The excitement and confusion over AI, primarily led by the text-drive Chat GPT, is running wild. Some are calling it an opportunity to replace all their writers and others are a little more conservative about their predictions, but who’s right and what do you really need to know about AI before you let it loose in your business?
We’ve done a round up of all the latest information available on AI and we’ve made it relevant to the context of using it for writing and developing content for your business. We’ll leave it to you to decide what’s the best course of action for your company.
What Is Artificial Intelligence?
Artificial intelligence is not, yet, intelligent. Instead, it is a form of software that relies on advanced pattern recognition algorithms. The bigger the data set that AI can review, the more patterns that it can discover and the more it can use those patterns to create works of art, written material, etc.
This may feel a lot like “intelligence” given that human beings also have advanced pattern recognition capabilities when compared to many other creatures. For example, if we see a flash of orange and black when walking through a jungle, we don’t have to see a whole cat to think “tiger”.
That’s pattern recognition at its simplest. The ability to infer something from limited data sets.
What’s The Biggest Limitation On Artificial Intelligence?
The biggest limitation on all artificial intelligence systems that rely on pattern recognition is the data set that they are trained on. For example, Chat GPT, is currently being trained off a snapshot of the Internet from 2021. It is unable to access the internet of today to update its data set.
That means Chat GPT is effectively blind to anything that happened after the cut off date in 2021. It has no idea about what’s going on in the world, right now, if scientific or political developments have taken place, it can’t tell and any other errors that were online that have now been corrected are still errors for Chat GPT.
At the moment this means that Chat GPT can’t tell what is really true, only what it infers to be true, and that means AI is as likely to tell you the wrong answer in a confident fashion as it is to tell you the right one.
What’s The Second Biggest Limitation On Artificial Intelligence?
We’re not going to spend a lot of time on this, but it’s worth noting that AI has one other major weakness when it comes to generating accurate results – the programmed biases in the software.
Chat GPT, for example, was famously happy to compose a poem praising Pol Pot, but refused to do so for Chairman Mao. This isn’t a limitation of the AI or the AI deciding that Donald Trump is a bad person, it’s been hard coded by the owners of Open AI (the system behind Chat GPT).
This is, actually, more dangerous than the first issue. You can overcome dataset limitations by fact checking, but there are proposals afoot to censor the web using AI. If that becomes common practice, anything that disagrees with the AI owner’s bias will simply not exist online.
You probably aren’t a fan of Chairman Mao or Pol Pot, we’re certainly not, but you can probably see the danger of an internet which pretends that one person is always bad and the other is always good. We suspect that this is going to be a topic of much debate in the coming years.
Can You Use AI To Generate Content For Your Business?
Yes. That’s the short answer. You’re free to use AI content for your business as long as the terms and conditions of the AI allow you to do so. So, for example, if you’re using GhatGPT then the answer is always going to be “yes”.
However, that doesn’t mean that you should use AI in your business and to really examine that more closely, we’re going to take a look at three important areas. The first is Search Engine Optimization (SEO) and the second is copyright, and third is your reputation.
AI And SEO Content (What You Need To Know)
There is no doubt that many of the SEO gurus have seen AI as a way to dramatically cut costs on their writers and many of them are already selling SEO writing services to their clients.
Now, before you join them, we’d encourage you to consider the following:
Google And AI Is A Huge Unknown
Some people claim to be ranking AI content with ease, others have seen it rank for a short time and then seen Google shut it down. Some point at Google’s “useful content” rule and claim that as long as they edit their AI copy, this means it’s fine to use and will always rank.
To this we say, maybe, maybe not.
No company other than Google can tell you what’s going on at Google. Google’s business model involves serving up web pages and selling advertising on them. That means Google needs not just well written pages that lead to a good user experience but it also needs human beings to write those pages.
Well, Chat GPT, the Internet’s prime AI writing tool (it’s the basis for nearly all writing tools such as Jasper.Ai too) is an aggregate of the knowledge it’s stealing from the internet. If all sites are written by AI? Then, there’s no incentive for human beings to create fresh content.
Any errors out there will never be corrected. New developments in fields are going to go unreported. And so on… over time, an AI built Internet is going to suck. So, Google is going to need to incentivize human content creation if Google wants to stay in business.
The Possible Google Penalties For AI
And when Google decides to smack, anything that it deems to be poor, content around it can do it in one of two ways:
It can penalize the individual piece of content. Think of how Google currently treats spam links – if your site is on the receiving end of such links, it reduces the site’s rankings in search. This prevents site owners from creating trash content to create cheap links to game Google’s algorithm.
It can go scorched Earth. The more AI content that you have, the more likely it is that if Google decides to penalize your content, that it just penalizes the whole site. In that it removes your site from the Google index, forever.
