Google’s new Generative AI has redefined metrics and created a stir in the SEO community. In order to stay at the top of the funnel, one needs to employ methods that align with the informational cycle.
In today’s episode, we navigate through the possibility of Google’s remodeling and how it can impact the SEO world. We also delve into securing your content by using our in-depth analysis tool, KeywordSpy, which indexes and boosts your page systematically.
Here’s What We Will Cover Today:
How AI Content Generators Work
Google’s Generative Search
How To Get Indexed Via KeywordSpy
Links:
Find this podcast here:https://www.youtube.com/live/evAqke90raY?feature=share
Newsletter:
Where to find Josh: joshbachynski@gmail.com
Where to find Greg: Support@handsoffpublishing.org
Referenced:
KeywordSpy: www.trykeywordspy.com
(0.00- 0.57)
Welcome to the SEO + AI show. My name is Josh Bachynski, and with me, as always, I have my co-host, Greg. Greg was just on vacation. How was your vacation, Greg? Did you actually get to take a vacation, or were you busy doing SEO during that time as well?
Greg: I was unplugged for ten days. I was able to log on for a couple of days to attend to a couple of different requests from various folks, and ultimately I had a great time. Hit up three islands that are Dutch-based islands.
So they love their drinking, they love beer, they love football or soccer. So I love all those things, and I love the beach and water. If you ever get the chance, take a cruise out of Miami and hit up the ABC islands.
(1.00- 1.50)
That's where I was. I'll let you look it up or let ChatGPT tell you what ABC islands are. Yeah, it was great. Thanks for asking. Glad to be back and fix bugs.
Josh: You get to fix bugs for the rest of your life. That's what we get to do. But that's what happens when you make an SEO tool. If you've never seen the SEO + AI show before, we are the best show online that talks about SEO and AI.
All the information you're seeing today comes from my Underground SEO University. It's aptly titled. Greg and I also are the chief partners on a new SEO+ AI tool called KeywordSpy.
If you want to try it free, go to try keywordspy.com and you can try the tool free for two weeks. It does everything that we're going to talk about today and more.
(1.51- 2.40)
And what are we talking about today? Today we're talking about AI-generated content. We talked about it before, but now everyone is implementing it these days.
If you have any questions, please put them in the chat. If you’re not watching live, why aren’t you watching us live at 10 a.m. Pacific time every Wednesday? You can ask questions live and we will actually do some SEO for you. You can ask about your page. You can ask us some questions.
We'll try to help you with your SEO, especially. You've been having trouble, and I know a lot of you out there have been having some trouble. So if you have any questions at all, please email me at joshbachynski@gmail.com
(2.41- 3.32)
Today's topic is AI-generated content and so I want to try some samples of AI-generated content out there now. Greg will attest that we are experts at AI-generated content.
We've been doing AI-generated content for some time and we've been doing it pretty seriously. I've been using GPT since before Chat GPT. I've been using it since it was GPT 3.0, the very beginning version. I’ve been making prompts for over six years now and checking to see how the AI-generated content works.
I’m talking right now about the text for the most part. We've done other videos on Mid Journey on how to do AI-generated images, which Midjourney is the best tool you want to be doing there. You're going to want to go back to my channel on Youtube.com/joshbachynski and check out my Midjourney episode that we did as well.
(3.35- 4.15)
It's really super cool, especially for SEO as well. MidJourney is not only for making pictures and editing them. No, you can use it for SEO, so go back and watch our video on that. Today, I want to focus on text-generated AI content and
Neil Patel's got a new one out that I wanted to hopefully look at. We will also look at our old friend Kyle Roof’s page optimizer pro tool. He's now doing AI-generated content. We were among some of the first, along with another company called Content Scale.
Justin runs Content Scale. We know him very well and we've been testing his content. He's been testing our content.
We've been doing AI content detection and AI content generation. So this is the new thing, guys. This is the new war.
(4.16- 5.06)
JM says, “I just sold three of my content sites that depended on AdSense. Not sure what SEO work to do now or where to invest the money SEO-related. I don’t like client work.”
So you probably saw my show last week and that's probably why you're mentioning it. When Greg was on vacation and I was talking about generative SEO, I was talking about generative search, the new Google Bard or Google Magi or generative search.
They're calling it generative search. That's what I'm calling it. It's the chatbot at the top of Google and how they're going to be dipping into informational sites. I said that if you had an informational site that is on the borderline, you might consider selling it. So it’s smart of you to implement it really quickly.
(5.07- 5.34)
There’s some risk there. It’s not the Armageddon people say it’s going to be, at least I don’t think so. There are ways around it for sure. You don't have to do client work, JM.
You could still build sites and sell them, but now you build them in the middle of the sales funnel or the sharp edge of the sales funnel, and you sell those sites. You can still build sites and sell them.
So it's still a valid way of making money in SEO. It's a great way of making money in SEO. You don't have to deal with clients. You don't have to deal with anyone arguing or complaining with you. Although I love my clients, don't get me wrong. But of course, client work is client work.
(5.35- 6.19)
You're your own boss. So it has pros and cons. The pros are that you work for yourself. You get up when you want. You build the sites as you want. You rank them as you want.
You get the success, you sell the sites, and you make the money. The cons are you do the SEO that you want, you got to try and sell the sites and convince someone to buy them and then you make the money. So the pros are the cons. It’s kind of a double-edged sword there. It's still viable, still possible.
I would just stay out of the informational space. For too many years, SEOs have been fetishizing, the information space. They've been spending too much time making too many information sites.
(6.20- 6.48)
The problem with that is you're informing the world for free. I’ve been telling my students in Underground SEO to not do that because I will not say I saw generative search coming.
