Alex says: “Do not let AI take over your content. Instead, be creative, helpful, and unique, and your storytelling will always shine through the mediocre efforts of others.”

What's the reason that you don't want to give AI full control over your content creation?

“Let's say I've got my brand and I am the founder of it. I'm the most passionate person about that business, that brand, and that product or service, so it stands to reason that I wouldn't want a machine, robot, or an LLM, at least now, to take over that work.

There are a couple of reasons why robotic LLMs are, in fact, robotic. Sometimes, when I read something, I know when it's AI content. So if it is AI content, but I'm being told it's from a person, I will think I am being duped and trust you and your brand less.

The other element is scaling content, which is quite a no. Again, if you're producing content primarily for the search engine or to get attention or visibility in the SERPs, in the long-term that will not happen for you. It might work in the short term, and I'm sure you'll see many people on LinkedIn and X showing an AI-powered niche site that they've got to show how well it's performing, but you should check back in three months and ask that person again how it's doing, because it probably won't be very well.

Then you go back to human-produced content. There's nothing more creative than the human brain, especially collaborative human brains, to create actual thought leadership pieces that are creative, informative, and primarily for the user, and not just for the search engine and the search results. As a result, that content will shine through because it will be deemed more helpful by the user. Therefore, it will be looked at favorably within Google's search rankings or systems, whatever you want to call it.”

One of the phrases you used there was that, at this point, the LLMs are producing a fairly robotic output. How long are we looking at it before passing the baton to the AI to create all the content we want?

“I guess that's two questions in a way. The first question is: how long will it take for the LLMs to be that much better that you cannot tell whatsoever whether it's a human or a robot? My answer would be within five years, because if these LLMs are going to be cleverer than every human on the planet, then I think they're going to be clever enough to be able to figure out content writing in the same amount of time. If they're that intelligent, which they will be.

The other question would be, should you move to letting AI take over your content? And even then, I'd say no. If you can help it, never do it. It should always be human output in a way that humans will always feel duped if they buy a book or a novel and read through it and think it is great. Afterward, you will find out that it was all created by AI. I feel psychologically as though I got done over because a robot led me through the story, there may have been a great story. But I feel like the human producing that story in the state of the art of storytelling will still be owned by the human for a long time.

I believe their greatest interpretation is what we've done thus far. Still, I don't think 40 years ago, if they existed, they would have created the next Tarantino in filmmaking, Tarantino, and only his mind could create that thing that created a new genre and influenced many other filmmakers after him. That's just one example of how human eyes can't do what AI can do. They can't make true creativity yet.”

LLMs are trained based on what humans have done in the past. The challenge is to get that creativity and come up with something that hasn't been done before.

“Yeah, they're trying to mimic us, but they aren't us, and unfortunately for them, no matter how much they have, maybe a mind or consciousness, they're never going to be us, they're never going to be the human brain. They can't replicate it exactly.

They can only keep guessing because that's what LLMs are - they're guessing the correct answer - and generally, they're right. That's all they can do: interpret what we are as a species or through our minds, and therefore, that's why it's so hard to regenerate content that would been done by humans because they can't get that 1% of what makes us.”

What's a good rule for incorporating AI into content creation?

“As I always say, think of it as your assistant. It can help in many ways, from ideation to stopping writer's block, creating summaries, or helping expand on something. There’s another tool called Perplexity that's another kind of like, ChatGPT, but it outputs as an experience, and it's very good for more academic research, that kind of stuff.

So, if you were looking at American history, for example, it would cite sources in a way that an SEO will love because it's bringing something from the web. It also gives you further reading. But what it's doing is complementing your thought process. That's what it should be doing and that's what you should be doing. It should be helping you lead your way. You should still be the decision maker, and you should still be the editor when you publish anything.

AI can also help out your day organization, that kind of thing. It helps you be your project manager. But when it comes to content creation, ideation, help with your creativity, and maybe thinking about things you may not have thought of by reading what you've written so far and asking what else you have not thought of thus far that maybe your users are going to find helpful?”

How do you know if you're using AI too much now? Is it all down to the quality of the content and perhaps the lack of individual opinion quotes or elements from thought leaders within your organization that aren't incorporated in your content?

“There will always be a ceiling that you'll hit to argue about capacity. And if you're scaling content to the degree that you're not going to get a team to bring out well-thought-out pieces, like you say, from citations from other sources or with input from other humans, it's always going to be hard to scale that content.

In addition, Google is cutting down and putting that in our spam policies, so you shouldn't be scaling content or creating content for content's sake, which has always been the ethos. We're going to go back to the original webmaster guidelines. Again, it's not veered away from its original guidelines that were made 20-odd years ago. Don't try to abuse the system; don't try to do something to be more visible on search engines because, in the long term, it will be much harder.

From what we've seen, if you keep on doing that, it won't just hurt those pages; it will hurt your entire domain, and it will hurt it until you rectify it. The worse you create the problem, the bigger the hole you'll have to dig yourself out of in X amount of months or years when Google finally fills whatever hole you're trying to fill. Some SEOs try to go through a loophole. Instead, don't think about the loopholes. Just do as they're saying, and you'll probably find that what you will be doing naturally to try and grow the visibility is the best and most natural way to do things, anyway.”

Can you teach AI to be ‘on brand’ and speak in a relevant and suitable manner for your organization?

“I think so. But again, it's not creating its information. It has to interpret what you've done so far. It can maybe take posts that I've written in the past and understand how I write, but again, I'm still going to check that to make sure that is what I wrote.

From a Yoast point of view, I could say check out yoast.com, read it all, ingest it, and understand the tone of voice and here's what we've got from our notes on brand guidelines for writing and editorial. Use that and adopt that when you then create this bit of content.

