How SEOs Can Leverage AI (Like ChatGPT & Bard) Today by Natalie Henley & Richard Lesher at Volume 9

Natalie and Richard of Volume Nine joined the webinar to share how to leverage AI for SEO.

They covered:

• A little history on AI

• Pros and Cons of AI

• Tactical Ideas and applications for using ChatGPT & Bard

• Speculation on where this is all headed

• Some of our favorite AI tools and resources

Check out Natalie and Richard’s slide deck.

And check out the resources Natalie and Richard shared below:

• Volume Nine’s ChatGPT Prompt Cheat Sheet

About Volume Nine:

Volume Nine has been helping clients with SEO since 2006. They help brands build their presence organically so that they don't have to pay for every click long term. Natalie & Richard have been speaking for years and have both been in the trenches for small businesses all the way up to Fortune 500s.

Follow Natalie on LinkedIn: https://www.linkedin.com/in/optimization/

Follow Richard on LinkedIn: https://www.linkedin.com/in/richard-lesher/

Natalie:

I want to start off with a little disclaimer. It is impossible to make in a presentation about AI and have that thing stay accurate for very long. We finished up these slides yesterday. And between yesterday, and today, AI has already changed. Just know that as a company and what Richard and I want to do today is we've been staying as up to date as we can as much as humanly possible on what's going on with AI and specifically how that affects digital marketing and very specifically how that affects the field of search engine optimization. I think it would be really, really impossible to give you everything about every AI tool and everything that's happened.

I mean, I think every tool on the planet has actually launched an AI application of some sort in the last three weeks. Today, what we want to do is bring you some information, give you some ideas, give you some fuel for thought, and then give you as many recommendations and ideas of how to tactically start using AI like ChatGPT and Bard in your different workflows. But I would definitely say that where Richard and I come from is we both have really strong SEO and digital marketing backgrounds. We've been really trying to figure out where this works well in workflows and where it doesn't. It's not just does the tool do this, but does it do it well?

And then also, I think we both have a level of healthy skepticism towards AI. I don't know if I'm going to go buy an OpenAI t-shirt anytime soon. Richard might. But I would definitely also caution to say that we are not AI experts. I think that we can bring you a lot of information, but neither of us work on a language model. We don't have a fancy database. Big asterisk for today is we're really trying to make this actionable and approachable and to synthesize a whole lot of information down into something useful for you. And then also I want to say, I think a lot of people as they start to learn a little bit about AI start to freak out.

I think the big thing is, is don't panic. The raw truth is, is that this is developing faster than really anyone human or even one company could really stay on top of. We're all in this together figuring this out. I think people are panicking about, am I going to lose my job? Is this the end of civilization as we know it? Is my company already behind? I think it's really hard when you go through times of massive change to just freak out about the future as much as start. Today, we really want to encourage you to take a look at some of the options of what's available and really start to figure out what you can use these AI tools to do, whether it's content, whether it's other more complex SEO workflow items.

I think today is about, can you try it? If you're trying it, can you lean in a little bit more? For today, we're going to cover some big sections. One, I want to give a little background on AI. I think some people are super in depth on this, but a lot of people have missed the key points of what's happened with AI, and then some of the pros and cons, because it's not all just roses and unicorns and confetti. There's some definite drawbacks and cons to artificial intelligence. And then we want to talk to you about just technically, how do we use artificial intelligence to build efficiency? How do we get it into our workflows?

What do these tools do? And then we're going to end with that great existential crisis. We're going to start talking about ethical issues, legal issues, things that are coming down the pipe. We're going to hopefully get everybody on a mind of not only your great existential crisis, but from our perspective, where does it make sense for you to be focused as you're thinking about these tools and the launch of artificial intelligence worldwide? A little history lesson. AI is not a new term. In fact, artificial intelligence was termed in 1956 at the Dartmouth Convention.

Between 1956 and 1997, artificial intelligence was really owned by science and different scientists trying to figure out, is there a way that we could get computers to think? Can we do something that feels like language modeling? How can we make this? Could computers even become sentient? And then in 1997, something really interesting happened. In 1997, IBM had been working on its own artificial intelligence, Deep Blue, and this artificial intelligence beat a champion chess player. For the first time people went, "Oh my god, this can think. It can critically come up with solutions, and it can outsmart a human."

Between 1997 and very recent history, I would term it as artificial intelligence was more or less controlled by big corporations, controlled by IBM, controlled by Google, and all these different companies were building up language models and trying to figure out how do we make software that's smarter. And then in 2015, OpenAI is founded on the whole principle as a nonprofit that they're going to make artificial intelligence accessible worldwide. It's interesting. A few years later, they became for-profit and then they started launching some different language models. That's when we started to see a rise in these we'll call them lower cost tools.