I definitely didn't see a generative search coming. Nobody saw ChatGPT coming. But if you want more information on that, watch my last episode on generative search.
You'll see what I'm talking about there. So here's the thing with AI-generated content: it's very dangerous. It's not created equal.
(6.50- 7.26)
If you've been watching anything online right now, you know that AI-generated content, sometimes works, sometimes it doesn't. It's quite dangerous to just use it. A lot of people are just firing up ChatGPT and just using that content verbatim and thinking it's going to help them rank. I did a poll recently at Underground SEO University.
There are folks who I’ve trained how to do AI content and they may either be doing the training in Underground SEO University or they might be doing keywords by one or the other. So I did a poll to see how well now this is around.
I mean, it's only 25 responses, but it's a pretty good sample of my student base. We’re looking at how well AI-generated content works from a statistical sample of my student base. A third of them are using AI content and seeing boosts.
(7.27- 8.12)
This is pretty statistically significant. I've talked with other SEO industry experts out there and they agree. Sometimes you can see AI-generated content working.
Greg, you've done sites with AI-generated content. I've been testing AI-generated content. Greg has had more luck with it than I have thus far, but I think it depends on the niche you're working in.
There's another rumor out there that you need to be in an informational niche for it to work well. I don’t know why that would be. I've speculated in Underground SEO University as to why that is.
(8.13- 9.10)
If anybody wants to know, ask in the chat and I'll tell you what my speculations are there. I’ll tell you why informational AI content would work, getting it indexed and ranked in Google. But it seems like the middle of the sales funnel and lower in the sales funnel. Some people say maybe that's harder.
That's not been proven, that's not been verified. I'm still testing this, but at least that's the rumor out there that people are saying. Overall, in the informal survey that I did, about a third of people from fairly high-level SEOs are getting AI content to work.
Another third roughly said 36% said yes, I've used it and got it indexed, but I have not seen boosts. So most people are getting it indexed, and half of those people see boosts. Around 68% are using AI content and seeing it indexed, and about half of those are seeing boosts.
(9.11- 10.05)
So that's pretty high, right? The other third is like, “I tried it but did not see boosts. I went back to handwritten content and I did see boosts.” 8% were too scared to try it, and 12% says, “I have a subject matter expert, so they write everything for me.”
I mean, this is pretty statistically significant based on the whole industry. I've checked and talked with people, and they got demotions, so it's very interesting to see those results. Greg, do you have any thoughts on those results yourself?
Greg: I'm thinking AI content at the top of the funnel is totally fair game. I don't see many issues so far from what I've been doing. Again, I think it's all in the tools that I use potentially to make it less AI or just the keywords that I'm using because of the tool we created. So it's hard to say.
(10.06- 10.49)
It's not an apples-to-apples comparison. If you're using ChatGPT and some of their prompts and some other tools, you may fail. Whereas maybe I'm succeeding because of the tools that I'm using. Does that answer the question?
Josh: Yeah, completely. That's a totally valid point. That's where I was going to get to you, which is that AI-generated content is not created equal. Some AI generation is far better than other AI generation, and you have to be very careful not to be in this red section here or these sections here. You have to be very careful not to use the wrong AI content generator.
(10.50- 11.40)
It's not all created equal. To prove your point, I wanted to show you guys how some of the different AI content works and how some of it looks. So let me just look at the questions here quickly.
JM says, “Yes, I'm focused more on the bottom of the funnel. I've always loved informational. But you're right, I agree. I still have more info sites, and I'm interested in why AI is working.”
Christian says, “ The top of the funnel is underserved.” Yeah, so that's very interesting comments I'm not going to repeat myself based on what I said previously in this show and in previous shows.
(11.41- 12.05)
Christian, I suggest you go back earlier in this episode when I was talking about how generative search, Google's Magi or Bard, or whatever you want to call the chat at the top of the search is going to start cutting into informational. As I said in a previous episode, it's largely on search queries.
We weren’t doing that good of a job anyway, so it might not hurt us too badly. I always position myself in the middle or the bottom of the sales funnel because that's where all the money is. Whether you're selling a site, doing client SEO, doing affiliate SEO, or doing in-house SEO, the middle of the funnel, the bottom of the sales funnel is where more money is. That's why I suggest it.
(12.06- 12.46)
JM has agreed. JM also asks,” I have more information sites, so I'm interested in why AI is working there.” Okay, this is a little bit nihilistic, and this is a little bit conspiracy theorist, but bear with me
This conspiracy theory, in all the testing we've done in Underground SEO University– we're going to get to actually showing you some of the different AI content generators in just a couple of seconds. That's what we're going to pivot to the rest of the show, which is showing the AI content generators, how they work, how good their AI content is, and contrast and compare. But before I do that, let me answer JM's question.
(12.50- 13.25)
Why would informational content be a little bit more allowed? It definitely seems to be there's definitely content that has been generated that has way worse factors, that Google still allowed to be indexed. That’s because of a number of reasons.
Mostly is because it is about your SEO metrics plus your HEAT metrics, plus your AI content generation metrics.
It's not all created equal. You're like, “What's HEAT?” HEAT is my own acronym. It's EAT plus helpfulness so it's helpfulness. Plus experience, expertise, authority, and trust. That's very important. They directly measure this.
(13.36- 14.06)
This is going to be a major ranking factor moving forward because all of the things are going to be created equal. Links are going to be a wishy-washy signal, and links are probably part of the HEAT signal anyway. So HEAT is going to be really the major SEO factor in the future moving forward, in my opinion.
Not now, but moving forward it will be. It's going to incrementally get more important every year. It also includes other factors that you'll learn about in Underground SEO University or about when you start using KeywordSpy as well.
