Understanding Semantic Search and AI Content to Drive Growth in 2023 by Tyson Stockton of Previsible
Bernard Huang
Tyson Stockton of Previsible joined us for a discussion on semantic search and AI content.
Tyson revisited what semantic search is and how search engines understand language by briefly looking at the history of semantic search.
He shared how AI content can fit into your content creation workflow before sharing 5 tips for AI content.
Beyond AI content creation
Human + AI
Use specific prompts
Google Sheets
Utilize existing scripts
Watch the full webinar
Click through Tyson's slide deck on SlideShare here.
Check out the resources Tyson shared below:
About Tyson Stockton:
For over 10 years Tyson has played an integral role in driving success within the fields of SEO & digital marketing for some of the top online businesses. Tyson has guided Fortune 500 companies to better utilize search data to drive successful SEO and content strategies.
Follow Tyson on LinkedIn: https://www.linkedin.com/in/tysonstockton
About Previsible:
Previsible specializes in providing educational and productive SEO advice and consulting to various global brands to build their online search strategy.
Read the transcript
Tyson:
Yeah, let's jump right into the slide. So, before we get into content, just a bit about me and Previsible.io. So, I've been in the SEO industry for a little over 10 years. So, I started on the in-house side with sports specialty e-commerce, leading that for 14 domains, both in the US as well as some time in Europe. I also worked at Searchmetrics as their VP of services, leading the service side of the business. And since then have started Previsible.io with Jordan Koene and David Bell.
Previsible, we're an SEO agency, and simply, our goal is to equip organizations with the tools and systems and personnel to thrive in the SEO environment. And we believe that, to help organizations, there's more than just traditional consulting services that's needed.
So, if we go to the next, the services and offerings that we provide from Previsible is, of course we offer a strategic SEO consultant. So, within that, the focus is within implementation. We don't just complete SEO audits, throw it over the fence, but we really work with the cross-functional teams on the implementation efforts.
We also provide resourcing and recruiting, so helping teams expand their SEO group as well as helping SEO individuals find new opportunities, and enable solutions of how we arm and support internal SEO teams to equip the stakeholders that they're working within the organization.
The next slide, these are a few of our clients. We work across the board with the enterprise, as well as startups, or really, if you'd like to learn more about us, we have some additional notes in the appendix of the slides that will go out in the recap of this presentation.
But to jump into the content, as I said, we really need to understand semantic search, and I think that's something that really helps give us the parameters and the playing field that we're operating with. So, by understanding where modern search engines are at, and then also understanding where the limits and capabilities of these AI tools are, we can then properly apply a strategy that's going to be effective in SEO today.
So, everyone's been hearing about this. You cannot go on LinkedIn and see things about Microsoft stealing market share from Google, open AI's advancements, Google's Bard, and there's a lot of commotion going on around this. And sometimes it feels a bit like noise, but there is a lot of potential, and this is going to be something that's really important moving forward in SEO.
So, going on to the next one, we have, in my opinion, two sides of how this impacts our industry. We have the SEO landscape side, and what I'd refer to as Next Generation Information Retrieval Systems. So, that's more of, how do we evolve and how are we changing how we interact with information retrieval systems, and that's more in the form of what Bing is released, the Google Bard, and this more chat orientation of getting information out of these systems. So, that to me is a shift in the landscape of how we're interacting.
Then, we also have more of the tactical side. And for us as SEO practitioners or content practitioners, that's really where we can have an immediate impact. And there is value in understanding and having strategies around how we compete in a shifting landscape, but today we're really going to hone in more on the SEO tactics side of how we can apply this to our work today.
Yeah, sorry, Travis. There'll be a few animations on these next ones, so we'll try to juggle through that. So, integrating AI and processes requires a few things. One, understanding how monitor search engines function today, and two, the current possibilities and limitations of the systems. And again, for me, I always think in the frame of, okay, what's the playing board that I'm operating within, what's the capabilities? Now, let me be creative and create strategies within those.
