Can AI perform SEO? OpenAI’s GPT-3 Experience

Artificial intelligence (AI) technology has made tremendous progress in recent years. It is now possible to assess their ability to perform specific tasks such as creating text, images and sound.

Now, what if we go further with more complex tests, like evaluating a job, for example, or more specifically, evaluating an artificial intelligence system based on its SEO ability?

Below, we will test the Generative Transformer 3 (GPT-3) generated by OpenAI.

Let’s keep in mind that the AI ​​system will simulate the data on which it is being trained.

SEO was built in tandem with the advancement of the search engine, and everything is well documented in blogs, books, and interviews.

Therefore, it is possible that the AI ​​has learned from all the SEO content available.

So the tests will measure the following indicators:

  • Consistency in pursuing the topic of SEO.
  • The importance of search engine optimization tweaks.
  • Ability to answer search engine optimization questions.

Consistency in pursuing the topic of search engine optimization

The following tests will focus on an AI system that understands the SEO situation and its ability to approach the topic clearly and consistently.

The first test will take examples of SEO horror stories – short stories about SEO disasters – and see how AI is performing.

I took two examples from the Aleyda Solis website and here’s what’s generated by artificial intelligence:

Two examples:

  • When all the internal pages of the site have been translated to the home page. # horror stories
  • When someone forgot to block the site of the test environment… I started indexing it. # horror stories

Results from OpenAI’s GPT-3 AI:

  • When someone said, “I have a great idea! Let’s put our site in the footer of an irrelevant site!” #Horror Stories
  • When someone forgot to check the robots.txt file for the staging environment before going live. # horror stories
  • When someone made changes to .htaccess without telling anyone #seohorrorstories
  • When we switched domains and didn’t redirect pages older than 60 days. # horror stories
  • When a customer created thousands of pages of thin content with no value #seohorrorstories
  • When the client insisted on using relational database tables as URLs (?tbl = table & id = 2) #seohorrorstories

Analysis: The resulting texts are diverse and coherent.

Moreover, after a little research, we can conclude that it is unique.

So, the AI ​​understood that it had to simulate a short text talking about SEO and add the hashtag #seohorrorstories every time.

This is a screenshot of the full example; We can create endless #horror stories.

Screenshot from OpenAI, April 2022

In terms of creativity, we can go a little further with our tests and try to gauge the consistency of creating a dialogue between two SEO experts.

For the next test, we’ll give the AI ​​a position and instructions, and then ask it to have a discussion.

This is the situation in which the AI ​​has to create a dialogue between two experts – one an experienced and one a novice: when someone forgot to block the site’s test environment… that started indexing.

Results from OpenAI’s GPT-3 AI:

Aaliyah: Oh no! When did you realize this was happening?
Paul: I just checked Google Analytics this morning and saw that we had some traffic coming from
High: That’s not good. Have you checked to see if the site’s test environment is still being indexed?
Paul: Yes, I did. and he.
Aaliyah: Well, we need to take care of this immediately. First, let’s check to see if there are any pages that have been indexed.
Paul: There are a few.
Aley: Awesome. Now we need to block the test environment until Google stops indexing it.
Paul: How do we do that?
High: We can add a robots.txt file to the test environment. This will tell Google not to index any of the pages on the site.
Paul: Well, that makes sense. Let’s do it.

SEO Dialogue Generated by GPT-3Screenshot from OpenAI, April 2022

Analysis: The dialogue remains very coherent

The AI ​​explains how to detect the problem and suggests a solution to a pre-production site indexing problem using robots.txt.

Next, let’s take one of the #seohorrorstories invented by AI and create the dialogue associated with it.

Here’s the new situation: when the client created thousands of pages with thin content without value.

Results from OpenAI’s GPT-3 AI:

High: Well, that’s somewhat of a problem. I’ve created a lot of thin content pages that don’t really offer any value to the users.
Paul: Yes, I know. I was trying to stuff as many keywords as possible on the website.
High: Unfortunately, this will not help you with your SEO. In fact, it can actually hurt your ranking.
Paul: So what should I do?
High: You need to create high-quality content that provides value to your users. This will help you rank higher in the search engines.