So go to keywordspy.com to try it free for two weeks. It's in beta, but it's still the best SEO tool on the planet, especially with the content generation and detection we're about to talk about.
(14.07- 14.49)
So if your quality is really good, for example, if your trust is really good, if your expertise and authority are really good, then Google will allow you to have worse AI-generated content. This is one theory that seems to be borne out with evidence as well.
Another theory and this is the more nihilistic one. Let’s think about how this happens. So Google puts out a generative search. Generative search starts to suck up 50% to 80% of informational queries answering that for users and therefore not sending 50% to 80% of traffic to web pages. Let's just say theoretically that happens.
I'm not sure it's going to at all, but let's say it does. There are many reasons why that wouldn't happen because a featured snippet answers a question way better than the generative search.
(14.50- 15.13)
The generative search is cumbersome. It's politically correct answers and slow and it's clunky.
I would rather just ask a simple question, and get a simple answer. I'd rather have that. So it's not in any way for sure that generative search is going to kill SEO or even kill informational queries, but that's what's on the table.
That's the risk, okay? To be perfectly honest, that is a risk. So let's assume that risk is true. Let's assume that the vast majority of informational queries are not really well served and underserved by blogs.
(15.14- 16.11)
Now let's say generative search, soaks that all up and they never click onto another web page. That means all–those business models, the AdSense business models, the ad business models, the top-end affiliate business models, the topical cluster business models that all want to build at the top end of the sales funnel– die.
They die because you're not getting traffic from search anymore. There's no point in making informational pages anymore in this hypothetical scenario. After 2023 or 2024, all of a sudden there's a drastic drop in the informational pages that are made and therefore a huge gap in the information that's available.
The information that is available on web pages is exactly the corpus Google uses to train its large language models. It's reciprocal.
(16.12- 16.51)
Google has these powerful large language models because we make web pages. If we stop making web pages in general or in a specific niche or a specific vertical or a specific end of the sales funnel or the top of the sales funnel.
Then Google suddenly doesn't have any informational answers for their generative search to say. You're going to ask them in 2025, “What's this new red widget that this new company made?” And they're going to go, “I don't know, I got no information about it because no one's making any web pages about it because no one's making money off of informational pages anymore.”
(16.52- 17.45)
That is going to be a big problem for Google. All they're going to have is the press releases and the marketing-speak that these companies put out, which quite frankly are nowhere near as good as the in-depth analysis that bloggers are doing on every single informational topic on the planet.
So bloggers decide not to do that anymore because they're not making any money anymore. Google shoots itself in the foot by making a generative search. Again, I told you this is a conspiracy theory but it’s a pretty good one.
This could be a reason why Google is really taken the dial in terms of the aggressiveness of policing AI content and really turned down that dial for informational. They don’t want to stop the informational cycle.
(17.46- 18.33)
They want you to use AI-generated content to spit out information on all everything that's going on and then human edit it a little bit for updated information so you can feed their generative search, so they still have information to tell users to stop them from going to your pages.
That could be the reason if there is any reason at all. Again, this is just an anecdote. This is another conspiracy theory of the Guessios. Everyone is like informational content is fine, you can do it. Hey, they might be right. They could be right, but there's no evidence to suggest that they are.
There's no hard evidence to suggest they are. This again is just still at the conspiracy theory stage. It is easier to get AI content indexed at the informational top end of the sales funnel.
(18.35- 19.01)
That is the only reason why I can think of that would be that Google does not have classifiers for information versus e-commerce versus review style. That's something that we SEOs do. We SEOs arbitrarily cut up web pages into e-commerce pages and review style pages and informational pages.
Google does not cut up pages that way. There's no evidence to suggest they do. I've been testing this for six years.
If you talk with Google, that's not how they talk about it. You talk with Google at conferences, as I have, and behind the scenes, it's not how they talk about it. They do not break up web pages as e-commerce or review style.
(19.02- 19.55)
That's what we do. They do cut it up in terms of informational, informational transactional, or transactional. That's our name for it. That's not what they call it either.
They call it know-queries where you want to know something. Know-do queries where they want to know something or do something, and then queries where they want to do something, and then navigational queries. They have four levels to their navigational, to their sales search funnel. And for them, it's not a sales funnel, it's a search funnel.
They have four categories in their search funnel. It's Know, Know- do, and navigational. So you're like, Know: what are red apples? Know-do: What are the best red apples? I want to buy red apples? Navigational, Josh is RedApples.com.
(19.56- 20.45)
That is the search funnel for Google and the types of queries you will get. Again, generative search is going to take most of the informational and it’s going to cut into the know-do as well, about 5% of those possibly across the board.
Does Google have any impetus to allow informational articles to be indexed, even if they are low quality and even if they are artificially generated? Yes.
So they can continue to fulfill their informational cycle so that their own generative search has a corpus to train on, and they have information to tell you in the generative search at the top. So take that conspiracy theory for what you will. But that was the conspiracy theory.
(20.46- 21.21)
It's all about AI content generation. One could make the argument that the SEO industry lives or dies through AI-generated content. It is one of the main ways by generating content faster, and with better ranking, and better, that we can stay afloat ahead of Google and make enough money to continue on in what might happen with generative search.
AI content generation is this trending thing that even if it doesn't work, people are going to do it like crazy. So, you have to be very careful about the AI content generator you use right now. KeywordSpy, our tool is very good.
It keeps the numbers very low. We're going to show you the low numbers in a second. And then when you human edit it afterward, or maybe with some ChatGPT magic edit it afterward, which we will then build into the system later.
(21.22- 22.01)
After we machine learn how all of this works. We're going to always make the tool better. We are always going to make the AI generate content better because it is the future of SEO. From doing spray and pray to staying ahead of Google and just for the fact that everyone else is going to do it.