A really common one that a lot of people have been using for a few years is Jasper.AI, where you could use Jasper.AI as an application. It's set on top of OpenAI's language model. It wasn't that expensive. The content written, when you read it, it still sounded like it was written by a robot, but it was pretty good and it was fast. You could take that content, work on it a little bit, and it passed that bar of good enough for a lot of people. And then in 2022, two things happened simultaneously that lit this whole thing on fire. One, OpenAI updated their language model to GPT-3.5.

And then on top of that language model, they launched a widely available free application ChatGPT, which I'm sure we've all heard of by now. Within a few months, the world and the conversation around artificial intelligence exploded. Even though artificial intelligence and language modeling isn't new, this is the first time that we've had a language model that when it writes content, it's really hard to judge it as being written by a robot. And then two, it's widely available and free. This is important, and this will be definitely important when we start helping you build up your new existential crisis that you're going to walk away with.

It's important to understand how these language models work and how AI works so that you can understand the breadth. Not only on capabilities of what an application like ChatGPT can do or Jasper.AI, but then where this might be going long term. You want to keep in mind that everything starts with inputs and information that these systems collect and they get stored in a database. This is going to be important later when we start talking about ethics, because one of the big reasons why ChatGPT was banned in Italy is we're really unsure where these inputs are coming from and if they're following data privacy laws.

Hold that thought. You have to first start with, where do we go get information from? It goes into this database. And then from there, these AI companies build a language model. OpenAI has built more language models than just GPT-3, dash 3.5, dash 4. It's got I think at least five different language models that are designed to do something different. What GPT-3.5 and 2 and 4 all are designed to do is predict the next word. It's a whole language model basically designed around how do I predict the next word? This is important because there's language models that are designed to do lots of different things.

Just because one application on one language model can't do something doesn't mean that it's impossible for AI to do it. It just means that the one tool and application you're trying and playing with doesn't do it. And then application set on top of that. An application example would be Jasper.AI. Another example would be ChatGPT. Another application example is Bard. It's what you and I would see interface with and use. The reason why this is going nuts right at the moment is, one, we know GPT-3 was a much better language model. Richard, can you click one more? I don't see any text.

There we go. Dash 3 was already a really good language model, which is what Jasper.AI Was using, and then we upgrade to 3.5, and upgrade then to 4. As they defined it, it's a 10X improvement every time they're updating the sucker. And then of course, they launched this free application on top of it. Now, taking that all the way into digital marketing and SEO even specifically, AI very quickly went from this, oh, a tool to play around with or a nice to have to basically table stakes. If your content isn't as good as what these guys can do, if your tool doesn't incorporate AI, all of a sudden you're behind.

A really great example is in November, ChatGPT launches and Bing announces that they're going to work that into Bing Search Engine. Google within a month and a half declares a code red. Pulls its founders back in. Estimates that this is going to undermine the very fundamentals of search, although that's not what they're saying in their press releases now. They're like, "Oh no! The world's open and we're so excited about AI," but a few months ago they were not. They rushed to launch Bard. We know that tech is working. And then as I mentioned, basically every tech tool on the planet has launched AI in the last few weeks.

That really has pulled us to the front of what's going on. Some pros and cons of artificial intelligence. Before you're like, "Okay, it's floating. Let's go. Let's get in there," I think it's important to understand where these tools are great and where there's some drawbacks. However you feel about the pros and cons, I honestly think this is one of the best quotes from the CMO at Zapier, which is, "We are experiencing a seismic shift in how the world works." That timeline I showed from 1956 to 2022, it went from science to corporations to finally it was open. That's very similar if you look back to the timeline of email.

I think as impactful as email was on how we do business and how we work and how our workflows go, I think this is going to be as big of a shift for us. I think when you fast-forward into the future a few years, this is going to be part and parcel and core to what we do all. Some pros. I think a lot of people are seeing the sunny side on the sky of this is building efficiency. We're able to stay way ahead of our competitors, especially if we're incorporating it and they're not. This is going to save us money, and frankly, we're just going to be able to do more. And that do more, I think, is one of the biggest.

Because we know that in marketing today, one of the trends that we are continually seeing in digital marketing is that marketers are being asked to do more with less constantly. Having a tool that helps you double your workflow is invaluable. But then the other side of that coin, because AI's not perfect is one, there's some risks. We're going to talk about at the end. There's legal risks. There's ethical risks. There's risks to your rankings. Richard, could you click the next slide? Thanks. There's ethical concerns about AI. A really big one is, okay, if it all starts on this database that I showed you in that model, where is that data coming from?

Is that data inherently unbiased? Is it inherently not racist? Is it inherently fair? That's really hard to judge. It's also really hard for us to tell because we don't get access to the data that they're using. We don't really know all the places that they're getting their data from. And then one of the things is, is that these bots, although they sound a lot like humans and I think there was one language model a few years ago that even convinced a Google engineer that it was sentient, they really still can't think critically. And then accuracy is a big thing that we're just noticing across the board, accuracy in stats, accuracy in research, accuracy of information.