So, now, going on, we're going to start with the first one of understanding semantic search. Now, AI is a super broad topic, and it's being used a lot today in more of a narrow focus. So, I think it is worth just level-setting and knowing, okay, these are more umbrella terms, and there's a lot of components and elements that follow within it.
So, if we look at AI, and what's shown on the screen here is not to say this is all of AI. The gray box area is simply elements that are relevant and found within semantic search. So, you have natural language processing, which is interaction between computers in human language, you have machine learning, which is training a computer system on a data set to perform a specific task, entity recognition, which is a really important one that we'll hit on a few times in this, and simply an entity is basically a noun.
It's something specifically definable in search. So, Paris, San Francisco, dog, myself, Tyson, all those are entities. And one of the important pieces of this is how these systems work as it's looking for the relationships between these entities. So, there's also expanding upon entities, sentiment analysis, so understanding human emotion behind things, and as well as looking at LSI, or latent semantic indexing, so understanding the relationship between words and phrases that changes the underlying concepts.
And what this is doing for search engines is, it's allowing them to understand intent better. The better that a search engine can understand our intent, the better that they can surface the right content to satisfy that need. Now, Google didn't start their journey as a semantic search engine. Simply put, this technology did not exist when Google was initially launched.
If we think way back in the early days of search engine optimization or just search engines in general, it was mainly keyword search engines. So, it was, I want to rank for iPhone, I'm going to rank for iPhone 200 times on a page and I have a good shot of performing on an iPhone. [inaudible 00:06:25] just back one more on that.
So, there's been this journey and this slow transition into where Google's at today, and I wanted to call out just a few key milestones, but also highlight, to the viewers, that this wasn't an overnight switch that Google just decided, "Hey, we're gonna be a semantic search engine now, we're moving away from keyword research or keyword structured piece."
So, they really started this journey back in the early mid-2000s. You see a representation of that with rich snippets where they're able to extract out pieces from it and surface it in the SERPs. But really, where the transition made a huge step was with the Hummingbird update in 2013. This was the first algorithm update that really put semantic search at the core of the algorithm system, and then we've seen it just step-charge since then.
So, RankBrain is another super commonly talked about one, the application of machine learning into it. So, as these underlying systems develop, you see these rollouts over time. So, Bert is another super common one within this, among even looking at between languages, between medium types. So, all of this is building into where Google's at today.
So, if we go to the next one and we look at, well, what's a visualization of semantic search? So, in this example query here, we have the best restaurants in San Francisco. Now, understanding that "best" also means the same thing as essential. So, the top, granted there's a geo bias on this, but when I did this, SF Eatery was the number one listing on it, but they're not saying "best" in it as you'd see on a keyword search engine. You're seeing the essential piece behind it.
Then, following that, you also can see things of, well by understanding "best" means similar to, essential, top-rated, things like that, Google's able to say, "Okay. Well, I can look at the restaurants within this area with the highest customer reviews, and I can surface that because that's gonna satisfy a similar intent from the other." So, Sotto Mare, if you're from San Francisco, North Beach, famous Seafood plot. Good call on their part for putting it in.
Then, going on to the next part, the other interesting piece that I think this highlights is, when you look at these "People also asked", to me, this is interesting when you're thinking about it in this lens of, okay, what does "best" mean? So, they're making this connection of "best" to "most famous", which is not the exact same thing, but you can see that there's a relationship to it. Then, with "famous" you have "famous people in San Francisco".
Now, we jump to Anthony Bourdain as being an authority in that best restaurants area, and then, "What should I not miss in San Francisco?" So, this is expanding out from that original query of "best", and connecting to other entities that are related to that because of the relationships that they have.
Going on to the next slide, this is a bit of an illustration of Google's Knowledge Graph. And simply, what a knowledge graph is doing is, it's the database of information and the entities and their relationship to one another. And just like in that example, if I'm looking at the visual on the screen here, the Eiffel Tower has a relationship because it's located in Paris. Paris is a city.
So, by having this context between these entities, Google and other search engines are able to understand relevancy, and then reward content that's dressing these entities and describing a topic more holistically.