It is the future of SEO. You've got to be really careful to make sure it works well because there's huge amounts of snake oil there. And hopefully today I hope we get some good examples to show you the kind of snake oil that you can show there.
So, Greg, did you have a chance to generate any content? I know you have a Neil Patel one there. Did you want to take a look at it and show it? I'm going to turn it on.
(22.02- 22.54)
Greg: Yeah, check discord. I sent three articles. Well, two are the same subject and then one isn't.
Josh: So here's one generated by Neil Patel's new tool. It’s Glastonbury tree removal, cost process, and safety tips. If you're a homeowner in Glastonbury and need a tree removed, it's important to understand the process and cost involved.
Yeah, standard SEO content. I could just read this and tell you it's AI-generated content. So this is from Neil Patel's tool.So, Greg, can you tell me a little bit about Neil Patel's tool and how easy or hard it was to use it?
Greg: It was quite easy. It does walk you through. I made some of these selections. I think it was just the H one and then some subheadings.
(22.55- 23.33)
Josh: So you had to choose these yourself?
Greg: Yes. They gave you a checklist or bullet point, whatever it's called.
Josh: So we do that a little bit better because we automatically do these for you.
Greg: So I went ahead and I selected what I thought, I did it quickly, to get it ready for this, and then it generated the paragraph below it.
So maybe if I selected more subheadings, it would have created more content. But I was doing it quickly. So that's what it gave you?
Josh: Yeah, that's perfectly fine.
(23.34- 24.30)
Well, how well does this content do? How good is this content in terms of AI-generated content? Well, let me go here to KeywordSpy.
Now, let me show you in AISpy, our AI detection. This is a GLTR-based AI detection. This is going to be the detection tool that wins in the end, because it's the one that is based on the most stringent factors, I think that is going to be what Google is using.
So let me change the colors here to show you what is bad. So I'll make red bad, yellow is almost as bad or let’s make that orange. Then I am going to make this a little bit lighter blue and I'm going to make this little one a bit darker blue, so you could change the colors there. Now, let's paste this as plain text here.
(24.31- 25.15)
Let's analyze the text. Now, in our testing, there are certain predictability levels you need to be under. Just immediately look at this, look at all the red and the orange that we see here and how little blue we see.
This is a highly predictable. Look how long it goes on. The worst predictability length is 24 tokens.
This is terrible. So I guarantee you unless you have the best, most trusted site on the planet, you would have some trouble getting this page indexed. And even if you did, it would be in the information space, which, quite frankly, is across the board.
(25.16- 25.57)
So look at this. It's quite tokens per average in these sentences. These red sentences you can see here is it goes 6.66. That's interesting. It's the token count of the devil. Greg, what do you think about that Neil Patel article? I think the metrics there were deplorable.
Greg: The overall predictable being over 80 is quite high. Again, I think people are going to use that out of the box, trusting Patel, and for all the reasons you just gave, it's going to be a problem.
Josh: It really is kind of sad that the SEO industry is full of such snake oil, quite frankly, which I've been trying to call out the Guessos and the snake oil for the entirety of my career.
(25.58- 26.40)
I'm going to try to do that. Neil Patel, sadly, is one of those snake oil salesmen. Even a blind pig finds an acorn, and Neil Patel is the blindest pig of them all, to use the metaphor.
He just makes so many pages that so many of them rank up highly. And you're like, “Josh, that's a strategy.” Yes, it is a strategy, but it doesn't mean he knows how to do SEO. Spray and Pray is just one method of SEO. It's not all the methods of SEO and I get it.
I get people look up to him. People want to be like him. His branding is on point, I'll give him that. But look at the stats. Look at the numbers here. 100% AI detectable from his article here. That came from his thing.
(26.41- 27.34)
KeywordSpy shows you why. It is over 84% predictable. It's 6.6% average predictability length. Look at all the red here. Look how bad this is. Look at how little the blue is and the worst predictability length is 24. From the writing style, you could tell it's AI generated as well.
This is not good news for how this thing works. Let's look at another article here that Greg sent. So tell me about this tool here, Greg.
Greg: It is a demo. If you go into their you sign up for a free account. This is the example demo article that they give you. It is 5000 words. So what you want to do is take about 500 words and plug it into AI spy because that's the max length that we will analyze just for the testing tool.
(27.35- 28.27)
Josh: This is on page AI.
Greg: This is on page AI. Like I said, it's a demo. I didn't create it myself, I just copied it or actually downloaded it.
Josh: So here is onpage AI. I think I have 500 words here. Well let’s see if I got a bit too much. Let's take a little bit off here. So we'll see how well this does.
Okay, so this one's better. It's got less red, it's got a little bit more orange and a little bit more blue. But still, look at all the red. Look at all the words that are in the the top ten most predictable words and word endings because tokens break up words into pieces. Look how predictable this was.
(28.28- 29.08)
This was 81% predictable for that word here. It's 73% predictable overall at 4.57–The worst predictability length, it goes on for 17 tokens.
Probably this sentence right here, 17 tokens. It goes on for without breaking the top ten predictability, which statistically only an AI could do. We've detected it was AI generated from the writing style as well.
So we have four different ways of detecting AI generated content. It's clearly AI generated. This would not do very well either.
Now, this is clearly a top of the sales funnel informational page. I've seen informational pages with these numbers get indexed before. If the site was highly trustworthy, they would get it indexed.
(29.09- 29.58)
But they still going to have that boat anchor of that distrust of it being AI generated content. They're still not going to rank as well with this. So if I take this and I paste this into originality AI and I scan it, originality AI also is going to detect this as 100% AI.
It is the darkest red here. It's like this is clearly AI generated. We know this very well.
