ChatGPT is an intelligent chatbot that can take directions and perform tasks such as writing articles. There are many issues to understand before making a decision on how to use it for content and SEO.
The quality of ChatGPT content is amazing, so the idea of using it for SEO purposes should be addressed.
Why ChatGPT Can Do What It Does
In short, ChatGPT is a machine learning model called Large Learning Model.
A major learning model is artificial intelligence that has been trained on large amounts of data that can predict what the next word in a sentence will be.
The more data it is trained on the more specific types of tasks it can perform (such as writing articles).
Sometimes great language types develop unexpected abilities.
Stanford University writes about the addition of training data that enabled GPT-3 to translate text from English to French, even though it was not properly trained for that task.
Major language versions such as GPT-3 (and GPT-3.5 which is under ChatGPT) are not trained to perform certain functions.
They are taught a variety of knowledge that they can then apply to other domains.
This is similar to how a person learns. For example, if a person has learned basic carving, they can use that knowledge to make a table even if that person has never been taught how to do it.
GPT-3 works like the human brain in that it contains all the information that can be used for many tasks.
A Stanford University article on GPT-3 explains:
“Unlike chess engines, which solve a particular problem, people are “generally” intelligent and can learn to do anything from writing poetry to playing soccer to writing taxes.
In contrast to most current AI systems, GPT-3 is approaching such intelligence…”
ChatGPT includes an extensive language model called InstructGPT, which is trained to get human input and long-term answers to complex questions.
This ability to follow commands allows ChatGPT to take commands to create an article on any topic and execute it in any specified manner.
It can write an essay within constraints such as word count and placement of specific topic points.
Six Things to Know About ChatGPT
ChatGPT can write an essay on any topic because it is taught on a variety of articles that are available to the public.
There are some limitations to ChatGPT that are important to be aware of before considering using it for an SEO project.
The main result is that ChatGPT is not reliable in providing accurate information. The reason it is wrong is that the model is only predicting which words should come after the last word in a sentence in a given sentence. It has nothing to do with accuracy.
It should be of great concern to anyone who is interested in creating quality content.
1. Programmed to Avoid Certain Kinds of Content
For example, ChatGPT is designed not to publish content on topics of violence, sex, and harmful content such as instructions for making an explosive device.
2. Unaware of Current Events
Another limitation is that it does not recognize any content created after 2021.
So if your content needs to be up to date and fresh then ChatGPT in its current form may not be useful.
3. Has Built-in Biases
The most important criterion for knowing is that you have been trained to be helpful, true, and harmless.
They are not just assumptions, they are unintentionally built into the machine.
It seems that the no-harm system makes the outcomes avoid injustice.
It’s a good thing but it also cleverly changes the text from what would otherwise be neutral.
So to speak one has to take the wheel and tell ChatGPT openly to drive in the direction you want.
Here is an example of how bias changes the results.
I asked ChatGPT to write a story in the style of Raymond Carver and one in the style of mystery writer Raymond Chandler.
Both stories had happy endings that were uncharacteristic of both writers.
In order to get an output that matched my expectations I had to guide ChatGPT in a very detailed way to avoid high endings and Carver-style endings to avoid solving the story because that’s how Raymond Carver’s stories played out.
The point is that ChatGPT is biased and one should know how to handle the results.
4. ChatGPT Requires Highly Detailed Instructions
ChatGPT requires detailed instructions in order to produce high-quality content that has a high chance of being original or taking a different point of view.
The more instructions it is given the more difficult it will be.
This is both a strength and a limitation of consciousness.
The few instructions in the content request make the output share the same as any other request.
As a test, I followed the question and the results posted by many people on Facebook.
When I asked ChatGPT the exact question the machine generated an original script that followed the same format.
The articles were different but shared the same structure and touched on the same subtopics but with 100% different words.
ChatGPT is designed to choose random words when predicting what the next word in a text should be, so it makes sense that it doesn’t fake itself.
But the fact that the same requests produce the same text confirms the limitations of simply asking “give me this.”
5. Can ChatGPT Content Be Identified?
Researchers at Google and other organizations have been working for years on algorithms to determine the effectiveness of artificial intelligence.
