enough to do some SEO and content research. So you can just go to GPT-4 and say here's my keywords, write me a content outline, copy and paste the output, give it to somebody to review, deliver, publish to your website. So you can certainly do that but if you're going to rely totally on the vast knowledge of GPT-4 and say, hey, build me a content brief, there's gonna be some issues that you might find with that process. So one of them is gonna be that you might have low confidence in the quality of your recommendations relative to the SERPs that you're targeting. So how exactly does GPT-4 know what's present and what elements Google is rewarding on those SERPs? So that's gonna be an issue. You're also potentially gonna have low confidence in the uniqueness of those recommendations. So if you can follow this very simple process, what's stopping your competitors from doing the exact same thing and coming up with a very similar output? And lastly, we get back to the EEAT principles. So what exactly about this process is going to demonstrate experience and expertise? It's going to be a very fast process but is it going to be good? Hard to say. So what we're gonna talk about today is the process and the framework that we've been using at SERP and that framework is called PARS. It stands for Populate, Activate, Retrieve and Scrutinize. So I'm gonna take you step by step through each level of this framework. One of the things we really like about this framework is that it's going to blend your data, your experience, your insights with the vast knowledge of all of those large language models. It's going to allow you to combine the power of those models with the existing and unique knowledge and their ability to ingest and contextualize new information. And lastly, it's going to come with a very robust review period that should make your compliance team very happy. So let's get into it step by step. So the first step is Populate. And let's take a step back. Imagine you were starting at SERP as a new hire. I wouldn't say go to the library and learn everything you can about everything. That person would come back a very intelligent person but they still wouldn't know our process and our approach and what our clients are expecting. So instead, what I would do is we would put them through our training program, share with them our training material. We would share with them our templates, our best in class examples so they can contextualize themselves with what it is we deliver to clients. So let's start by doing that with GPT. We can import all of that information directly to GPT-4 so it really understands the task at hand. Next, we wanna make sure that the output we're building is as high quality as it can be. And so we're not just going to give it keywords. We're going to designate, here's the primary keywords, here's the secondary keywords, here's data we've pulled from the SERPs to say here's the features that are present so we wanna make sure we're optimizing for them. Here's people also ask questions so we understand what the audience is really looking for with this content and here's related keywords. And we're not going to stop there either. We also wanna make sure that we're integrating the subject matter experts into the content process. So how would we do that without GPT? I would have my team talk to people like sales representatives, customer service representatives, or directly with our clients' audiences to better understand their pain points, their motivations and their triggers. And so all of the outputs of those conversations can also be ingested directly into GPT. Lastly, we wanna make sure that the content we're creating aligns with the brand guidelines, the brand vision and the brand tone. That's really important. That's the first step. If you were to just stop there, you're gonna have a pretty good output, but there's three more steps. The second is activate. And this is basically what people are referring to when they talk about prompt engineering. So there's a ton of information out there about prompt engineering. Mike King has a great resource that he's built for the community that you should definitely check out when it comes to prompt engineering. So I'm just going to focus on the high level steps that we're taking here at Sierra. The first thing is you wanna make sure that you are diagnosing the problem that GPT needs to solve for you. The second step is to break down that problem into functional elements. The third step is to reframe that problem. So say it a different way to make sure that GPT understands the context. Lastly, you're going to outline the constraints. So what are the must haves and what are the nice to haves? So the third step is retrieve. If you've done step one and two very well, retrieve is gonna be very easy. What you're going to do is just copy and paste the output. You can either copy and paste it directly from chat GPT or you can export it to a file type like a PDF, a Word doc, or an Excel, or you could be working directly within GPT for sheets for an easier output. The last step is scrutinize. So this is the review period. What's really exciting is there is a very near future where GPT is gonna be able to do this step for us. For now, what we're gonna focus on is the who, the what, and the how. And so first of all, who should review it? When you're initially building out this framework, you should be having anybody review it who has a stake in the final output. And so that could be SEOs, that could be subject matter experts, product managers, anybody who wants the best possible output for this work product. The next step is gonna be the what. So what exactly are you reviewing for? You wanna review for everything. You wanna review for grammar, make sure the information is accurate, make sure the brand guidelines are followed, make sure we've got the subject matter expertise incorporated, and most importantly, make sure that the output is unique. You can be doing this with a variety of tools and people. Lastly, how do you do this? Ideally, you just latch this on to your existing acceptance criteria or SOPs in order to create a checklist to run through. The last tip here is any feedback that you identify as part of that review period, you can feedback to step one or step two and make sure that you're applying that feedback to the framework to make it better for next time. And so that is the process. That's pars, populate, activate, retrieve, and scrutinize. And this is a process, again, that you can use to leverage generative AI for SEO and content in a way that is both safe but also effective. So thank you for spending your time here with me today. For more ideas like this, please follow us on the SEAR newsletter. Thanks.