Especially if you're in an industry where it's critically important that you have accurate information, I would say it's always critically important to have accurate information, but especially if you have something like a compliance layer, you don't want to have a legal site and have inaccurate legal information go up. All right, so a few use cases. I'm going to break this down a little differently. I know there's some standard categories of how people are thinking about this, like, oh, these are good for Q&A, but I wanted to break it down into categories that make sense a little more for digital and SEO. Here's the five things that I think the AI does really well in categories.

The first is research. I think it does a really nice job helping you lay out research. With a big asterisk of when it gets all the way down to a data point, you have to double check that data point. But if you're like, okay, how do I outline this, and what are key things I should be including, and what are key questions I should be answering in my content, it does a nice job. I'd also like to note as we're going through these that all these pictures of the robots in this section were generated by AI, different than ChatGPT, but this was all generated via AI. I think a lot of us are seeing it's good for ideas.

Sometimes we're just not in the mood to come up with 50 blog topics for X, Y, Z, or we're just coming out of a data analysis. Crap, it's time for the calendar. Just to help you come up and brainstorm with ideas, this tool does a really nice job of just helping you think broader and get you into the mindset of coming up with new concepts and ideas. Next slide. We all figured this one. It writes a lot of content, small copy, long copy. This is probably one of the main applications that I think most digital marketers are using the tool for. Writing code, I'm going to put a big asterisks here because I promised I have a friend who's a coder and he asked me to just give this disclaimer, it can write code.

Please don't write code with this tool if you don't know how to write code. What's happening is people are going in and they're getting GPT to spit out code for them. One, the code doesn't always work. Two, it's not particularly secure. And then third, it's bloated, which are all not ideal things from a coding standpoint. I am not a coding expert, so I am not going to take us down a path on PHV, but it can help you with your coding and cleaning stuff up, which is a big asterisk here of if you're not used to writing coding or you don't know what you're doing or you don't know what you're looking at, maybe don't start there.

And then finally, data analysis with a big asterisk on ChatGPT and Bard specifically it. That data analysis is really text-based data. But a cool application, I did a customer survey with our clients and it was a mix of quantitative and qualitative. I copy and paste of the whole thing into ChatGPT and asked it to analyze it for me, and it did. It gave me advice on things that we could do to improve our customer satisfaction. It absolutely can do text data analysis and a little bit of data analysis, but there's a big note there. It's not really designed for big, high level, big data or quant.

Richard:

Perfect. Then I will jump in and take us through some different channel tactics. As we've played with the different tools and ChatGPT and Bard and some of the different applications of what it does well, what it doesn't do well, we put together just some different ideas on how, again, you can use it today and really leverage the tools to create those efficiencies and workflows. Certainly the number one piece is mastering your prompts. Going back to the earlier slide of inputs, this is a not good in, not good out to keep it family friendly exercise. As you look at prompts, it's really learning how to enter prompts, seeing what makes a difference in the final output, the more information you can provide.

There was actually a recent LinkedIn post from the New York University Dean of Students where he was giving direction on how students should be using AI as opposed to choosing evil and having a kick up their big research paper. In a good way, how can you converse with AI? How can you have it help with rewrites? How can you have it help with research? When you look at prompts, outlining what you want to accomplish and not being afraid to use that conversational aspect, which is really what this is all based on, being able to provide additional follow-up, provide examples to the AI to help get to the answer that you need or as close to the answer as you can possibly get.

And then of course, there's always that human element of providing feedback. As we look at learning models, of course, all of the feedback is going to make it better. I say all. Most of the feedback provided to AI engines is going to make it better and provide better answers later on as you continue to use it and continue to try new things. As an example of really looking at those prompts, there's a lot of inputs when you're thinking about trying to get to something like in this example email topics for a drip campaign. As you look through, hey, this is who we are, this is what we do, that's all pertinent information for the output.

But then saying, we want to write a three email drip series, and then also saying that we're looking for topics at the end of the day from the output as it relates to creative generation, that idea of getting us brainstorming. I think the number one thing that we hear from the team as they try it is really helps with writer's block, really helps with that moment where you sit down, you know you have to be creative in that moment and you just can't find the inspiration. AI and really working with prompts and trying to hone in that final messaging has been really beneficial in that process.

Then as a gift, I'll take you through some specific examples, but if you go to v9digital.com/ai, there is ability to download a resource that will give you the prompts that worked and give you some examples of prompts that get us closer to the final output. Because as you play with it and as you've had experience in ChatGPT and Bard is different. We'll save some time to show Bard for those who haven't been able to access it, but really understanding how to enter the input. I think the one big downside of this where integrations are going to help is that you get one bar and a submit button and you just never feel like you can include everything you want to say in there.