So, again, this is not going to do that well in terms of AI generated content. Now I just show originality AI because they're my second favorite. I like ours the best.
There is no guarantee that originality AI is the way Google is going to do it. The way originality AI does their content detection is totally mathematical. On their own machine learning model, they ran over 10,000 articles through that were GPT generated.
(30.01- 30.41)
They ran a control. They didn't see how big the control was of human generation and they got the exact numbers here in terms of the token replacement, they got the exact numbers of what these need to be, which I found interesting. We're doing something similar by breaking it down into 1-10, 11 to 100, 101 to 1000 and 1001+. They know the exact numbers here and it's their own secret sauce.
We're running tests to see if we can find out what these exact numbers are here. But these numbers are partially showing the story of what originality of AI finds as well. So it's very interesting.
The science of AI content generation and AI content detection. This is the new SEO war. This is the way it's going to be.
(30.42- 31.45)
KeywordSpy here has generated article. So we did the generate article here on Glastonbury tree removal. We did it out of the box in KeywordSpy and look at how well it works without any human editing afterward. This is the other thing I’ve come up with.
So here's the second poll I asked folks another question. Getting AI generated content indexed across the board is quite difficult. For example, this is a piece of writing that could get indexed across the board that still originality AI detects is 100% AI generated, but it's got really interesting writing in it.
So if I go to KeywordSpy here and I paste it in, this is what you would need to do to get the numbers you need to kind of get indexed.
(31.46- 32.23)
Here it is only 60% predictable, it's under the 60% mark and it's under the three for average predictability length. Look at all the blue and how little the red is. The worst predictability length is six in terms.
We can also detect it as AI content in terms of the writing type in our special methods of detecting AI content, which everyone is going secret now with how they do it because it's such a highly coveted and lucrative thing to know. These are the kind of numbers you would need to be able to generate AI generated content indexed across the board, no matter where you are in the sales funnel. And even that originality AI still detects it as 100% AI generated content.
(32.34- 33.01)
For that reason, I did another poll with my group and I'd like to hear the people listening to the call, what you guys think? This is the poll that I did and I said if I said automatic awesome SEO guaranteed AI content with a little human editing is unfeasible. We mostly want human editing anyway to make sure it has that personal touch.
But if I said that AI will generate all of it, you just edit it. This means that SEO+ AI content will take you about an hour per page. How would you feel about that is what I asked? And 50% of people said, “ I would feel awesome about that.”
Only about an hour for full SEO and writing of a page. That will do well.
(33.02- 34.01)
Another 35% said, “Okay, I wanted full AI writing and SEO insertion, but I will rank well. This is a good deal.”
So the vast majority of people are like, yeah, you know what? We don't really want AI generated content. Yes, I know. Everyone is tripping over themselves to make the AI generated content work for 100% of people, 100% of time.
The question is not to AI content or to not AI content. That is not the question and that was never the question and never will be the question. The question is, one, what percentage of AI generation are you going to use to generate your content? Two, how much is that going to get you indexed? Three, how long does it take you? Four, how high are you going to rank by it? If those four factors that are important, again, what percentage of your content is going to be AI generated?
(34.02- 34.59)
Just 5% of it? Like your SEO keywords or 95% of it? Plus your SEO keywords, plus the writing, plus the headings, plus the headers, plus all the HEAT stuff you need to put in there, all the helpfulness, all the expertise, authority factors in there. The tables, the charts, the graphs to make it good for users.
Did you have to make that or did AI generate that all for you? Super fast , and you just edited 5% of it at the end and will index for you. And will it boost you? Because at the end of the day, if automatic AI content does not boost you the vast majority of the time, then what is the use of it? Then it has no use. And so I asked my students and they gave me a pretty statistically good response.
They're like, “The vast majority of them are cool with the way we're doing it in Keyword Spy, which is we will generate the content for you right away. All these headings are automatically generated for you as you go.
(35.01- 35.51)
It has innumerable NLP based keywords and Google semantic keywords. We take all of the available AIs to put in all the keywords for you that you need to put in. We did that for you automatically.
Greg, how long did it take you to generate this article?
Greg: It literally took two to three minutes. Our wizard lets you create the header, but I went back and created one because I forgot and the subheadings were created and I even went in if you click on the SEO tab, you'll see a URL title tag and meta description done for you. It's so quick.
I saw the dials at over 100. I can dial this back a little bit, get it in the green. The HEAT score already came out higher than your competition, so I was happy with that.
(35.52- 36.48)
I probably would want to fix that long 36 and that's it. It's definitely going to be on this one less than an hour. So if I can get this done in 20 minutes, this client probably has 20 pages. That’s not even a day’s task and it’s done.
Josh: Yeah, I agree completely. I love the meta description. It looks great. It has a perk in it.
Perks are great in there. We're still in beta, folks, but we're doing really well.
You just got to remember to insert your phone number here. It says insert phone number. A human does need to look at this and give it their personal touch. Quite frankly, we're still going to strive for fully AI content, but there are valid arguments to not even do that. There are valid arguments to say, you know what? This is good enough. Stop here.
(36.50- 37.29)
I still want to strive to have AI generated full content. Statistically speaking, pushing that dial higher statistically that we do it better for more people, more a time in more verticals to open up and spray and pray more for people.
This is automatically produced so we can go into keyword discovery here and then generate all of our keywords in keyword discovery. Let me show you keyword discovery here.
So if I go into keyword discovery, you go into one of these things, you check out the keyword difficulty, and you say you go, “Okay, look how easy it is or how hard it is to rank in this niche.” No one else in the industry does this anywhere.
(37.30- 38.04)
It does its own special AI based analysis, like an AI consultant would, of how hard or how easy your niche is by looking at the EMQ in the page. Is the EMQ in the title? Is the EMQ in the URL? No one else checks for this. What is the domain rank?