Results generated by artificial intelligenceScreenshot from OpenAI, April 2022

Analysis: Once again, the AI ​​was able to create real dialogue by explaining why creating thin content with SEO didn’t work and even recommending solutions.

We can conclude that an AI system can talk about SEO rigorously and accurately, but can it also make relevant SEO adjustments to text?

As it happens, GPT-3 offers a text editing mode that has been available since March 2022.

Let’s try.

Importance of SEO tweaks

The first test would ask the AI ​​to replace the anchor with something more marketable to better highlight the product.

Results from OpenAI’s GPT-3 AI:

SEO version with GPT-3Screenshot from OpenAI, April 2022

Analysis: The test is relatively successful.

Here, AI replaces “mahogany-paneled steak” with “best New York steak,” given context.

The solution works regardless of the number of links.

Then, the second test consists of completing two simultaneous tasks: asking the AI ​​to change the link anchor while adding an H1 tag with an address to the beginning of the document.

Results from OpenAI’s GPT-3 AI:

AI generated results for H1 . tagScreenshot from OpenAI, April 2022

Analysis: The test is successful again, and you can apply the results to thousands of documents. A title that summarizes the document is added in the appropriate place.

Finally, we will ask the AI ​​to add a list of three related links using tags

  • .

    Create AI results to add relevant linksScreenshot from OpenAI, April 2022

    Analysis: Again, the format is well respected and the anchors are consistent.

    The links are fake, but GPT-3 can easily be selected from the link database to use real links.

    In short, GPT-3 is good at editing text by context when given the right instructions. This can be a huge time-saving measure for large-scale adjustments.

    Ability to answer SEO questions

    We will now test the SEO knowledge of the AI ​​system.

    I chose to use BrightEdge’s SEO test, but I could have used any test:

    As with each of the previous tests, we used GPT-3 OpenAI technology to answer questions in SEO MCQ. The number of correct answers measures the quality of the test.

    AI chooses the best answerScreenshot from OpenAI, April 2022

    Below, we see how each question is presented to the AI; The text highlighted in green represents the answer he chose.

    BrightEdge provides a table of questions with the correct answers in one column and the answer chosen by the AI ​​in the next column:

    AI can answer your SEO testShot by author, April 2022

    GPT-3 results are encouraging: AI scores 16/20, and even then, it lacks context in which to go wrong.

    Now, let’s ask AI about the latest updates from Google regarding the release of Google MUM on May 18, 2021.

    google mom questionScreenshot from OpenAI, April 2022

    What is the primary purpose of Google MUM?

    Again, you see, AI answers perfectly.

    In short, we have a real SEO response engine if we ask very specific questions.


    Now, we can take a look at the limitations of AI-powered SEO.

    Here is a non-exhaustive list:

    The ability of AI to fix unknown or poorly documented problems.

    Oftentimes, in the SEO field, understanding what really happened when a problem occurs is time consuming and tedious, and sometimes there will be no explanation. AI cannot find solutions with data that is not part of its analysis.

    The ability to take into account everything that exists.

    There are often many limitations to what has already been done or tested on a site or group of sites. Therefore, it is currently impossible to provide all the source codes and project documentation for the AI ​​system to find an accurate answer that takes into account all the existing information.

    The ability to measure the true quality of a content or idea by its originality or its representation of the current situational context.

    AI systems don’t work in real time yet, and they lack originality and creativity. As a result, the true measure of content quality is currently minimal. Quality content is often judged by humans. AI systems generally detect spam and poor quality content.

    To sum up, an AI system can outperform a novice SEO or someone with little knowledge of SEO, but still have a long way to go to replace an experienced SEO expert.

    In general, the duo of AI and humans will always be the best combination, with humans focusing on tasks with the highest added value.

    In conclusion, all the seasoned SEO experts are far above AI SEO.

    More resources:

    Featured Image: Shutterstock / metamorworks

Leave a Comment

Your email address will not be published.