One of the things that is definitely true, you certainly can. You can put every prompt you could think of in that bar, but just the framework of how you interact with it and how you create the correct prompt to get to the final result is helpful. There's a resource available to everyone. As we looked at really in the SEO space, and I'm an old SEO guy, I've been doing it for 13 years. When I think about AI, I tend to not go as existential of, boy, this is going to do everything I know how to do. I look at it more from the perspective of, oh man, I don't have to write meta descriptions anymore? This is great.

But as we look at different tactics and different buckets, things like overall SEO strategy, where we have asterisks in the deck, it's where AI has proven to be more helpful and just provide better results. There are things where you can look at things like overall best practices for SEO. Do research if you're not as comfortable in SEO tactics or things that would benefit a client site or your site as you're working on it. It can provide some best practices, but really when you start to dive into those specific ideas, help setting goals, KPIs, personas as to what you're working against, there are definitely areas that AI currently does a lot better than say structuring your SEO strategy.

It does do, it can do it, and there are going to be some prompts for that. It's just a matter of overtime more inputs, great improvement, create a stronger ability to support in these areas.

Natalie:

As you go through the prompts, I hope everyone enjoys my pithy feedback on the ones that I thought were garbage versus the ones that I thought were decent. Just because it can do it doesn't mean it should do it.

Richard:

I think keywords is a good example. There's a lot of pushback around keyword research specifically and more the traditional sense of keyword research. We're looking at Keyword Planner or looking at Semrush data or whatever it may be as you're researching overall keyword opportunities. Topically, it does better. But if you ask for top thousand keywords and search volume, it basically is pulling from all these different resources, but you get a little bit of Semrush data and it doesn't all match up. And so then it starts to get a little bit weird.

But if you're looking for what are primary topics that I should focus on as a X company, it can give you some good insight into at least where to go from there as it relates to keywords.

Natalie:

Oh, go ahead.

Richard:

No, go ahead.

Natalie:

I'll give a little plug here to Clearscope because we really try to get the tool to give us semantic keywords and we tried it a few different ways. It's just nothing comparative to what you'll get through a tool like Clearscope. It'll give you something, but it's not going to be as in depth or robust or, frankly, as helpful.

Richard:

And then optimization, again, there's really an input-output piece here. But when you look at optimization recommendations, keyword insertion, content expansion, it will help with that. Again, I think a lot of it is that it serves as a nice guide, not necessarily a firm solution for those pieces. But some of the more technical pieces like generating schema, of course, there are already tools out there where you can generate your script if you're hard coding it and typing it out yourself. There are already tools for that. But also we have found that AI will generate that schema for you and generate it correctly.

Metadata caveat here is that Bard is not as great with metadata. Maybe it's because Google doesn't want to help us. Maybe that's it. We'll look at an example of that as well. But again, creating meta descriptions or creating titles, it'll give you pretty basic elements to those pieces now, especially in ChatGPT. There can be some helpful elements to create those efficiencies. As you look at content development, which is the area that I think everyone just naturally leans into because it's just so impressive when you see a thousand word blog in 10 seconds, but things like creative inspiration, super helpful.

Finding content topics, different titles, even messaging, get some different examples of messaging for those titles, creating outlines for topics. A good example is you have a post and you can say just, "Review it. Let me know if there's anything else I should add to it," or if you very explicitly ask it, "What questions should I be answering here as part of this content? Take a look at what I've already written," it can definitely help in that perspective.

And that example with NYU, rewriting copy or rewriting your paper, rewriting certain paragraphs, just taking that feedback and expounding upon it to look for areas of improvement in the existing copy that you either already wrote or you leveraged AI to write the first half and went through and adjusted it to your liking. And then local SEO far-reaching pieces for AI, certainly just in SEO, digital marketing, and the world. But local SEO it also touches on as far as optimizing your profile information, Q&As, business updates in post and helping create messaging that goes out through Google My Business, if they still call it that.

But really the idea is finding directories, which I know is here and there as far as how much people focus on directories these days, but very specific directories I know are still definitely a huge focus. I don't know that anyone's kicking down the door of city pages and white pages and all the many, many directories that we used to work with back in the day, but relevant directories to the industry and helping you identify those as far as where you can be listed. Definitely a lot of things where AI does a lot of the legwork, which I appreciate.

Especially on the research side when you would do it manually, having that automation certainly can assist to create those efficiencies and allow you to focus on larger, more complex type items. And then link earning or link building or we'll drop it in here just for fun, link juice. There are some pieces where it can help with the research of finding sites for links. One of the interesting pieces that I know we've tried out is different ideas for being newsworthy, different ideas for content that's going to have that appeal to be shared. We're in those links more naturally. Writing outreach emails, link earning idea is pretty generic at this point, but a lot of that is going to be based on certainly the inputs that it gets back.