Bringing it back, and I know for some this might seem, yeah, obvious, we've been in this state for a long time, but I talk to a lot of individuals and maybe they're at different stages in their career path within SEO, that there's still some underlying mentality of, "I need to use the exact keyword word, I need to put it in my H1s, I need to use it however many times in the text."
But it's not simply just the one keyword that you're trying to rank for. The way a search engine, today, is looking at it is in this lens of semantic search. So, if I want to rank for iPhone, well, I need to be addressing these similar related entities to it, and so the keywords that's going to build relevance is going to be "recent models". So, iPhone 14, iPhone 14 Pro, 13 SE, the prices, screen resolution, storage capacity, and all these things. But in the past it would be more of, I found this many times.
Next piece here, I grazed over this concept of semantic search, and I hope it gives just some reminder or context to the audience of how the pieces fit together, but I think it's an important framework to think within. And there is a ton of really great information about it. If you are interested in doing more of a deep dive in this, I'd highly recommend checking out SEO by the Sea.
There's just a great body of work here by what Bill Slawski did by looking at the Google patents, interpreting those, and offering just a crazy amount of knowledge and breadth of information out there to the world. So, if you are interested, check out this website. I'll have a few other plugs of other resources that you guys can dive a bit deeper.
Now, let's talk about where AI is today. So, AI has been gaining a lot of popularity, but this isn't necessarily a new concept within SEO. So, there's new formations, and we're at a new stage of it. But when we look back to that early slide in the deck of AI and how semantic search fits within AI, and even looking at Google's timeline, we can see that these machine learning and these sub elements within AI have been applied at different stages.
So, if we're thinking about, regardless of what we're showing in search, image and visual search and all these other things, these are all different variations. And one of the things that's driving this increased interest or awareness today is the moving and shifts from classification task to generative AI, and that's what we see with ChatGPT.
So, AI is constantly improved and performing tasks over time, and it can't really take on replacing an entire process. So, we're still in this stage of being able to perform specific tasks that we're giving it, with prompts, and then, eventually working to that owning or being able to drive an entire process intent.
So, systems like ChatGPT, they can be very powerful and effective, but you still have specific inputs to then determine what those outputs are. And another way that this is driving awareness and driving applications within the industry today is this low code AI development. As it's speeding up, it's lowering the bar of entry to leverage and utilize, and that's opening up a ton of opportunities for us within SEO and other applications within our world.
So, let's talk about AI and SEO. We've seen it a bit from the search engine side, but from the operators or us practitioners, we see it in a variety of tools. We see it in keyword research tools, content optimization tools, link building tools, rank tracking tools, content creation tools, and analytics tools. Being that AI can perform these specific tasks, we're seeing it in a variety of forms.
So, the main point that I want to drive home by this is, when we're thinking about AI within SEO, we need to also have a broader lens of, hey, this is not just for content creation, but it has other applications for it. But of course, today, content creation is the main driver and the main piece that we're seeing in the media and just conversations at large.
So, going on to the next one, if anyone's been on LinkedIn, you've probably seen this just flurry of AI content and just AI ChatGPT information. And there's a lot of really solid content in here. And some of the screenshots that I grabbed and I threw up here are great articles and things to take a look at, but also, to me, a lot of the narrative, again, is just pretty narrow-focused into this assumption of, well, is AI content good or bad?
And the main thing I want to drive home with this is, there's plenty of different applications of it, and using our creativity we can really unlock a lot more potential than just, "Hey, is this good or bad?" We can see too, AI SEO search demand versus AI content. AI content has just exploded. So, there is a lot of value to it, there are a lot of capabilities to it, but again, it is just scratching the surface on what the overall applications are within our industry.
Now, let's dive further into these content creation tools, and let's talk about how we make the best or how we should be thinking about the limitations of this and how we can then get into a state of actual usefulness. So, ChatGPT sparks this debate, and again, one of the huge value ads to it is the accessibility. It's so easy to get into ChatGPT or open AI. Now, we're seeing it and Bing, we're having Bard come out.
So, being able to just quickly and easily access this without having to have familiarity with coding is just opening up the doors for this opportunity. It makes it so anyone can jump in there and... You've probably also seen debates and things. I've seen stories of colleges removing entry essays because of the fear of kids just cheating and using ChatGPT to write their essays.