What is the page focus of the main topic? What is the site focus of the main topic? What is the title focus of the main topic? How many backlinks do you have? And then it's also asking you, are you going to use AI content? There's a modifier for that.
Are you a top level expert in this? There's a modifier for that. Do you have a powerful site? There's a modifier for that. And it'll tell you how difficult or easy it is to rank for all of these niches, and it'll show you the easiest one that you probably can have the most likelihood of beating.
(38.05- 38.45)
This is super easy at seven. This is easy at -75 because they are way over optimized in terms of their EMQS. This is probably pretty easy because they're also way over optimized in terms of their EMQs.
This is where you would start in KeywordSpy. This is the best keyword difficulty planner on the market. That's already based into Keyword spy.
But in the future, my point is, what we want to do is just have a build me the site button. It does this automatically with the AI content you see generated here with all these levels, which I'm going to go through and explain in a second. I haven't told you how awesome this all is.
(38.46- 39.33)
It's going to go through and generate all these and it's just going to produce a site for you. That is my dream to do in the future. It's not quite there yet because we need a little bit more AI engineering.
We need a little bit more testing. And quite frankly, there's reasons why SEOs are like, “Hey, Josh, don't AI me out of a job. I still want to do some stuff here. I need to take a look at this to eyeball this to make sure it's doing SEO just the way I like to do it.” Let me go through why this is so good.
First off, the Google semantic AI score here, this is our patented scoring system. I call it patented. It's not patented.
This is our exclusive scoring system that we use in the tool. Let me show you the reports. This is how that looks.
(39.34- 40.18)
This is so awesome how this works in here. Let me just check the reports here just for a different keyword.
Greg: It's going to be a different search because I just shared the link with you, remember? You don't have access to, obviously, my dashboard.
Josh: Yeah, this is a different search. But look at the way we score and check it. We check BERT. We are the only tool who actually takes Google face value and doesn't use some off the shelf NLP program that is GPT-2 annd the corpus was cut.
The AI was produced like five years ago. We are the only AI SEO based tool that doesn't do that. All the rest of them do that.
(40.19- 40.48)
We actually build a model for BERT for you. We actually build a model for neural matching for you. We actually build a rank brain model for you.
We score based on all of these AIs to get you the actual score of how useful these phrases are and how useful these entities are. Based on this score is what we use to make this dial here.
You can see my competitors for this one that Greg sent me is only around 49%. We're at 102. We're fine, we're golden, we're great.
(40.49- 41.26)
We're way above them, right? That’s why you want to be way above them. Because now what we're going to do is I'm going to start, for example, highlighting this and clicking the Improve button. After that finishes working, it's going to reduce this and it's going to reduce these numbers here.
But what it's going to do is it's going to make the AI content numbers better.
Greg: Try that one more time. It might. Because you're view only?
Josh: Yeah, I'm in view only because this is your page you were working on.
Greg: Yeah, that's my page. Can you try the improvement one more time?
(41.27- 42.03)
Josh: Yeah, it's not going to work because I don't have free to write access to it because you sent it to me. I'm just saying as an example, I press the Improve button, this would get shorter and it would make these numbers go down, but it would also make these numbers go down.
So these numbers going down is what you want. You want this number to go down for it to be less predictable and more index worthy.
You want this 74% to go down. Overall predictability, for it to be less predictable and more index worthy is what you want to do. You want to keep these numbers high into the green here, which is kind of a rough number, but you want it to be higher than the competitors, at the end of the day.
(42.04- 42.42)
Now, this NLP is just for over optimization. We just use a common measuring stick, which is super fast and super easy for over optimization. It is this Google semantic AI that I'm focusing the most on, and that again is based on the BERT test we do for you.
The model we build, neuromatching model we build, the Rank Brain model we build. You can check Rank Brain, Neural matching tests individually. We also check the image entity vector for you, which is another AI people don't even know about.
We check the Dom vector for you, which is another AI that people don't even know about. But here's another thing that people don't know about that we do. Check out the HEAT thing that we do.
(42.43- 43.27)
We do a full analysis in all of the actual questions that Google asks. If you've looked up the helpful AI, if you've checked up the review AI, if you've checked the expertise, experience, authority, and trust AI that Google is using, these are the exact questions that they ask.
We added one in the end, as per one of our customers asked us to add this because it was important for them not to show any unfair bias against BIPOC, which is fine.
So we add that in there to check whether your content is doing that which is probably a good thing for your money and your life content too. We give you an actual score on each of these things so you know how to make your content better. You can see what has the lowest score.
(43.28- 44.01)
You can see seven. You can see, “ I would not expect to see this content in a reference by a printed magazine, encyclopedia or book.” That means you should be improving this content.
You got a low seven here. Or if someone researched the site producing the content, would they come away with an impression that it is well trusted or widely recognized as an authority audits topic? That's where you would know how to improve exactly how to improve the content. No one else does this for you.
No one else checks this for you. You can also export this to CSV or you can share this whole page the way that Google did. He shared it with me somewhere.
(44.02- 44.30)
There's a share button.
Greg: No, it's disabled because I shared it with you. You can't share it again.
Josh: Since I don't have write access to this, you could send this to your writer and say, “Hey, improve this and send it back to me. And they can copy and paste it and prove it, send it back and you'll know exactly what they need to improve because we're doing the heavy lifting here for you.
In terms of the presentation and production questions, where is any good in terms of presentation or production? No.
(44.31- 44.59)
Is the content mass produced? Yes. This is a bad score for that. It seems to be mass produced.
Interestingly, this score is always low. When these numbers are always high, it's consistent. Okay, look at this. Yes. It's primarily made for search engines. That's not a good thing.