There aren't really a ton of resources around here's how to have very specific link or any ideas for your niche. It's going to give you more broad best practice type information. But again, in the document that we'll share, there will be some different prompts to show you how to get to that information. But on the tech SEO, not super helpful. On the actual review of the site or looking at technical SEO best practices, it's not great, but things like creating hreflang script or creating a robust set .txt file, it certainly will do those things just to support. But your technical SEO audits are still a thing. It's not going to do that for you at this point. Maybe tomorrow.

Natalie:

It's manually writing robots.txt. Stop it.

Richard:

And then reporting. Similar to the overall sentiment analysis of data entry and text-based data, reporting certainly there are elements where AI has been proven to help in the early stages. Best practices for reporting, events set up ideas for a site, UTM creation. Of course, there are going to be tools and best practices for that as well, but there are some good pieces for for SEO reporting that you can take advantage of now to, again, create those efficiencies in the workflow. And then the part that I swear isn't meant to cause an existential crisis, but almost always does is, where is this headed? Where are we going?

Naturally, you see these applications that right out of the gate are super helpful in creating those efficiencies. But as you look at where this is headed and the note of this has changed since we finished these slides, this will change tomorrow, there will be new pieces, really taking in where this is going and the different language models and the true understanding of the landscape as to where it is now. ChatGPT, of course, is that one language model. They have multiple language models through OpenAI, but the one that obviously everyone's using is the predictive text-based. LaMDA is Google. BigScience, DeepMind, Stability.

Adobe is about to launch their own model, which won't be a language model naturally, but specifically for graphic design. There's new things developing every day. The big corporations that can put out the learning or AI are doing so, and then the tools that can leverage certainly the different language models are hopping on that train. It's moving fast.

Natalie:

And also throw a quick note in because I know there's a lot of questions in the chat about AI images. That last one, Stable Diffusion, is the language model that powers Canva's new text to image, which if you haven't used it yet, I like it. We've played around with a few different ones like the Jasper Art one and the OpenAI one, but that one's a really cool one and it's working really well. It's nice because it's there inside your Canva account. Turning that into corporate level graphics is really helpful.

Richard:

There's also the one that went really viral on social, DALL-E, which a lot of folks started using. I think on the image side, it's really interesting for applications because we initially looked at it like, man, that's great. We could get really creative, really beautiful images. But the integration with frameworks is where images is really likely to take off. That idea that you could put in specific colors and a logo and then your prompts are just based on what you're looking for as the output will be super helpful. Right now, those models, you can look at professional and artistic and you get some different prompts, but a little bit more control over the output is...

I was joking with our social team that I feel like that's two months out. I mean, it could be two hours out as fast as this is moving for more efficient outputs for images or more control over those final pieces. Plugins is a big piece. Of course, the quote from Zapier early on in the presentation, Zapier announced hundreds of integration capabilities with ChatGPT across different tools. The online plugin component of this and really the application for ChatGPT is moving at a quick pace as well. The ability to integrate it into a website, integrate it into more tools and really leverage the full breadth of its capabilities is on the very, very near horizon and already happening in many ways.

Then a little bit boring part, but kind of interesting, the legal ramification. One of the biggest things, and I'll show you an example as far as what Bard thinks of the legal ramifications, but the big issue here is when we look at the future of AI and especially the ability to create content, the ability to create imagery based on prompts from other artists, the intellectual property component of this, in that Bard very rarely or ChatGPT very rarely will give you a source for where it found that information and just explains it as we found it in a lot of places. Stop asking.

And then of course, in the EU, data protection and privacy, which naturally is becoming a bigger focus in the US, and surveillance, which, of course, is becoming a bigger factor in the US, but as we think about the different impacts here, even going beyond that, there's the constant debate that... I mean, I might be the only one who watches it because I find it interesting, but Section 230 of, does this make Google a publisher? Does this now mean that Google is responsible for the content that their AI creates as opposed to just a directory of websites where the website is liable for that content?

If they're sourcing it and they're creating and pushing content, does that make them a publisher, and thus, liable to the potential for incorrect answers, which, of course, are going to be rampant early on? And then the ethics of it. One of the big pieces in the image space that first happened was when DALL-E first launched, there were a lot of questions about, is this ethical? If you say make me an image based on the stylings of Van Gogh, is that ethical to then use that image for business, to use that image for profit, to publish it on your website? Is it a breach of that intellectual property? Is it right that you have that ability?

When we look at the overall ethics, these are far ranging and this is usually where the existential crisis comes in, but the job displacement. There is a huge AI economy that has been stood up in the last month and a half where you have companies that are leveraging it to increase profitability. You have the companies that are creating tools to try to combat it, especially at the school level. I think in conversations recently, I was pointing out that especially at schools, that's the one that makes it on the news are students using AI to create their research paper.

Again, to that idea of this being a seismic ship, the same conversations happened when Yahoo and Google became a mainstay in our lives. Does that remove critical thinking? I feel like that was still a debate when AI launched of, is the ability that everything on the web is accessible at our fingertips bad for critical thinking? Is it bad for really the society as a whole? AI is just having that same conversation as most new technological advancements will have. As we look at misinformation, fake content, that's a concern.