There are many research papers on the topic and I will mention one from March 2022 that used the results from GPT-2 and GPT-3.
The research paper is titled, Adversarial Robustness of Neural-Statistical Features in Detection of Generative Transformers (PDF).
The researchers were trying to determine what type of analysis could detect AI generated content that used algorithms designed to avoid detection.
They tried methods such as using BERT algorithms to replace words with synonyms, one added spelling, among other ideas.
What they found was that some of the AI statistics that created the text such as the Gunning-Fog Index and Flesch Index scores were important for predicting whether the text was made by a computer, even if that text used an algorithm designed to avoid detection.
6. Invisible Watermarking
Of particular interest is that OpenAI researchers have developed cryptographic watermarking that is useful in identifying content created through OpenAI products such as ChatGPT.
The latest article called attention to a conversation with an OpenAI researcher found on a video titled, Scott Aaronson Talks AI Security.
The researcher says traditional AI practices such as watermarking could turn into an industry standard in the same way that Robots.txt became the standard for crawling.
“… we have seen in the last 30 years that the big Internet companies can agree on a few criteria, either because of the fear of being judged, because they want to be seen as a responsible player, or something else.
Another simple example would be robots.txt: if you want your website not to be indexed by search engines, you can specify that, and the major search engines will respect it.
In the same way, you can imagine something like watermarking – if we can prove and prove that it works and that it is cheap and does not harm the quality of the output and does not need a lot of computation and so on- that. it would become an industry standard, and anyone who wants to be considered a responsible player will include it. “
The watermarking developed by the researcher is based on cryptography. Anyone with the key can test the document to see if it has a digital watermark that indicates it was created by AI.
The code may be in the form of punctuation or word choice, for example.
He explained how watermarking works and why it is important:
“My biggest project so far has been a tool to check watermark output for text formats like GPT.
Basically, whenever GPT generates some long characters, we want there to be some invisible secret symbol in its choice of words, which you can use to show later that, yes, this came from GPT.
We want it to be very difficult to take a GPT output and pass it off as if it came from a human.
This can help prevent academic fraud, obviously, but also, for example, a lot of fake generation—you know, spamming every blog with seemingly high-profile messages that support Russia’s attack on Ukraine, without even a building full of trolls in Moscow.
Or impersonating someone else’s handwriting to accuse them.
These are all things that one might want to complicate, right? “
The researcher shared that watermarking overcomes algorithmic attempts to avoid detection.
But he also said that it is possible to overcome watermarking:
“Now, all this can be overcome with enough effort.
For example, if you used some AI to interpret the results of GPT-well, we will not be able to see it. “
The developer announced that the goal is to release watermarking in future releases of GPT.
Should You Use AI for SEO Purposes?
AI Content is Detectable
Many people say that there is no way for Google to know if the content was created using AI.
I don’t understand why anyone would make that point because seeing AI is a really cracked problem.
Even content that uses anti-detection algorithms can be viewed (as mentioned in the research paper I linked above).
Searching for machine content has been the subject of research for many years back, including research into how to recognize content that has been translated from another language.
Autogenerated Content Violates Google’s Guidelines
Google claims that the AI created violates Google’s guidelines. So it’s important to keep that in mind.
ChatGPT May at Some Point Contain a Watermark
Finally, an OpenAI researcher said (a few weeks before the release of ChatGPT) that watermarking was “hopefully” coming in the next version of GPT.
So ChatGPT can at some point be enhanced with watermarking, if it is not already watermarked.
The Best Use of AI for SEO
The best use of AI tools is to measure SEO in a way that makes the user more efficient. That usually involves letting the AI do the tedious work of searching and analyzing.
Shortening web pages to create a meta description can be an acceptable use, as Google specifically states that is not against its guidelines.
Using ChatGPT to create an article or summary of content can be an interesting use.
But assigning content creation to AI and advertising it as-is may not be the best use of AI for many reasons, including the possibility of it being seen and making the site receptive to manual (restricted) activity.
Image courtesy of Shutterstock/Roman Samborskyi