Is AI capable of performing SEO, and how does it stack up against the labor of seasoned SEO experts? Check out the results of this OpenAI GPT-3 experiment.
Artificial intelligence (AI) technology has advanced dramatically in recent years. It is now feasible to evaluate its ability to do certain jobs such as text, image, and sound generation.
What if we take it a step further and conduct more complex tests, such as evaluating a job or, more specifically, analyzing an AI system’s capacity to perform SEO?
We’ll put OpenAI’s Generative Pre-trained Transformer 3 (GPT-3) to the test.
Please remember that an AI system will imitate the data it was trained on.
SEO has evolved in lockstep with the evolution of search engines, and everything is widely recorded in blogs, books, and interviews.
As a result, the AI will most likely have learned from all of the SEO information available.
As a result, the tests will assess the following factors:
• Maintaining a consistent SEO topic.
• The significance of SEO changes.
• The ability to respond to SEO-related questions.
The Consistency Required To Maintain An SEO Topic
The AI system’s capacity to comprehend an SEO circumstance and address the topic rationally and consistently will be tested in the following tests.
The AI will be put to the first test by using instances of SEO horror stories – that is, short stories about SEO mishaps – to evaluate how well it works.
Here’s what the AI came up with using two examples from Aleyda Solis’ website:
Here are two examples:
• When the site’s internal pages were canonicalized to the home page. #seohorrorstories
• When someone failed to block the site’s test environment, it became indexed. #seohorrorstories
OpenAI’s GPT-3 AI produced the following results:
• When someone said, “I’ve got a brilliant idea!” Let’s put our site in the footer of a completely unrelated site!” #seohorrorstories
• When someone forgot to check the staging environment’s robots.txt before going live. #seohorrorstories
• When someone made.htaccess tweaks without informing everyone #seohorrorstories
• When we changed domains and didn’t send the previous pages for over 60 days. #seohorrorstories
• When the customer built thousands of worthless thin content pages #seohorrorstories
• When the customer insisted on using?tbl=table&id=2 as a URL for relational database tables #seohorrorstories
Analysis: The texts produced are diverse and well-organized.
Furthermore, after some investigation, we can determine that they are one-of-a-kind.
In terms of inventiveness, we may take our testing a step further and try to quantify the consistency of developing a discussion between two SEO specialists.
We’ll give the AI a situation and instructions for the next test, and then ask it to construct a dialogue.
Here’s how the AI will have to generate discourse between two specialists, one seasoned and the other inexperienced: When someone forgot to block the site’s test environment, it became indexable.
OpenAI’s GPT-3 AI produced the following results:
Oh oh, Aleyda! When did you know something was wrong?
Paul: This morning, I checked Google Analytics and noticed that we were receiving some traffic from google.com.
That’s not good, Aleyda. Have you double-checked that the site test environment is still being indexed?
Paul: Of course, I did. Yes, it is.
Aleyda: All right, we need to have this taken care of right immediately. Let’s start by checking to see whether any pages have been indexed.
There are a few, Paul.
Aleyda: That’s fantastic. Now we must block the test environment so that it is not indexed by Google.
Paul: How do we go about doing that?
Aleyda: In the test environment, we can add a robots.txt file. This tells Google not to index any of the site’s pages.
Paul: Okay, I see what you’re saying. Let’s get started.
Analysis: The conversation is still pretty well-organized.
The AI system even discusses how the problem was discovered and suggests a solution for indexing a pre-production site with a robots.txt file.
Let’s now take one of the #seohorrorstories created by AI and build the dialogue that goes with it.
The new situation is as follows: When the client built thousands of useless thin content pages.
OpenAI’s GPT-3 AI produced the following results:
Aleyda: Well, there’s a problem here. You’ve made several pages with thin content that doesn’t provide much value to users.
Paul: I understand. I was simply attempting to cram as many keywords as possible onto the site.
Aleyda: Regrettably, that will not boost your SEO. It may be detrimental to your rating.
Paul: So, what should I do?
Aleyda: You must provide higher-quality material that is useful to users. That is what will assist you in achieving a higher search engine ranking.
Analysis: Once again, the AI manages to start a meaningful conversation by explaining why thin material isn’t good for SEO and even offering suggestions.
We can deduce that the AI system can speak fluently and accurately about SEO, but can it also apply meaningful SEO changes to a text?
GPT-3, on the other hand, has had a text editing mode available since March 2022.
Let’s get started with the testing.
The Importance Of SEO Changes
In the first test, the AI will be asked to substitute a link anchor with something more marketable to better emphasize a product.
The test was relatively successful, according to the results.
Given the context, the AI replaces “mahogany-paneled steakhouses” with “greatest steaks in New York.”
Regardless of the number of linkages, the solution works.
The second test entails completing two activities at the same time: Requesting that the AI alter the link anchor while also putting an H1 tag with a title at the top of the document.
Analysis: The test was successful once again, and the results may be applied to hundreds of papers. In the appropriate spot, a title that summarises the paper is appended.
Finally, we’ll use the ul>li> tags to ask the AI to add a list of three relevant links.
Analysis: The format is respected once again, and the anchors are consistent.
The links are fake, but genuine links may easily be selected from a link database using GPT-3.
When given the proper instructions, GPT-3 is capable of contextually modifying text. In the case of large-scale changes, this might be a wonderful time-saving technique.
Ability to Respond to SEO Questions
Now we’ll put an AI system’s SEO understanding to the test.
I went with BrightEdge’s SEO test, although any test would have sufficed:
We used the GPT-3 OpenAI technology to answer the questions in an SEO MCQ, just as we did in the prior tests. Test quality is determined by the number of correct answers.
The AI is given with each question as shown below; the text highlighted in green is the answer it has selected.
The Limitations
We may now look at the AI-driven SEO’s limits.
The following is a partial list:
The AI’s ability to solve challenges that are unknown or inadequately documented.
When an issue happens in the SEO industry, it can be time-consuming and laborious to figure out what happened, and sometimes there is no answer. With data that isn’t part of the study, AI won’t be able to find solutions.
The ability to consider everything that is available.
What has already been done or tested on a site or a set of sites is frequently constrained. As a result, it is currently impossible to supply all of the source codes and project documentation to an AI system for it to provide an accurate answer that takes into account all of the available data.
Being able to assess the genuine quality of a piece of material or an idea based on its originality or portrayal of the current situation.
AI systems lack originality and innovation since they do not work in real-time. As a result, the genuine measure of content quality is now negligible. Humans are frequently the ones who determine whether or not a piece of material is of high quality. Artificial intelligence can usually recognize spam and low-quality content.
To summarise, an AI system can outperform a beginner SEO or someone with limited SEO knowledge, but it will still have a long way to go before it can fully replace a seasoned SEO specialist.
Overall, the AI plus human team will always be the best combination, with the human concentrating on the tasks that provide the most value.
Finally, all seasoned SEO experts are considerably superior to an AI SEO.
