Julia McCoy says “It's 2023 and AI is here. I can't believe I'm saying that, because I saw what happened with “Article Spinners” back in 2015, and I was just horrified and I thought that AI had no place in SEO writing and quality content. But now GPT-4 is here and I'm actually the president at an AI company called Content at Scale, so I'm fully in! My tip in using SEO in 2023 is to definitely use AI in your process. For some context, I used to run an SEO content agency for 10 years, and we were doing 10 million words a year. But now I'm in the mix of over 40 million words of AI production per month. So that's four years of content in one month, because of AI.

So what I see is that you can't just publish content straight from AI. There is a misconception, especially in new comers to SEO, that I can just go to ChatGPT or any AI writer, and say, ‘write a 500-word blog post on this keyword’, and then go take it and publish it. But that's not something you can ever do for many reasons.

First of all, that content might not even rank because it doesn't have the qualities Google is looking for. Google is all about EEAT (there’s a new E now which Google recently announced) - Experience, Expertise, Authority and Trust. You have to embed those in your content, which comes back to the human touch. So on one hand, yes, embrace AI, because we're seeing anywhere from 5x to 25x the reduction on time and cost that it would typically take to do human-only content creation, which is crazy. So by using AI, you're able to get way more profitable and efficient.

But what I'm teaching a Content at Scale is something called AIO, which is where you're taking content from AI and you're humanly optimizing it. So the humanly optimized process is built around a framework we call CRAFT, which just five simple letters, and it takes you through the process of humanly making that content just 10 times better, which you have to do whenever it comes from AI.

So the first step is to ‘cut the fluff’ and ruthlessly edit AI, which it's famous for repeating the same thing over and over which it's funny, because that's what a novice writer would do, as well. So we have the same problems with AI that we did when starting out with human writing. You’ve got to ruthlessly edit.”

Is the fluff going to be there moving forward, or is AI not getting better and better, so the fluff is getting less and less?

“I think it is getting less and less, and our dream with what we're actually building at Content at Scale is that you hit a button, and the fluff is removed. So yeah, that's a very real possibility. All steps in the CRAFT framework could be removed, except for the last one, which is trust building. And that has to come from your personal experience, because ChatGPT and AI in general will never know your human experience. So that last step of trust building, I think we can eliminate all the other ones eventually, and have AI do it.

But when it comes to building trust, which is like the biggest reason why people buy, subscribe, move forward. it's your most important element, and that has to come from personal experience. It's like what did David go through to achieve what he's achieving today? I want to know that story. But AI is not going to be us and it will say something completely false. Unless you feed it your story, which means you’ve got to train it.”

So at the moment, humans have to be involved a little bit more. And your second step in CRAFT is reviewing and optimizing SEO to be as human as possible. So why doesn’t AI content sound like a human at the moment, and what do humans actually have to do to make it sound more human?

“So what we're seeing with the AI output is it's going to get better and better. But at the moment, what we are seeing is a lot of keyword stuffing or just awkward ways the AI is saying the keyword that you would never want to read. You can see when its keyword stuff just for a backlink. So that's not human, so you've got to look at those keywords and ask if they are written naturally. And this is something that I work with our writers on all the time. Naturally written keywords are the end goal, which is what Google prefers too, not just us humans. So that's the second step (in CRAFT) - reviewing how those keywords read and how they're put in there, e.g. is it stuffed or is it natural, because we want to make it natural.”

The ‘A’ if CRAFT is adding visuals and media to tell the story with pictures as well. Do tables and data within cells actually form part of what humans can add as well, was that something that you rely on AI for?

“I think the data is definitely something that you want a human to either lead or assist. There's a lot of AI tools out there that can either put together data or even go through it and give you a summary, even ChatGPT can do that. So I'll ask it to look at the top books on Amazon in marketing, and give me the top 10 negative things people are saying. That used to be manual work, and it's all done now. So the research side of tasks is something which you can actually delegate to AI.

But when it comes to the final things that we know from the data, that's still a very subjective, human, personal experience type of thing. How to present that data is something that I don't see AI replacing anytime soon, because the more innovative you get with presenting your data, and the more innovative it is, the more it stands out. And that's just that's going in the human lane again. So data, yes, you can have AI do the legwork of it, but to present it, you definitely need a human that knows what they're doing.”

Images can be generated using AI now as well, is that something you're actively doing?

“We’re testing the waters of that and we're pushing the limits of what it can do. It's funny, because time and time again, we come back to the human designer, because if you’ve seen it, Midjourney is notorious for adding all the extra hands and fingers, and it just ruins the image and you can't fix it, and we end up going back to using a picture of a real human.

So I don’t think we are there yet with images, and that's where we keep coming back to over and over, whether it's a blog header or a visual data piece that we want to put into the blog to drive the point home. So our illustrators and our designers have gone nowhere, in fact they've probably gotten more work than they usually get, because AI imagery is just not there yet.”

Moving forward, it might be a little bit more challenging for visual search engines to identify which images have been generated by AI and which haven't, so is that something that should be a concern for search engines, and also people that want to use images on their sites?

“I think with OpenAI announcing, they're going to be watermarking the written content, that's the one that's been first to market, and imagery has definitely come second. It seems like it's harder to get AI images to market than it is the written word. So we have AI writing content better than a human writer, but we also have images that are not designed as well as a human designer. But it's going to get there, I believe it's just a matter of time.

I think the watermarking on the written word has created a standard that's going to follow with images. So we're going to see images potentially watermarked as well, by Midjourney or whichever the AI will be, and then that will give Google a benchmark to decide was this humanly creative, or was it AI generated? I don't know if that will harm your rankings because Google has really gone back and forth on whether they like that or not, and right now it seems like they like it, but basically what they're saying is just don't spam people.