And another reason why this is boiling up and being so pointed in the industry today is the significant strides and improvement in the quality within the GPT architecture. So, the quality of output between these tools is really strong right now. And it's, "Sure, you can do it."
And there's stories of, sometimes, people being, "I got wrong factual information. And this doesn't sound good." So, yeah, it's not perfect, and this is still an emerging and a developing area, but it's something that, if you look back to where it was even just a few years ago, the benefits are just so much greater and the quality is so much greater than it was that it's at a stage that has a lot of real applications to it.
So, what are some of the limitations though within it? So, some limitations occur, and it sounds obvious, but the inputs are going to control the outputs from this. Sorry, let's go back one more time. So, the inputs and the outputs, and I'm going to come back to this because being that it's still task-specific and it can't create this overall process, that's really significant because it keeps us grounded in it, and we have to be consciously thinking about what direction are we giving the tool, and then that's going to determine the quality of the output.
Another limitation that we need to be aware of is, these models are being trained on different data sets. And these are massive, massive data sets. And I'm sure some people, if you've been playing around with ChatGPT, you can ask, just sometimes, a question and it will say, "I'm sorry, I was trained in this time period. I don't have the most recent information," or something along those lines.
So, you do have a limitation of what training these models have gone through to the surface what it's able to provide out. Another area of a potential limitation from this is the need for validation. So, if you ask ChatGPT a simple question like, what time is it in Berlin? And actually that's probably a bad one because I was playing with it and it had some struggles there, but who was the first President of the United States?
You can look across the web or across the train data set it was on, see a high volume of the same answers being responded to, so therefore it has confidence and therefore it can surface it. So, not only do you have a limit of what is the data set that it was trained on, but also, it's limited to what can be validated. So, it has these limits of creation of original thought.
So, that's going to be an important piece that we come back to because original thought and original human creativity is an area that we own right now as far as the capabilities of the system. So, if you know that and alluding ahead to it, that's where a lot of our focus needs to be then, is in this human intervention, this human aspect of it. So, if we know, okay, it can provide these validated sources and systems and then know where to put in our human judgment, we can already be off to a good start. Dog in the background.
So, let's go to the next one. Another resource that I'd recommend looking at, a little while back, I think it was late December, I got a chance to jump on with Bernard Huang, co-founder of Clear Scope, and had him on the Voice of Search Podcast and really had a great conversation. I loved connecting with Bernard, and we talked about the same concept of where it is today.
Yes, we can dig that up and get a link to the episode there. Yeah, maybe when we're doing another Q&A, I can throw it in there. So, this episode, it's roughly 35 minutes, I'd say. Great listen. Bernard's super knowledgeable on the subject as well, so I'd encourage you guys to check out and hear what he said on the subject. But again, it touches on, where are the limits of the system and what limitations? Does that exist for having success in SEO today with content creation?
Now, let's think of these potential challenges a bit more. Thank you, Travis, for throwing that in this chat. AI content has limited information that's already been published. So, I talked about that, but it's a really important piece because again, that's the boundaries of where we're operating. You have this other huge benefit of, hey, AI reduces the cost of production, and with that cost of production decrease, you have the potential and the ability to just explode volume of content on the web.
And sure, everyone's not using this yet, but if you think of it, hey, it used to cost X amount to produce one piece of content or have a writer come in and do that. Now, it's, well, heck, with proper prompts, I can just crank out 1,000 pages in no time. So, this is an interesting piece because it creates a challenging situation for search engines.
So, you have the limit on what's available to it, then you have the lower barriers of even, hey, anyone can jump in and use this, the lower barrier in cost, and so you have this, now, potential to have a high volume of content on the web that's basically homogenous. Now, we think about, okay, well, what happens if most of the content online is essentially the same because it's only addressing all the known entities that can be validated through these trained systems?