That means each piece of AI content you upload for SEO is potentially going to kill your site in the long-term. There’s no coming back from a sitewide penalty either, you’d need to start a brand new site from scratch.
We question whether it’s worth risking the reputation that you’ve already built for this.
Google Has Already Hurt Some Sites With Large Volumes Of AI Content
You’ll find a lot of people online bragging about how awesome SEO content is and how everyone should be using it. What they all seem to have in common is that they are selling SEO content or “SEO Get Rich Quick” schemes.
The truth is that while we believe some sites are doing well with AI content others have been hit hard by Google over it. Kevin Indig writing on Medium cites the example of a site “throughtheclutter.com” which had nothing but AI content on it and which was rapidly ranked by Google when that content was published.
In fact, for a few months it did brilliantly, and it peaked at 1 million views a month. Then Google’s algorithms were clearly updated to seek out AI text and it went from 1 million views a month to next to zero views a month.
Mark Williams-Cook, another SEO expert saw a similar issue on a site, undisclosed to the public, which he’d published 10,000 pieces of AI content on. The traffic spiked rapidly and then collapsed to zero.
We would note, however, that this content was published extremely rapidly – this alone may be enough of a signal to allow Google to realize you’re using AI to generate content. Normal writers don’t slap 10,000 pieces of content together in a month.
On The Other Hand, Some AI Content Is Ranking Well In Google
Yes, there are a bunch of dodgy SEO merchants out there trying to flog AI as the best solution for all your content’s needs but there is some evidence that AI content does pretty well in Google.
However, there seem to be some common themes in this content:
The volume of AI content isn’t on a massive scale (this reduces the chance of detection and the chance of a site-wide penalty)
The content is of a higher quality than in the previous two examples (it can still be pure AI, but tends to consist of work with much more attention paid to the prompts used and the scale at which the AI model is deployed)
The content appears in low competition niches (the higher the competition, the more Google pays attention, good luck ranking number one in Google for “auto insurance quotes” with AI content, for example).
The content is typically shorter (if you use AI to write a 3,000 word essay, we guarantee it will read like it was written by AI, if, on the other hand, you let it write a 100 word answer to a brief question? It will be much harder to detect).
Google hasn’t caught them. Well, doh, right? Of course, Google may not be looking for them but we suspect that it is.
The implications of these two things are quite major. If you want to fight Google on this battlefield, you are entering the territory of Black Hat SEO and as with every major Black Hat SEO win of the last two decades, eventually Google is going to catch up with you and you will be penalized.
We would recommend if you wanted to create large volumes of AI material for SEO that you start a separate website for it. Continue with your own website creating content as you always have and use the AI site to create business leads, sales, etc. separately from the main site.
Then, when Google does catch up and applies a sitewide penalty to the AI content? You won’t have to worry about losing all of your business, just the part built under the Black Hat SEO techniques.
Google Speaks Out On AI
Google Search Liaison (the official search Twitter account) Tweeted, “As said before when asked about AI, content created primarily for search engine rankings, however it is done, is against our guidance. If content is helpful & created for people first, that’s not an issue.”
Then, they published a whole blog post on AI and the Google experience on February 8th, 2023.
It doesn’t say much more than that original Tweet, mind you, but it’s nice to know they’re trying, right?
This has made a lot of SEOs believe that they’ve found a magic permission slip from Google to use as much AI as they like as long as they deem it “helpful”.
The trouble with this is twofold:
Google’s definition of “helpful” evolves constantly, we’d be willing to bet money that your definition of “helpful” and theirs not only varies today but over time as more AI content floods the web, that they diverge dramatically.
Google has a history of saying one thing and then, unapologetically, doing something entirely different.
And as many internet users have been quick to point out, “Everybody wants to use AI content, but do you know what? Nobody wants to read it.”
AI content is thin. Really thin content. It makes sense when you read it through but that’s because mainly it doesn’t say very much at all.
If you put AI content head-to-head with handwritten content? (The guys at Income School did this for you). It’s very, very clear what was written by a human being and what was written by computer.
AI Content Detection Is Coming
There are already AI content detection tools online. The only trouble is that they’re all a bit on the garbage-side at detecting AI content. The scores they offer are a programmer’s best guess and nothing categoric.
They certainly don’t relate to how Google views your content. It’s also worth noting that Google has almost certainly built its own AI detection tools and these are almost certainly better than the content detectors on the market. This is because Google has more money than all those detection companies combined. They may never allow public access to these tools, either.