In the next five years, that will become much better. But as time passes, you must also think and be specific in some of these algorithms. You have to be specific because the vaguer you are, the bigger the chances of hallucination; therefore, there's more editorial. By the end, you'll have spent 80 percent of the time trying to make AI deal with this piece of code; I could have just written it myself.

At that point, it would have again been more natural had the AI platform just helped you do your job better rather than delegating that job to the machine that will make your content come out a bit more than everyone else who may be doing scaled AI content as well. Remember, mediocrity will become more common as time goes on.

As well as that, real estate on SERPs is going to be harder where that is the difference and SGE will work exactly how it's been showing. It's not ten links anymore, now it's two or three. That means that your fight to get onto those rankings on those three positions is much harder than being positioned eight on the tenth. It's therefore going to be more valuable to do more unique and natural SEO, that's helpful for the user, which again, AI is going not to compliment as much as if you're leading it.”

Mediocracy is becoming more common. I love that saying, but I think it's motivating. It shows you, as a brand marketer, that if you do something a little bit different, you still have the opportunity to bring in that organic traffic because maybe other people aren't trying as hard. They're just using AI a little bit too much and just being the same as everyone else as well.

I also liked your advice about brand guidelines and having content that talks about what is right for your brand. Normally, you would have that brand guideline as an internal resource document. But if you have that online, some orphan page, then you can give that LLM access to that or not, and hopefully, better guide them towards writing on behalf of what you're trying to achieve.

“Yeah, exactly. It will get better to do that. If an LLM is writing something, I wouldn't just hit publish. I know that there have been platforms that offer that and they pulled those things back, or they're not scaling it as much as they were six months or nine months ago, and that's because I think people are saying that maybe you can't just say write ten pieces of content for me every single day and connect it via install whatever you're doing to publish something automatically.

But what's the point when the bigger the brand, the more at stake for your domain’s authoritativeness? If I were a big brand, I would never sacrifice the whole sites authority for the sake of being lazy, which is what you may be doing. I think listening to AI makes your content lazy.

You can't complain that AI is taking your jobs in two years when you've delegated that job to AI. There's got to be a limit to what you do. You have to be the best at using AI to your advantage rather than letting AI do your job because it will take it over in that case.”

Is them the right pronoun for an LLM?

“I don't know, but I think it's a thing. Because it isn't alive, it isn't a biological form. It's a machine. It's it. That's all I take from it. But then, when I say they or them, I think that LLMs are a collective of many different things, like synthetic brains. So, if you know what I mean, I think of them as an audience, not one element.

You can ask it two very different questions, instantly giving you the answer as if it's the same person. Right? That's what you're conversing about. So, you can ask about creative content and then tell it to diagnose PHP errors in the same conversation. So, it's like having many minds to help you out. Which is why I say that.”

Your general advice is to be more creative, helpful, and unique in your storytelling. How do you measure its impact? How do you decide that you have achieved some form of success in doing that?

“Brand visibility. Don't just think of SERPs and search results and now I'm ranking higher and ranking for more things. I've been saying that a lot, but especially more recently, it's not just SEO anymore. In my opinion, SEO is an element of a whole search experience. The ranking is just one element. Ranking for a keyword is one element of the tough battle.

You have to get a conversion or whatever the goal is for you, whether that's collecting an email address, buying products, contacting you, or whatever the event or the CTA or the conversion may be. I think those are the best things to measure, but also the brand recognition and the people within it.

If we talked about authority and EEAT, which isn't a ranking factor, I still think it's really important for people. The success of social platforms and visits around them is as important as generic organic visits from the SERP. So, think wider on the KPIs. I think about it from even The SEO Update that we do at Yoast; we do it every month, and it's going up every month.

And I'd love to attribute all that success to me or Carolyn as a host. But I am also aware that other variables are at play, like Neringa, who sorts out all the email marketing for it; you know, she's doing an improved job over the months. So is it that? Or is it a combination of all of these things? Is it what we're covering? There are so many things that you can't measure what that growth is.

But as long as you are a part of the bigger team looking at the growth together, it's good, and you should be working more. You shouldn't just be isolated in SEO; you should get into the marketing and product teams and have input because it is important. If the leadership doesn't think you're important.

You have to remind yourself you are being paid to be there to have that knowledge. And it should be partaking by others because you do know more than the people above you and because they're concentrating on other elements of the business. Our specialization is the search experience in general. And it's not just an engine we need to optimize or just one engine we need to optimize. It's moving forward; you've got things like TikTok, which some people consider a search engine. I still call it a social platform, a very popular social platform, but a social platform nevertheless.

But if people are talking about TikTok SEO, those two words don't go hand in hand; it's TikTok optimization because that's the platform. And I think that's how it'll go wider and wider. As more things come out, more AI tools are used, and it'll be more important, I say, to have links to things like schema and connect all of these entities. That will be important for LLMs because structured rather than flat data will be easier to read.”

You shared what SEO should be doing in 2024. Now, let's talk about what SEO shouldn't be doing. What's seductive in terms of time but ultimately counterproductive? What's something that SEO shouldn't be doing in 2024?

“Content automation is kind of what I was insinuating before. Don't put your site into cruise control in any way. If you're automating it, be clever and do something creative around it. But don't just let an LLM connect to your WordPress site, for example, and post daily by asking it to talk about X based on our tone of voice and what we sell in the brand and then publish it.

I would say write less. Less is more at the moment. If you have a great thought leadership piece that covers a specific topic, concentrate on that instead of getting five blogs out this month because you'll do that later. KPI is for the search engine, and the format is for users. And the first will cement your brand. It will give you more chances for links, give you more authority, and therefore, give you more chances of being cited in things like an LLM.”

Alex Moss is Principal SEO at Yoast and Co-Founder at FireCask, and you can find him at alex-moss.co.uk.