Can AI Do SEO: Using OpenAI’s GPT-3 for Experiment

Data Science, Machine Learning, Science and Technology, SEO

AI (Artificial Intelligence) technology has advanced dramatically in recent years. It is now feasible to measure its ability to do certain tasks such as text generation, image generation, and sound generation.

You can open Table of Contents show

What if we take it a step further and conduct more complex assessments, such as evaluating a job or, more specifically, evaluating an AI system on its capacity to perform SEO?

In this section, we will put OpenAI’s Generative Pre-trained Transformer 3 (GPT-3) to the test.

Remember that an AI system will mirror the data on which it was taught.

SEO has evolved in tandem with search engine advancement, and everything is widely documented in blogs, books, and interviews.

As a result, the AI will most likely learn from all of the available SEO content.

As a result, the tests will assess the following indicators:

Maintaining a consistent approach to an SEO topic.
The significance of SEO changes.
The ability to respond to SEO questions.

The Consistency To Continue An SEO Topic

The next tests will focus on the AI system’s comprehension of an SEO issue as well as its ability to treat the topic intelligibly and consistently.

The first test will use instances of SEO horror stories – that is, short stories about SEO mishaps – to observe how the AI performs.

I picked two examples from Aleyda Solis’ website, and here’s what the AI came up with:

Two examples:

When all of the site’s internal pages were canonicalized to the home page. #seohorrorstories
When someone forgets to disable the site’s test environment… that began to be indexed #seohorrorstories

Results generated by OpenAI’s GPT-3 AI:

When someone said, “I have a fantastic idea!” Let’s put our site in the footer of a completely unrelated website!” #seohorrorstories
When someone fails to check the staging environment’s robots.txt before going live. #seohorrorstories
When someone changed the contents of.access without informing anyone #seohorrorstories
When we changed domains and failed to send the old pages for more than 60 days. #seohorrorstories
When the customer built hundreds of useless thin content pages #seohorrorstories
When the client insisted on utilizing relational database tables as URLs (?tbl=table&id=2)

Analysis: The texts produced are diverse and coherent.

Furthermore, after conducting some investigation, we can determine that they are one-of-a-kind.

As a result, the AI realized that it needed to duplicate a brief essay on SEO and include the hashtag #seohorrorstories every time.

Here’s a screenshot of the entire example; we can create an infinite number of #seohorrorstories.

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.

For the next test, we will provide the AI with a scenario and instructions, and then ask it to construct a dialogue.

Here is a scenario in which the AI will have to facilitate a conversation between two specialists — one experienced and one inexperienced: When someone forgets to disable the site’s test environment… that began to be indexed

OpenAI’s GPT-3 AI produced the following results:

Oh, no! Aleyda: When did you know what was going on?
Paul: I just checked Google Analytics this morning and noticed that we were getting some traffic from google.com.
That’s not good, Aleyda. Checked to see if the site test environment was still being indexed.
Paul: I did, indeed. Yes, it is.
Aleyda: Okay, we need to have this taken care of right immediately. First, let’s see whether there are any pages that have already been indexed.
There are a few, Paul.
Aleyda: Excellent. Now we must prevent Google from indexing the test environment.
Paul: How do we go about doing that?
We can add a robots.txt file to the test environment, Aleyda. This instructs Google not to index any of the site’s pages.
Paul: All right, it makes sense. Let’s get started.

Analysis: The dialogue is still extremely coherent.

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.

Next, let’s choose one of AI’s #seohorrorstories and develop the conversation for it.

Here’s the current situation: the customer built thousands of useless thin content pages.

OpenAI’s GPT-3 AI produced the following results:

Aleyda: Well, this is a little difficult. You’ve made a lot of shallow content pages that don’t provide any value to users.
Paul: I understand. I was simply attempting to cram as many keywords as possible onto the website.
Aleyda: Unfortunately, that will not assist your SEO. In fact, it may have a negative impact on your rating.
So, Paul, what should I do?
Aleyda: You must provide higher-quality content that adds value to users’ lives. That is what will assist you in ranking better in search engines.