Worst of all, Chat GPT addicts, watermarking is coming. Open AI has stated that in the next release, all Chat GPT text will contain a “fingerprint” marking it as AI generated. This will be done in a way that involves word usage but one that can’t be detected by the human eye.
The only way to avoid this fingerprint would be to rewrite all the text in your own words, which kind of defeats the purpose of the AI writing for you in the first place.
This has some pretty stark implications for using AI in your Black Hat SEO world too – it means you can’t rely on running your AI generated material through an AI content detection software and assuming a low score means Google can’t detect that AI.
It means that contrary to popular belief, editing and fact checking AI generated content simply won’t be enough in the long run to avoid being caught by the fingerprint of watermarking.
You should assume that some watermarking is already taking place even if the official launch of that service hasn’t arrived because it probably is.
We would, however, recommend running any content that you buy for SEO purposes through an AI detector. This doesn’t mean that if it gets a high score that it has been AI generated but if it consistently trips alarms on the detector over time? You want to have a strong talk with your SEO content supplier – especially if they’re charging you for writing their own copy.
Does This Mean Avoid AI Completely For SEO?
No. It doesn’t. What it does mean is that you want to proceed with caution. Don’t forget that you’re the expert in your field, not an AI.
It makes sense to generate AI content for any boilerplate pages that you may have (privacy, t’s and c’s, etc.)
It also makes sense to use it sparingly to generate areas for further research, ideas for keywords (though be warned when tested against premium keyword generation tools ChatGPT does not perform very well), and even article structure.
What it doesn’t make sense to do, is to fill your site with cheap AI SEO material. The odds are pretty good that it won’t last long and could permanently damage your website.
If you really want to test large volumes of AI content, again, we strongly recommend creating a second site and separating it entirely from your current site. This means that when the inevitable penalties are applied, it won’t hurt your existing business.
AI And Copyright
The next hurdle to AI content is going to be even bigger than that for SEO and it’s the thorny issue of AI and copyright.
Copyright is the intellectual property which is used to protect the author or creator of a work in law. So, when a human being writes an article for your blog, they own the copyright, unless you buy that copyright from them in some form.
And the key phrase there is “human being”.
Copyright Ownership Isn’t For The Birds, The Monkeys or The AI
There is a famous case of Naruto, the monkey. David John Slater, a photographer, gave the monkey a camera and it snapped a selfie of itself. It’s one of the most famous images on the internet, in fact.
Slater claimed the copyright. PETA (the People for the Ethical Treatment of Animals) decided they would sue on Naruto’s behalf claiming that as the creator of the work, the monkey must own the copyright.
The courts threw out PETA’s claim. They noted that there was no obvious intention in the framing of the law for anything other than a human being to own the copyright of an image.
The image is still considered to be Slater’s copyright by Slater, himself, but many argue that the image has no copyright at all. After all, Slater did not take the photograph himself. Wikipedia falls into this “uncopyrightable” camp and continues to use the image even though Slater has demanded that they take it down.
We tend to agree with Wikipedia on this. Copyright law is clear that the act of creation must take place for you to own the copyright. Slater may have made the circumstances for the photo to be taken but he did not take the photo.
What Does That Mean For AI Content?
Well, it means that AI cannot own the copyright of any material that it creates. The office of copyright in the United States is clear that it will not allow copyright registration of material that is produced by a mechanical process and “without any creative input or intervention from a human author”.
But how much “creative input or intervention from a human author” is required to distinguish human work from AI work? Nobody knows. Copyright law is deliberately vague. It is meant to be tested in court. Nobody’s tested this, yet.
ChatGPT and OpenAI Content
OpenAI has this lovely statement in its terms; “Your Content. You may provide input to the Services (“Input”), and receive output generated and returned by the Services based on the Input (“Output”). Input and Output are collectively “Content.” As between the parties and to the extent permitted by applicable law, you own all Input, and subject to your compliance with these Terms, OpenAI hereby assigns to you all its right, title and interest in and to Output. OpenAI may use Content as necessary to provide and maintain the Services, comply with applicable law, and enforce our policies. You are responsible for Content, including for ensuring that it does not violate any applicable law or these Terms.”
There are some problems with this. Firstly, it’s not clear that OpenAI actually has any rights to transfer to you. Again, copyright belongs to humans not to software tools.
Secondly. OpenAI bases its answers on other people’s answers to the extent that they may go from “plagiarism” (which is unethical but still legal) to “copyright infringement”.
This is touched on its terms too: “Copyright Complaints. If you believe that your intellectual property rights have been infringed, please send notice to the address below. We may delete or disable content alleged to be infringing and may terminate accounts of repeat infringers.