Are you producing lots of content on many different topics? Yes. Are you using extensive automation? Yes. There's a low scores of seven.
You would want to see eight, nines or tens here. Are you mainly summarizing what others have to say without adding much value? Yes. It's showing you all the ways to make your content better.
(45.01- 45.43)
Nobody else does this. There is no other tool that does this out there. So we have a HEAT score and we also check to see how good on HEAT your competitors are.
If they're only 58%. You could say, “Well, I'm at 77%, I'm better than they are and they're at 58.” You can leave it at that or you can follow the suggestions and make your content better.
That's another reason why the human touch is important. We want to be all things to all people. We want to just have a fresh content editor here you can use and all you use KeywordSpy for is to generate your SEO tags and maybe you get it to generate your H1.
You can go to content and get it to generate your H1 here or in the outline and get it to generate your H1.
(45.44- 46.15)
Greg: I think it’s because you're logged in that demo account. I honestly think that's causing that issue there.
Josh: Yeah, that's the reason.
Greg: Yeah, it's conflicting because you're logged in because normally someone getting this share link won't even have an account. We'll correct that.
Josh: It's just because of the way we're doing this demo, you wouldn't normally see all those errors pop up. Normally you might see other errors pop up. As we said, we're still in beta. This thing is still in beta. It's got the occasional bug. But Greg is even working on his holidays and fixing this thing.
(46.17- 47.12)
Even when he's 1000 miles out at sea, he's fixing this thing. This is getting better by leaps and bounds. We've had a lot of people using this and getting good results.
I don't want to just share it because I'll show you the Wins channel. I don't want to leak some of our tactics in the Wins channel here in SEO University, but a user just posted a Win from using it today.
Greg: That was actually me. Was it the helpful content?
Josh: Well, you posted that, but also Simon posted another one as well.
Greg: Oh, yeah. I think they were helpful content too. I think this is like a hidden gem that we haven't even focused a lot of our talk on. Is this helpful content? That's all I did to the page. It literally improved it from, I don’t know, but it was a major improvement.
(47.14- 47.59)
I think more to come on that.
Josh: You notice how it also checks your headings and you have the right amount of headings and it checks your words. Now the word count is really high and that's on purpose, so that now we can go through.
We can just use the improve button. It'll shorten it, it will reduce these numbers slightly, but as long as you're over your competitors, then you'll be fine. It'll bring these numbers down to a level that is viable to get indexed across the board.
We won't have the same trouble as like, Neil Patel's content that is automatically detected or the kind of content that we're seeing from these other content generators.
(47.60- 48.42)
I hope you appreciated that talk about AI content. It is the way of the future. Let me see here. Let me share my screen here and look at the questions you guys got.
Doran says, “You're feeding the machine.” Yes, exactly. You're feeding their machine when you're making informational content.
That's why I might think you might not want to be doing that anymore.
Greg: I won't comment on that. So if I'm thinking about it, the way my brain works and logically, if I'm Google and I'm consuming something, if I just see the same thing over and over again, just spun, I'm going to take that. But I'm probably going to throw out 99% of it. What I really want is new information.
(48.43- 49.28)
I want something new and current that really perks my ears. I mean, that's how a human works.
I'm wondering if Google is working the same way, and therefore as an SEO, you should really focus in on that. That's something I wanted to get your thoughts on.
Josh: Yeah, so that's a great question. That's going back to the theory of why Google might be allowing AI generated content to index easier. It's because they need it to feed their own generative LLM. So the question might be why and how are they going to deal with new topics? Because they're going to have a poverty of information.
(49.29- 50.15)
They have a wealth of information of all the users making informational pages to service informational queries. It's the lifecycle of Google that we've been doing for the last 20 years.
And we make the content, they rank us, they reward us by giving us traffic. We monetize the traffic. When that starts to go away, probably in some ways at the top of the sales funnel, especially then people are going to stop making informational pages.
So then the question I think Greg is asking here is, how is Google going to stay up to date then? Even if they're allowing AI content to be easily indexed, so they still have some informational content, they're not going to have the wealth of information they currently have to service queries.
(50.16- 50.50)
The answer is, you're right, they're not. They're not going to have that.
They're going to shoot themselves in the foot. But I think Google knows this. This is why Google is being so very careful in how they roll this out, because it is changing their base business model.
It is changing the ten blue links business model. They've based their entire 100 billion dollar a year profit monopoly they've been running for the last 20 years.
It changes all that. So now they got to be very careful about how they roll this out. I think you're right, Greg.
(50.51- 51.20)
It's like, how are they going to stay up to date with new information? Well, there's all kinds of ways they can do it. Their own AIs can write the articles, their own AIS can generate the information based on press releases and the information that humans would have.
They might be able to displace humans out of the loop entirely, in which case they won’t need us anymore to make AI generated content. They won't be indexing it anymore because it costs them money. It costs them millions of dollars a year to index content anywhere. They can save money from indexing, they will.
So that is a major problem. We've been on an indexing freeze and people have been having trouble getting any content indexed. I got some content indexed from sitemaps, but people have been having trouble getting any content indexed for the last 70 to 140 hours, depending on who you ask.
(51.21- 51.55)
They’re obviously huge change to the corpus right now. They're making some giant, huge change to their corpus right now. They're reprocessing the corpus because they want to limit the amount of pages that get into it, because I need to reprocess the whole thing for their generative search.
And this does not always happen on the fly, guys. This is interesting. What will happen in the future? It's still very difficult to guess.
I think whatever can deceive us is what Google will do, because that's kind of their modus operandi. They will do whatever gives the best answer to the users.