Certainly I think the biggest thing that is important to consider, especially with Bard, just because Bard does access Google search results, is that the common joke is, oh sure, everything on the web is true. I know it's true because I saw it on the internet. It's important to consider that these AI tools are reading all those things too, and they are processing those things as well. The question of is this going to be the thing that replaces search engines, guaranteeing accuracy based on the fact that the internet's not a super accurate place and when Bard opens up to millions of people and people aren't the most reliable source of information all the time, what is it learning?

When you ask those questions later on, what information is it giving back? There's certainly a lot of pieces that crop up as you think about AI as a whole and how we integrate it naturally, as we look at channel tactics, as we look at personally when we're using AI to write those meta descriptions. It's super helpful now, but of course, there's this bigger conversation that goes into it. I think the easy applications in digital specifically are in those areas that whether you're talking to clients or you're talking to your team internally of how do we get these things done that no one wants to do, write four blogs a month about garage doors, a thousand word guide step by step.

Leveraging it in those spaces is where it's super exciting. I think when you think about job displacement, there was a quote from Hank Green who said his freelancers and his contractors were reaching out of, "Are you going to use this to replace me?" He said, "No, but you should think about what you do better." That's the piece, right? What do you do better than the bot at this point? There's a lot that the bot can't do. Where it goes from here is anyone's guess or hypothesis. Privacy and security, of course, is going to be at the forefront of this conversation as well, as it is with anything in the online space.

What to think about without the existential crisis, if anyone went into an existential crisis, I'm sorry, but there is the potential that you lose unbranded SEO or paid search traffic. The idea that if someone can just use Bard to ask a question, get the answer, the same things that happened with featured snippets in search, that call out at the beginning where it takes content from the page and places that. It had already answered the question. But user behavior dictated that that actually got more traffic when they did that and people did click on those search results. User behavior is going to drive a lot of this.

Building up the brand and notoriety around the brand. Because naturally if you lose informational type searches and there isn't as much traffic to go around, building a strong brand that people can rely on and go to is going to be a important piece. Building your first party data or really that ability to control a bit more of your marketing. If we're not able to just as easily attract people on informational posts or informational type content that they then lean on AI for, having that first party data is going to be essential moving forward as it relates to your marketing efforts, as your ability to connect directly with consumers.

And then to that earlier point, up-leveling the skills, how can you leverage AI to make your work more efficient, more valuable? What do you do better than AI? Even in those examples of the initial tests of trying to get it to do certain things within the SEO or digital marketing space, there are areas where we know we do better. Leveraging those areas, but not being afraid of AI. I think I probably am more in the it's inevitable and let's get on board camp than even Natalie is. But I think there is a component to this that there is so much investment, there's so much going on with AI right now that it's a freight train.

Best not to stand on the tracks, but see if you can catch a ride. I took that metaphor all over to the end. Resources to help you keep momentum and then we'll look at some quick examples and leave some time for questions, but different tools certainly. Canva recently integrated the AI components, but Fathom is another tool that we used for call recordings and transcripts to summarize our meetings. It's free because venture capitalists don't care about making money these days. Check it out. And then of course, ChatGPT and Bard. I'll show you Bard.

There's a lot of different tools that can be super helpful that already integrate some component of AI or integrate some component of complex data analysis or content analysis into your workflows.

Natalie:

I'll make a note too because it's becoming my new pet peeve of someone sharing a list of a hundred AI tools you should check out. I'm like, oh cool. You click on one and the first one's garbage and the second one is pretty garbage, and then you're signing up for accounts and you're trying it because someone recommended this is a great video account and it's terrible. These are all different ones that we've actually tried and used. I would say Fathom is one of our new favorite ones that we're using a lot because it's recording our Zoom meetings and then summarizing them for us and then highlighting key takeaways for us. That's been great.

Richard:

And then some different sites just to stay up to date on AI and the changes that are happening, additional resources there. If you need a hand, we did include some contact information and connect and engage. I am going to switch over real quick just to show you some Bard examples because I know that is not a tool that everyone has access to at this point, but I thought there were some interesting pieces, so I seeded a couple different tests of interacting with Bard. It is very similar in the framework to ChatGPT, but things like if you're not as familiar, how long should a blog post be, which of course, everyone always asks, because again, no one wants to write them and they're hoping you say 100 words.

How do I make sure it shows up in search? It provides an answer for that. This is really that idea of keeping it conversational, where if you can't think of everything to prompt it with in the first piece, you can just keep going. What if I'm writing about a niche B2B topic with a small audience, does that make a huge difference? It provides resources and information. Other examples that I pulled were more in the weeds with Bard of, are you going to turn into a search engine? How can you ensure the accuracy of information provided? And then getting it to explain how it understands Section 230, which is super interesting.