3180 18th St
San Francisco, CA 94110
Attn: General Counsel / Copyright Agent
Written claims concerning copyright infringement must include the following information:
A physical or electronic signature of the person authorized to act on behalf of the owner of the copyright interest;
A description of the copyrighted work that you claim has been infringed upon;
A description of where the material that you claim is infringing is located on the site;
Your address, telephone number, and e-mail address;
A statement by you that you have a good-faith belief that the disputed use is not authorized by the copyright owner, its agent, or the law; and
A statement by you, made under penalty of perjury, that the above information in your notice is accurate and that you are the copyright owner or authorized to act on the copyright owner’s behalf.”
There is, however, nothing in its terms about OpenAI protecting its users against such copyright claims. And given that OpenAI is being trained on a copy of the Internet, it’s a guarantee that some of the material that it creates will infringe the copyright of others. Are you ready to be sued for that?
Thirdly, also in those terms of service you will find, “Due to the nature of machine learning, Output may not be unique across users and the Services may generate the same or similar output for OpenAI or a third party. For example, you may provide input to a model such as “What color is the sky?” and receive output such as “The sky is blue.” Other users may also ask similar questions and receive the same response. Responses that are requested by and generated for other users are not considered your Content.”
Yes, you’re reading that correctly. Open AI may be generating identical content to other people’s content and that, from their perspective is fine, they’re not asserting any rights over the content and they’re giving it away and yes, that means they can give it to more than one person.
That means you and anybody else with that content can use it on your site. Going back to Google SEO for a minute, Google doesn’t like repeated content particularly if it’s unaccredited, but how will you know that your AI content is repeated unless you search for all of it in Google? That’s going to be a time-consuming pain.
Worse, because there is no copyright on any of the content that you produce with OpenAI? Anyone can steal your content and use it in any way that they like. You will have no legal comeback at all.
AI And Your Reputation
This brings us to our last and shortest section on this new technology, but it’s important. The risk that the use of AI poses to your reputation.
Businesses live and die based on their reputation. For some businesses, particularly those selling products, their reputation is in Amazon reviews or star ratings on various other webstores.
For the majority of us, however, our reputation is online. When somebody Googles your name or your company’s name, what can they see?
Worse, what happens if they find your name and your AI content on other less salubrious sites? You won’t be able to do anything about this, after all, as the copyright will belong to nobody.
If it’s a bunch of stories about copyright theft and your long-standing website is nowhere to be seen because Google deindexed it, what is your reputation now?
Now, we’re not saying “never use AI”. In fact, we believe AI can make your life much easier carrying out certain tasks.
What we are saying is “do you want to bet your hard earned reputation on an AI software tool that won’t provide you with any legal or reputation protection if things go wrong?”
If you’re concerned about being left behind in a rush of AI content flooding the internet, again we strongly recommend building a separate website for this content. That way if everything remains beneficial you can reap the rewards, if it doesn’t and we don’t think that it is likely to remain beneficial, you’re insulating your main business from the fallout.
You can, however, deploy AI tools to assist with research before you start writing, you can use them to get a feel, for example, for how a specific person might craft a phrase relating to your topic (this can help create an extra-dimension to blog posts) and so on.
Scalezilla is ultimately a personal brand management company. We want our clients to thrive, but we don’t want them to gamble everything they’ve worked for over the years to win a quick but unsustainable advantage with AI.
Final Thoughts On Using AI Content For Your Business
We have been hugely impressed with the progress of AI content generation over the last 2-3 years. They have rapidly improved from sounding like someone who barely speaks English and who doesn’t understand the subject matter to providing clear and concise answers that are fully intelligible.
However, AIs still produce “thin content” which lacks any form of expertise or understanding of the content itself. They can produce answers that are factually incorrect. And the grammar in many AI generated posts leaves a lot to be desired.
This has led to many content companies claiming that all you need to do is grammar check and fact check AI content and it’s “good to go”. We’re not so sure that’s true.
We think that Google will ultimately rule out most AI content and that sites which depend on this content are likely to be deindexed. After all, this content is simply plagiarizing content that already exists on the internet, content that was created by real people for the benefit of others.
also extremely concerned about the copyright minefield. This is no
small thing. Copyright cases tend to involve large businesses raining
fire on small businesses as they are expensive to litigate.
What would you do if News International or CNN decided they wanted compensation for your AI content stealing theirs? The legal implications could put many smaller businesses into bankruptcy.
Most of all, though, we’re concerned about the reputational risks involved in content that you have no formal intellectual property rights for. It’s simply too big a risk to rely on AI writing tools at this moment in time.
Here at Scalezilla we’ll be happy to talk to you about creating rapid content generation using AI for a test site that is not your main site or how you can integrate AI into your regular content generation without creating substantial risks to your business.Scale