Analysis: Once again, the AI manages to start a real conversation by explaining why providing thin material is bad for SEO and even suggesting solutions.

We can deduce that the AI system can speak rigorously and accurately about SEO, but can it also apply relevant SEO changes to a text?

GPT-3, on the other hand, features a text editing mode that has been accessible since March 2022.

Let us begin testing.

The Relevance Of SEO Modifications

In the first test, the AI will be asked to replace a link anchor with something more marketable in order to better emphasize a product.

OpenAI’s GPT-3 AI produced the following results:

The test was relatively successful, according to the analysis.

Given the context, the AI replaces “mahogany-paneled steakhouses” with “greatest steaks in New York.”

The solution is applicable regardless of the number of linkages.

The second test consists of accomplishing two tasks at the same time: asking the AI to alter the link anchor while also inserting an H1 tag with a title to the beginning of the document.

OpenAI’s GPT-3 AI produced the following results:

Analysis: Once again, the test is successful, and the results may be applied to hundreds of papers. In the appropriate spot, a title that summarizes the paper is appended.

Finally, we’ll instruct the AI to insert a list of three relevant links between the ul>li> tags.

Analysis: The format is honored once more, and the anchors are consistent.

The links are fake, but GPT-3 could easily use a link database to select real ones.

In short, when given the proper instructions, GPT-3 is capable of contextually modifying text. When it comes to large-scale changes, this could be a terrific time-saving technique.

The Ability To Answer SEO Questions

Now we’ll put an AI system’s SEO understanding to the test.

I picked 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 like we did in the prior tests. The quantity of right answers determines the test’s quality.

The image below shows how each question is presented to the AI; the text highlighted in green is the answer it has selected.

BrightEdge gives a table of questions with correct answers in one column and the AI’s choice in the adjacent column:

The GPT -3 findings are encouraging: the AI receives a score of 16/20, but even so, it lacks context for where it goes wrong.

Now, question the AI about the most recent Google updates surrounding the launching of Google MUM on May 18, 2021.

What is the primary goal of Google MUM?

Again, as you can see, the AI responds flawlessly.

In short, if we ask very targeted inquiries, we have a real SEO response engine.

The Limitations

We may now look at the limitations of AI-powered SEO.

The following is a non-exhaustive list:

The AI’s capability to fix unknown or poorly documented problems.

When an issue happens in the SEO industry, it is sometimes time-consuming and tiresome to figure out what happened, and occasionally there is no answer. AI will not be able to identify solutions if the data it uses is not part of its analysis.

The ability to take into account everything that exists.

There are frequently numerous limits on what has previously been done or tested on a site or set of sites. As a result, it is currently difficult to supply all of the source code and project documentation to an AI system in order for it to obtain an accurate answer that takes into account all of the available information.

Being able to measure the true quality of content or an idea by its originality or its representation of the current situational context.

AI systems are still not real-time and lack novelty and inventiveness. As a result, the genuine measure of content quality is now minimal. Humans are frequently used to assess the quality of content. AIs are generally capable of detecting spam and low-quality content.

To summarize, while an AI system can outperform a beginner SEO or a person with limited SEO knowledge, it will still be a long way from replacing an experienced SEO specialist.

Overall, the AI-human combination will always be the best combination, with the human-focused on jobs with the highest added value.

To summarize, all seasoned SEO experts are considerably superior to an AI SEO.

Apart from this. if you are interested, you can also read Entertainment, Numerology, Tech, and Health-related articles here: Yellowstone season 5, Ozark season 4 part 2, PUBG Redeem Code, Is Windows 11 Free, KProxy, How to play Xbox Games on PC, Ozark Season 5, How to cancel YouTube TV, Churchill Car insurance, The Rookie Season 5, Downloadhub, Ssr Movies, 7starhd, Movieswood, How to Remove Bookmarks on Mac, Outer Banks Season 4, How to block a website on Chrome, How to watch NFL games for free, DesireMovies, How to watch NFL games without cable, How to unlock iPhone, How to cancel ESPN+, How to turn on Bluetooth on Windows 10, Outer Banks Season 3,