I think the big piece down here of Bard may display inaccurate or offensive information that doesn't represent Google's views, this is 100% that case. I don't think anyone at Google would ever want to say this as it relates to the potential implications of AI. But having those conversations I think it is very easy. We were joking before the call that I got access to Bard a few weeks ago and just started testing prompts. And then the next thing you know, I'm calling the AIU. I'm having a full-blown conversation, just keep it going, to the point where Bard sincerely apologized to me about a source they provided me.

It's just really interesting to play with, especially in that conversational space. I did have a couple examples of where Bard is will say better for today, but the current information is where Bard is going to be a little bit better. ChatGPT's database obviously ended or had a time limit to it, where Bard is currently accessing search results. You're able to ask things like what happened yesterday, or better for current events, things that are currently happening that have that understanding. And then where Bard has seen challenges were initially you were able to put in a link and have it look at that link and give you information based on it.

In the past week or so, that functionality has turned into, sorry, I can't help you. But if you reframed that prompt to say, "Can you give me a page title and meta description for the homepage of this brand," it'll do I. Where ChatGPT, you can certainly paste in 10 links and have it kick out this information. Of course, you can see this very basic optimization. But the other thing that Bard will do say going back to this example is for every prompt it'll give you three different variations. There are some pieces here where one thing that I noticed, especially in contact production, is that the length of article you tell it to write can impact the quality of the article.

If you say that I'm looking for a blog of 1,000 words and you find that it's repeating itself, very similar to when you ask a writer to write a blog for 1,000 words and they're trying to hit that number. Even adjusting that to say 750, 500 can give you different variations and also seeing the drafts of different options on how it chooses to approach that topic. Jump back over to this to get to the thank you slide. We landed it with 10 minutes to spare.

Travis:

Awesome. Great job. Appreciate that. We did have a couple questions. If anyone else has additional questions, drop them in the Q&A box. We have about 10 minutes left. The first one is from Sarah. Can you recommend a good free course on AI?

Richard:

I don't have a good free course, but some of the different resources, I think a lot of it is going to be reading material, but Natalie, if you have any?

Natalie:

I don't. This is back to my pet peeve. We've had a lot of AI experts pop up overnight, a lot of paid courses, a lot of different information, even people publishing information that is just wrong. I don't know that I have one off the top of my head. I think when we went on our journey to figure out AI, we got really frustrated really fast with bad information, and so we just started to dig in and really figure out what we liked. But right now we seem to be swimming, at least in the SEO side, in a world of, hey, I found out one cool thing I could do with ChatGPT, take a look. We're up to our eyeballs and content like that, or people saying things like, "Hey, ChatGPT can do keyword research."

Just because it can do it doesn't mean it should do it. We're not getting that much of a differentiation. I think everybody's just in this like "oh my God, look what I got it to do" mode. We're trying to weed through that as much as we can. I'm sorry, that's not a great answer. I did see one come through my inbox that was a very expensive training, but it seemed really intense. Someone like teaching you how to program tools in AI. I can dig that out and send that out, but I don't think I have a good one. Plus, it's also good to note that any course that gets created is going to be outdated real fast.

Even as Richard articulated, these tools are updated regularly. Even though they're like, oh this is the official GPT-4 iteration, they even have a note that they're constantly working on it, and it's doing things differently than it did a week ago or two weeks ago. It's going to be really hard to have a real up-to-date course.

Travis:

Yeah, that's a good point. And then another question is more on the publisher side, but what are your thoughts on sites blocking crawlers used by the AI tools and ChatGPT specifically?

Richard:

I would say there is always going to be an urge to fight back and say, "I don't want you to use my information as a source for AI generated answers," or whatever it may be. Blocking crawlers or tools to try to say this is definitely AI or I think the tools generally say this is 80% AI when you put them in. There's an economy to that. There's a concept behind it, I would say. Whether you block AI and decide not to play in the space, the space is going to happen around you. I think there's a component of users are going to use AI, the idea of blocking it.

Unless everyone blocked it, I think the big piece will be that users and businesses will decide how far this goes, right? From an application perspective and if people say, "This makes me uncomfy and we're going to block AI from being able to access information on our site," you can certainly go that route. But if you have someone who's researching your business using AI very similar to SEO, you want to be there.

Natalie:

I mean too, I also think about, okay, so you're going to block AI, but I mean, we've all seen these websites that pop up overseas that are basically duplicates of your site. We know Google doesn't rank them, but then the content then exists. How do you prevent that from happening and then the AI not accessing that site? And then how do you start doing a good job of blocking specific AI bots without damaging your brand by blocking something like Bing or Google, which if you're a publisher is going to be how you get paid?

If you can't get visibility from major searching, just direct traffic to your publishing site, then why do you have a publishing site? It's an interesting, I guess, intellectual conversation, but I'm just not actually sure that that's possible.