So, let's jump to the next one. So, walking through it a bit more. First response I would have is, okay, well let's run through this thought process. All the content, it's not worded the exact same, so it's not duplicate, but it's saying the same thing. Well, doesn't Google have 1,000s of ranking factors that then it can still determine best content by quality of backlinks, authority of a domain, markup on a page, social signals going to it?
So, it's, yeah, okay, so we go through this and we could still sort and order and we could prioritize who's the best still, even if it's almost the same content. Well, but then that still gives a challenge, and you have this other notion of, well, Google rewards originality, and everyone, pretty much in SEO, knows duplicate content, bad. There's no debate around that.
We're debating whether or not AI content's good or whether AI content's bad, but no one debates whether or not duplicate content's bad. So, if we think about that, it's, well, why does Google not want to reward duplicate content? I'd love to say it, but it's, "Yeah, it's not just because Google's great and great intentions and they're doing everything for us." It's, "No, that's a bad user experience."
A lot of things within searches, you could be thinking about Google in the sense of their business, what their product is, the user experience when someone submits a query and finds what they're looking for. Well, from a user experience perspective, if they land on a search results page and every single page has the exact same content and it's duplicated, that's not a quality user experience.
So in the same mindset, if all of the articles are saying essentially the same thing in different words, but it's still the same information, is that a quality user experience? So, I believe that from this, this is part of the reason, and it's part of this challenge of, one, Google's faced with, hey, there's potential for an explosion of content, so crawling, indexing, all these costs could significantly increase from that.
You also have the thing of, well, what user experiences is this going to be if everyone's saying the same thing in different words? Then, it's, how do we award originality and how do we provide the best user experience? So, I think this is really a key piece. And we've seen in the communication that Google initially mentioned the helpful content, and I believe that was the prompt of why Bernard and I had that conversation on the Voice of Search . It was coming right off the heels of the helpful content update.
And it's, okay, well helpful content, shouldn't that be all of Google's algorithm updates? Is it all trying to surface helpful content? But in that, a lot of the early messaging was around machine content. And since then, Google has retracted and stepped back on some of their PR messaging from this. But I don't believe that it's necessarily saying machine content is bad. There's been applications of machine content for a while. There's plenty of people that have come up on LinkedIn and Twitter and saying, "Hey, here's an example of a company that uses AI content that's still succeeding very well."
So, it's more of this notion of, how do you award and how do you give credit to originality and human thought that may be in addition? So, if we go into the next one here... Actually, sorry, before we jump into the tips, the piece that would leave this, if we go back to thinking about the knowledge graph, semantic search engine, you're evaluating and you're rewarding websites or content that's addressing the known entities around the topic.
So, again, if I'm Google, I want to surface the best content that's going to make sure my users are finding what they're looking for. Best chance for that is that they're addressing everything that I can identify and I know about it, and that's going to be the most boring, but you're not progressing the knowledge graph, you're not progressing the known information. So, adding originality and adding and introducing new entities to this concept creates more of an authority, and it gives a reason to give credit or give preference to content that's adding something new.
So, let's close this out into tips for AI content, and we're going to leave a nice little chunk of time for a Q&A after this. But the first tip I have for thinking about AI and content is to not... So, think about AI content as beyond just written copy. And being that the systems and a lot of these examples are skewed to ChatGPT because of the power of it and just recency, but to be thinking about content creation within your organizations and within your efforts, not just of what is the writing process of it, but what's my entire process to get to that publication going live to site.
So, within this, we have a strategy aspect of identifying keywords. We have, well, how are we clustering those keywords to have a representative actual search demand that we can prioritize our efforts? We have limitations of what topics we want to address, where they're going to live in the site, process for getting it live to site, and so we have this need of creating the roadmap for it. You'll have optimizations of metadata's titles, whatever.
So, think about AI's capabilities within content creation around your entire process, and then be using it to identify areas for opportunity to improve volume and pace and what you're able to get to, and then be applying your own judgment on top of that. And I think the best way to think about ChatGPT, or anything else, is to think of it as a tool. It's a tool in our toolbox and it's something that we can use.