(52.00- 52.29)
If it ends up that they cannot beat us in terms of the quality of the information we produce, they will keep us in the loop.
If they can beat us in terms of the quality of the answers that they can produce, then they don't need us in the loop anymore.
Greg: I said we're out.
Josh: Yeah, possibly, quite possibly at the top level of the informational level, that's where I'd be worried the most. But you bring up a valid point, statistically speaking, there are still going to be huge cracks in the information scape.
(52.30- 53.05)
Without humans making content, there's going to be huge cracks in the available information and the analysis of said information that, quite frankly, is going to underserve their users in generative search, meaning the whole thing will be harder to get people to adopt.
Like I said in last week’s episode while you were on vacation, that there's nothing beating the featured snippet.
You ask a simple question, you get a simple answer. It's fast, it's easy, we're used to it. Those are times where we don't even want to use the generative search. I'm not in the group that's saying it's over completely, and I'm not in the group that's saying, don't worry about it at all.
(53.06- 53.53)
I'm a little closer to that end, saying, I think it's not going to be the armageddon we think it is. But if I had an information site that was barely ranking or already having trouble getting indexed, that the generative search could very easily answer the questions of my site, I would sell that site right now, is what I would recommend for everybody.
Otherwise you need to get in on an AI tool. Now, ChatGPT content is just as bad as Neil Patel's, and that's what Neil Patel is using. Neil Patel is going to be using ChatGPT content.
That's where it comes from, right? That's what everyone is using. So that's why I'm not surprised at all that it's all going to be the same.
(53.54- 54.31)
So let me just answer any more questions you guys have here in the chat. Let me see, Inquirer says, “I tried using KeywordSpy a few times. Too many errors for now.” Yes, we're still in beta, so KeywordSpy is still in beta.
I would get in the grandfathered price now, though, be a beta user. Help us fix the bugs. We give you a huge break on the tool, on the price to do that, and a huge amount of credits to do it.
Because once this tool goes live, it's going to cost a lot more money than $144 a month. So you can try it free for two weeks. Go to trykeywordspy.com and try the tool. Free for two weeks. You will enjoy it.
(54.32- 55.16)
There are the occasional bug, but I wouldn't wait too long, because when we suddenly jump that price up, I'm not going to give breaks to people. That's going to be the price, For example, content at scale.
I haven't checked the pricing lately, but last time I looked to do the same thing that we're doing is $1,400 a month. $1,400 a month for worse content. That ranks worse, that indexes worse, that has worse metrics is $1,400 a month.
What we're going to give you now for $144 a month, we're 10% of the price right now, so we're going to jump up to some number like that in the future. I would get on the grandfathered pricing now, and I would get into AI now. I would get proficient with the tools now.
(55.20- 56.07)
You will be well ahead of the game. If you have any further questions, let me know.
Doran says, “I really believe leveraging AI content doing the heavy listing is great when it's human edited and the outlines are done manually. My concern is this, but what is your thoughts on this”
Yeah, I agree, but we do the outlines for you automatically, too. Our outlines are better than everybody else's.
You can also do them manually if you want, but our outlines are better than everybody else's, so no one else can do outlines very well because they don't really know how to program AI. I do. Greg does.
So check it out. Our outlines are better, but I agree. Human edited is the way to go across the board is what I would recommend right now.
(56.08- 56.50)
Major AI generated for the content, both the keywords tables, lists, everything, and for the main body of the article, and then human edited after that. To get the right numbers is what I would do.
Doran also says, “A lot of new AI tools are marketing one click SEO, optimized articles. This flooding the web with bloat and junk AI. That's what happens. It rubs me the wrong way.”
I agree. That's why you have to watch out for the snake oil out there. We are not snake oil. We are trying to make the best product. It's attached with my name. You guys know that I'm a straight shooter despite the beers I'm drinking here, I'm a straight shooter.
(56.51- 57.17)
That I tell you the truth and that I'm seriously concerned with ethics. I studied ethics at university. The only one that I can say works pretty well is ours because I've shown you the poll.
33% of people using it out of the box are fine. The other 33% have to human edit it. So most people are 60% are doing it and getting it to work for them in some ways.
So that's what I would say. Do the human editing after you do the generation. Our tool is primed for you to do that.
(57.18- 58.24)
Let me see what else other questions there are. Sifola asks, “Can we use that content on agencies websites?” The answer is yes, of course you can use our content from KeywordSpy on agency websites for sure. Greg, do you have any final comments?
Greg: No, just if people want to leave some comments about maybe different topics they want to hear about or different questions and then also I was thinking, what do you think about this having potentially some guests come on and talk about some of the things that they're doing in AI and SEO. I thought that potentially might be a good idea for sure.
Other than that, I just think related to the KeywordSpy tool, definitely this week we made a lot of updates. We're going to continue to update and it should be working in a much better way this week and still give it a try and then email us if you have any questions or concerns.
(58.25- 58.54)
We have 20 or 30 grandfathered spots left, so get in on it now because when we say, beta is done, bugs are fixed. Here's the real price. It's going to shoot up, guys, so make sure you get in on it now.
It is the best tool out there because it does things another tool does and it's based on actual science that's proven to rank you. And it's only going to get better from here. There's going to be real one click magic in it that actually is going to work.
So I would get on it sooner rather than later.
(58.55- 59.21)
All right, folks. Well, thank you for watching.
Thanks for listening. Thanks for paying attention. As Greg said, email us with any SEO questions at joshbachynski@gmail.com.
And if you want to email Greg about any keyword spy specific technical questions, email Greg at hello@trykeywordspy.com. All right? Thanks for watching.
Make sure to click like and subscribe and click the bell, if you don't mind. And if you want to watch, live and ask questions and we'll see you next time.
Share this post