Travis:

Makes sense. And then another question or a question from Katie, is there a graphic AI for creating maps? Have you seen any of those?

Natalie:

I don't know that I have. Have you, Richard?

Richard:

I would bet a lot of those image sites would at least try. At this point, I don't know that there's a specific one that that'll spit out maps right now. Again, it could have happened while we were talking.

Travis:

That's very true.

Richard:

But I think those image components with prompts, I would bet that it would at least try to kick on a map for you.

Natalie:

I think in general the AI for image generation, it's good, but it's not amazing.

Richard:

They can't do human hands, which I think is hilarious.

Natalie:

I only have one hand and it's over here, but I think it's getting there fast, and I think that's part of the point of leaning in. All those different language models we shared are all just designed to do different things. The image stuff's getting really good and they're developing it every day. Canva launched something, then Adobe launched something, and then there's just all these different... One of those language models was just designed to get the AI to multitask. It was playing Atari while doing a math problem. You just start thinking about all these different language models as they continue to adapt. It might not be weird as today, but it could be later today or shortly in the future.

Travis:

Yeah, yeah, it's a good point. This could be the last question. It's not directly related to AI, but it's interesting and it corresponds and correlates y'all's expertise. But what are your thoughts on LinkedIn collaborative articles? They're working well for them, but seem risky as an SEO content strategy in some spaces like medical, for example?

Natalie:

In general, I would say AI generated content, especially stuff that's generated through ChatGPT, is a risk for SEO. I wrote a whole article on our site about that. I could send that to you too, Travis, for the follow-up. A lot of it's conjecture. I'm connecting some dots. But in general, you have to imagine that Google's now launched an AI tool. It wants you to use Bard. They don't want you to use ChatGPT. We have a tool like Originality.AI where a few guys got together and figured out or already figured out if it's GPT-3.5 content.

It's not going to be that much harder to figure out, okay, this content was GPT-4, whatever the next version was, and that Google's on top of that because they don't want you using a competitive product. My argument is that there's a little bit of inherent risk of just throwing something into an AI tool, getting it to generate your content. CNET found this when they were using Jasper.AI content. And not only was it factually incorrect, they also hit a bunch of copyright violations. If you look at their rank profile in a tool like Semrush, you could see that they just tanked the site. They are a publisher, so that was a big deal for them.

You've got the copyright issues. You've got the data accuracy issues, and then you just have the general idea of you just want to be really careful about ChatGPT specifically writing your content. It doesn't mean you have to. My conjecture is in general, be careful from an SEO perspective, but a LinkedIn article is a little bit different. You don't care as much about Google algorithms when you're writing LinkedIn articles. You care more about connecting with your LinkedIn audience. I think that those are really great with the big asterisks that what plays well in LinkedIn is thought leadership and these tools aren't quite there in terms of giving you thought leadership.

I think they could support a thought leadership type piece, but I don't think that they could have really cool, unique, thoughtful thoughts for you. But I think if you're meshing your critical thinking along with the tools, I think LinkedIn articles is one of the many places that this content makes a ton of sense. In general, wherever you're publishing content and especially if you're a medical or legal or even for us for SEO, we feel a lot more comfortable writing content through Bard and ChatGPT because we understand what we're reading. If it gives me a bad SEO strategy, I know what I'm looking at right away.

If it gives me a terrible meta description, I know what I'm looking for. It's kind of that same thing of you'd want someone who knew what they were looking for in legal content or medical content, double checking, being like, yes, they defined that procedure accurately. Yes, that's a medical journal we would link to, not a crazy one. I mean, I think in general, you just have to be really careful because it'll create something, it'll do something for you. If you don't know what you're looking at, I think you could have really big inherent brand reputation problems wherever you publish that content, whether it's on LinkedIn or your site or social or wherever.

Travis:

Awesome. Excellent point. That's a good one to end on too. Thank you so much for your time, Natalie and Richard. Any last parting words before we give everyone their day back?

Natalie:

I hope everybody is inspired to go try something new. I hope whether it was our prompt sheet or an idea we came up with that you go try something new. If you have follow up questions, I know because I'm seeing tons, people had tons of questions throughout, so if you just have a burning question still after this, just ping us at this email address, we'll get them answered.

But maybe to give a little hope, one of the cool things that AI is doing is recently there's some really cool artificial intelligence in the medical space. Recently, artificial intelligence was used to predict a woman's breast cancer four years before we could have detected it with our current equipment and technology. It's not all horrible. There's some really great...

Richard:

And it's not sentient. Don't believe it. Don't fall for it. I was joking with someone yesterday that predictive text has been on your cell phone forever, when you're seeing those tweets that say start a text message with this and then autocomplete it. That is what it is doing. It is doing it very well. It is doing it in a way that can make people uncomfy, but it's not doing anything until we tell it to. It's not figuring out the world's problems in the background and then going to take over.