So, we went from building a house with a hammer and nail to now we have a nail gun. It doesn't necessarily mean that we're going to build a better house, but it means that we can build it quicker. So, you still have to have this mindset on quality and having those quality assurances, but there's also a variety of ways that you can use this. So, think open-minded on this, really sit down and look at what process you have today, and then where can you apply these into even pieces of it that's going to allow you to save time, increase efficiency, and really drive success.
Next tip, the human aspect. So, I was really hitting home on this before of where the limitations are, and I think one of my top recommendations in this area is, know the limits of AI systems, and capitalize on the strength or what we have to offer on the human side. So, focus it on originality, creativity, introduction of new topics to this, and this is going to help your content set aside.
It doesn't mean there aren't queries and there aren't spaces where strictly AI content can still succeed, but at this stage of the industry, I wouldn't say that that's a sustainable strategy to it. So, within certain queries, there's still human aspects. Add human QA to it. QA measures are super important on content production, because we have this ability to generate thousands and millions of pages, but if our quality isn't there, we have risk within how we perform in search.
So, you pair human and AI together, focus on how humans can add originality and something new, and then, also make sure that we're applying the proper Q&A measures, and then use human judgment to validate some of what we're putting out there.
Use specific prompts. Now, this again comes back to where the technology is. So, the quality of the output that you're going to get from ChatGPT is going to be strictly related to the quality of your prompts going into it. And this is, I think, like something that a lot of us have seen, and it's something that you can really experiment with.
When you're in those early stages of defining what you're going to be, how you're going to be using this, don't just do one prompt and say, "Great, looks good. I'm gonna run it through 1,000 pages now." Have some iterations, play around with several versions, and then evaluate which one's give the best quality, and then scale out through this.
So, if you're going through an exercise to rewrite meta descriptions for 1,000 pages, great. Play around with it, and write an enticing meta description. So, objective meta description, not a ranking factor, has a play-on click-through rate. If we can encourage people to click through to our site more, we can have a positive impact on our rankings. Also, we're driving more traffic.
So, key there is, make it enticing, make it catchy. That could be another prompt to do it. You could also use things like, "Write as an SEO expert," or, "Write an SEO optimized," or, "Write as a New York Times journalist." So, I think that grounding can really improve the quality of outputs. Test, test, test, then rollout.
Next tip, bring it outside of ChatGpT. ChatGPT is awesome. Often I have just a tab open and there'll be a random thing and I jump in, I do a one-of. But again, you're doing a one-off in that nature. Super powerful. You get a lot of good information. There's ways of getting information that you may not know. It's not in the tips, but there's been a lot of talk-to as well as, "Hey, you can use it to get a schema markup pulled from it?. I don't know how to code. Cool. ChatGPT's gonna do it for me."
But again, you're still in this one-off. So, I think there's a limitation of really capitalizing on the technology here because the technology allows us to scale. It really allows us to do things at a far greater rate than we ever were possible doing. But to do that, you can't be doing it one-off. So, an easy low technical ability is to bring that with the open AI API, bring it into Google Sheets, and then you can have one prompt for it, scale it across the entire sheet, and now you're operating a larger scale.
If you're not familiar with doing it, Tharindu has a great article here on his SearchMinistry. He walks through and gives the prompts of how to make the connection with open AI and how to bring that into it. He also gives an example for rewriting meta descriptions. You could just make some simple tweaks. And again, you don't have to be an expert in writing API calls or interacting with these systems, but he does a really great job at breaking this down.
So, I think this is a super resource that if you've been interested in it but you're nervous of, "I don't really know exactly what I'm doing," great resource for it. And this again is, it's going to unlock the scalability of it.
Next tip, utilizing scripts. Now, we're going a bit more into the technical realm, but again, the capabilities and the potential from this is so much greater because instead of having just one prompt, one output, you're able to string together different prompts from it. So, Kristin has been throwing out a ton of awesome content on LinkedIn, and she's been creating these scripts and just posting out, offering them for free.
We have an amazing SEO community of people that are willing to help out each other. It's one of the things that at Previsible we really want to lean into and also give back into the SEO community. Part of why I want to give shout-outs to these other people, I don't have an affiliation with them. I came across their content and thought it was helpful. So, I've been just trying to also spread the word of it.