At the end of the day it comes down to testing. We're testing pages all the time seeing if it's going to rank putting out there, and what we see is that if you implement this CRAFT framework, if you go through these steps and make sure the content is accurate, if it's true to your personal experience, you've added the right visuals, the right things, the right elements, you've chopped off the fluff, then that content will rank, and it will convert, and it will succeed. You can't not apply the human touch, and I think that will separate you from the ‘watermarking potential danger zone’ as well, because you don't want to be in that danger zone as a brand.”

Going back to your framework, your F is for fact checking and obviously AI can spit out content that sounds very definitive, but isn't necessarily completely 100% correct. So do you absolutely have to check every single fact? Or is there any particular type of facts that AI tends to spit out that is less likely to be accurate?

“There’s definitely a divide there, and what drew me to highlighting the factual accuracy was something OpenAI actually says on their blog which is that GPT-4 /GPT-3, have no known source of truth. So they're essentially pulling billions of data points, and truth is not a meter, it doesn't matter. So if we know that, then we should know that we have to fact check this stuff, and we can’t let it fly without that. What I'm seeing is the newer the technology, the more you need to fact check how it's described. So for example, it'll say that an AI detector is a security measure, but it's not. It's a way to classify content. But GPT-4 doesn't know that yet, because it's not as current as it should be. So anything that's new and innovative, you better fact check the heck out of it. And this also includes older facts as well, such as ‘when was SEO coined?’, because I found that our tool Content at Scale wrote accurate content, it found the correct date, found the correct time it was first mentioned, and all of that was correct. But I verified it. So even though the facts that were correct in this instance, you still want to verify it.”

And T is trust building with personal style. But is it not the case that you can actually ask AI to talk in a certain manner in a certain style and retain your existing style?

“You can't do that off the bat do that right now, but I'm close to some experts that are actually training ChatGPT on their personal style. But what I'm seeing is that to achieve that certain tone style, from ChatGPT, they're spending hundreds of hours on it. The question should really be - do we want to do that? Or do we want to do some human editing for one hour instead?”

I was referring more to content based upon famous personalities, because obviously you can ask it to mimic maybe Donald Trump or some particularly famous person in terms of writing style or speaking style. But I guess if you've got a honed bespoke style for your agency, and it's not obviously based upon an individual, then that's more challenging to train AI on you're saying?

“Yes, and that's what we're seeing the need for this is. Let’s say you’re an influencer and you have a known following – then how do you get this AI to write like you? I think we're approaching the point where you can type in ‘Amy Porterfield’ and the AI will know your voice. We're not there yet, but we're almost there, because it knows you sound like Donald Trump or The Rock. So it knows super famous people, but it’s not there yet for the influencers, which would be a lot more practical, because who's going to write like The Rock on their personal brand? Some people might.

What we want to know is how do we train the AI to know our style. So at Content at Scale, it's actually something the founder and the dev team is working on, having a tool to learn the style of these influencers, get to know them, crawl their content, learn their style, and then replicate it for each new piece of content. I think it's just a matter of time until we get there.”

So you've shares what SEOs should be doing in 2023. Now let's talk about what SEOs shouldn't be doing. So what's something that's productive in terms of time, but ultimately counterproductive? What's something that SEOs shouldn't be doing in 2023?

“I would say don't overthink it when it comes to production. I've been on many coaching calls in the early months of 2023, and what I'm seeing is a lot of people that are stepping into SEO and doing content at scale. We’ve got to face the fact that we have to do a lot more content than just a few years ago, because we’ve got to actually compete. So in order to do that, we’ve got to do more than one blog post a week. We have to do a couple, maybe three, maybe four.

So what I'm seeing is people stepping into this new age of doing content and SEO, and they're pausing to get every little thing right before they hit publish. Overthinking it, which I'm guilty of too, can be such an obstacle to your traffic growth and your revenue success. So you have to execute and get that content out. If you're using AI, make sure you're applying the human touch. Don't sit on a piece of content. Hit publish, get it out there, and rinse and repeat. So the more you build, the more pieces of content you publish, the more your site has the chance to rank in Google, and gain authority.”

What percentage of your time do you spend on quality versus quantity?

“I would say that we have two buckets right now that we're dividing the quality up between. So one bucket is thought leadership, and we spend maybe 2-3x the amount of time on that. And the other one is bulk SEO traffic gains. If we're going after those keyword gains, then we have a series of very simple edits to that piece, we generate it in Content at Scale with our own AI writer, then we move it to a human writer who spends about 30-40 minutes, and then we ship it. But with the thought leadership stuff, sometimes one piece will take a day, because maybe it's a case study of somebody that tripled their business using our technology. So something like that we have to pause, go get the Zoom call and that's something we want to spend more time on. So if you look at your content in that way, which are the ‘Rockstar pieces’ that maybe I want to put in an evergreen email campaign, I want to put in my signature, okay, and spend a lot of time on those pieces. But the other pieces, maybe I want to get to the top of Google for this keyword, so let's do a 3,000-word piece using AI to write the bulk of it, and then just spend an hour editing it. Put a timer on yourself so you don't overdo it, and then ship it. Thinking of content in terms of quality and quantity is a really good divide, because you want to have both, but I would say that quantity now matters just as much because you have to compete in order to get those rankings and you're competing, oftentimes against a lot of other content pieces.”

Julia McCoy is the President at Content at Scale, and you can find her over at contentatscale.ai.