Within the example or the screenshot I have is one of the prompts that Kristin shared with me. It's creating an entire content plan. And it's not just saying, "Great, I already have these pieces," is, it has prompts, command, output, and it builds upon it, but even, again, if you're not super technical, you're feeling, "I wouldn't be able to write that code myself," these prompts are really clean. She clearly marks where you're making inputs to it, and it allows you to actually capitalize on it.
So, again, if you're not technical, and I know some technical people on this call are, "Yeah, okay, I can write that myself." If so, awesome. Use this then for getting ideas of doing it. She has really creative ways of giving context in her prompts, and so I think, even if you're coding, you're familiar with how to do this yourself, you could probably get some good ideas of just seeing the context and the parameters to the prompts, and then how she's streaming them together.
Almost all the prompts that I've seen from her are all in the relation or in the vein of this, news, journalism, content creation. So, again, tons of great resources here from Kristin on that.
So, to close things out, I would revisit and I would say, when you're thinking about AI content, one, keep in mind AI is very vast and broad. Everything from self-driving cars, visual search, even your auto complete, is variations from that.
So, there's a huge range of it, and we're just really scratching on the surface where it is. So, be constantly keeping an open mind and looking for new applications and where it is. Second one, be thinking of where the search engines are at. We do have this conversation that's valid and warrants what's happening on the landscape, how are we interacting with the next generations of information retrieval systems.
That's something important too, but we need to think, okay, this is how a modern search engine like Google works, this is how they're using entities, how they have this basis of relationships, how they want to reward this, we understand the challenges that they have as a business, and then, from within that, we can operate more effectively. We also need to know where the limits are on Ai.
Where I specified, and a lot of my recommendations are based upon where the technology's at today, this is not going to be the same as where it's going to be in maybe 12 months, 24 months, maybe even the end of the year. So, again, keep this as more of a mindset of, I need to understand this side, this side, within it I can create our strategies, honor strategies. Think of it broadly. Again, use specific prompts for it, combined with human intervention. Be smarter about how we use our time.
Maybe AI and these tools cover 80% of the production work, and we can then apply our time into 20% across each, and we just expanded the reach of our impact. Be thinking of ways to do that with scale through things like Google Sheets integrations or custom prompts. And these elements are really going to allow us to magnify our impact, magnify the body of work that we're getting live to site, and ultimately, drive more traffic and success for the businesses that we're operating within.
So, that wraps up my presentation. I hope everyone's found that interesting or at least a reminder of how to be thinking about these topics. And if anyone has any questions, happy to go through it.
Travis:
Awesome. Yeah, that was super helpful. Also, pretty topical with ChatGPT releasing their API, I think, yesterday. We do have a couple questions. And kicking it off, you mentioned using human input for or making the content original. Do you have any ways or any strategies on how to do that when you're scaling content with AI?
Tyson:
Yeah. And it could be through an editorial process, so maybe you have, and I've seen people do this, whether it's through, "Hey, we're gonna have AI create an outline, and then we're going to add to that, or, "Hey, we're gonna have Chat GBT create an initial first draft," and then you have the human piece of one looking at that, "Great. Is this our brand voice? Does it match how we want to be presented?"
But you can also then look at what's covered and use that human creativity and originality to then add something to it. So, let me see if I can think of something off the top of my head. Okay, let's go back to the best San Francisco restaurants. So, obviously, the top-rated ones are always going to be on the list. They have that through structured data on customer reviews. You also can look across a variety of different blogs, and it's going to be a lot of the same restaurants.
But then, if you go in and say, "Okay, I'm not gonna pick, necessarily, the highest rated ones, but I'm gonna pick a new restaurant that maybe doesn't have as many reviews or it's kind of more trending within the social," and so you could be looking in other data sets to get that, you could have just personal expertise on the matter. So, if I'm writing an article of best restaurants in San Francisco, and then I'm adding a section that is "best new restaurants that have launched within the last six months", or whatever the new trending restaurants within San Francisco.
This is going to be something that's net new, that maybe outside of that trained knowledge of the current systems. It's also something that maybe isn't validated because not enough people are talking about these new restaurants, but it allows me to add something new to the knowledge graph, and it allows me to introduce a new entity in its relationship, which then I can set myself apart with other ones.
So, it's almost like that 80-20 rule of, "Great, I'm gonna use this tool for 80% of the work, and I'm going to add that 20% of humans on top." And I think that's a really strong way to be thinking about this. It's not at a point that it owns the entire process again. It's specific tasks, and we have to be strategic in how we use those tasks.
Travis:
Nice. Yeah, that's helpful. And we have another question. So, this is from Anonymous. They say they're new to this side of the SEO side of marketing, but any thoughts on things or factors that shouldn't be overlooked at all, from an SEO perspective, while creating a blog post? A back-to-basics question.
Tyson:
Yeah. And sorry, can you repeat? I missed one piece in the middle of that. Can you say that one more time?
Travis:
Yeah, yeah. Any thoughts on things or factors that shouldn't be overlooked at all, from an SEO perspective, while creating a blog post?
Tyson:
Yeah. So, that's back to our fundamentals. For fun, you could throw that in ChatGPT and see what the output is from it. It's going to be a lot of your just standard basics. So, you have, on one level, the metadata that you're providing to it. Obviously, you have your titles, your H1s, your meta descriptions, those elements. You have the actual content which we talked about, are you addressing everything that's around the topic, and how well are you holistically addressing a topic.
You also shouldn't forget things like structure. This can be really helpful in getting into position zero on queries. It's also a way to be more competitive in certain areas. Another area too, and it gets a bit outside of this, I would be thinking about content too beyond just written copy, especially as search engines are evolving with image recognition, video's performing better and better in search.
So, think about your content beyond just written copy and what other media elements that can be included in it. And then you have an entire technical side, which is just more of, do I have any limitations that are going to hold back my content? Overall though, I would say, and there's 1,000s of ranking factors out there, the biggest area is to be thinking about the quality of the content, and then making sure we don't have those limitations for it.
But again, break the mindset of, I just need to repeat this keyword X amount of times. That's something we should be universally aligned on as duplicate content is bad.
Travis:
I agree with that. And this might be the last question. You gave some resources, like Kristin's LinkedIn is fantastic, and ChatGPT put this into the foray as far as content or prompt chaining, but what are some other resources to learn more about prompt chaining or prompt engineering?
Tyson:
Yeah, I'd have to... Feel free, anyone on here, reach out to me on LinkedIn. I wouldn't be able to just throw them out off the top of my head, so I'd have to follow up with that. Happy to do so. So, yeah, I'm most active on LinkedIn. I am on Twitter as well. But anyone's welcome to reach out to me, hit me up directly, happy to follow up on these.
But I think, too, I would do a few different searches for this, and then, you could get an idea of just the basic structure of it, or there also is a bunch of people out there that are just sharing, "Hey, this is what I've already created or done." So, sorry no more help on that, but yeah, hit me up, and I'm happy to share some links afterwards.
Travis:
Awesome. And I dropped Tyson's LinkedIn in the chat. Definitely let him know what you thought about today's webinar, give him some shout-outs. And Tyson, do you have anything you wanted to share before we give everyone their time back today?
Tyson:
No. I appreciate being on, and really like the work that has been going on at Clearscope. So, yeah, happy to be a friend on it, having you guys on the podcast doing this here. So, thanks for having me on. If anyone's interested, from career development, recruiting offers, check out Previsible. We offer SEO consulting, recruiting. So, if you need to expand your team, we work with people to understand the needs of the role and what they're trying to achieve in SEO, to then help them define the exact type of SEO they want, and then we help secure those.
Also, if you're an SEO looking for work, feel free to reach out to me. I love helping other SEOs find their next opportunities. So, regardless if it's something that we're working on recruiting or just some introductions that I can make, I'm happy to help. And yeah, check us out @previsible.io.
Travis:
Awesome. Well, appreciate your time, Tyson, and thank you, everybody, for your time. Have a good day. Bye.
Tyson:
Awesome. Thanks, everyone.
