Figure 2: AI-written content showed positive impacts in a case study conducted by Nathan Gotch and showed negative impacts in a case study conducted by Glenn Gabe and Reboot .
Given this background, cmlabs decided to conduct a similar experiment comparing the performance of SEO articles written by AI and humans on Google search results pages using the keyword rank matrix .
The main goal of this experiment is to find out whether articles written by AI can rank higher on Google than articles written by humans.. The following are the stages of the experiment that we conducted.
Stages of AI vs Human Experimentation
All stages of the experiment were conducted fairly, so there was no distinction or special treatment for human-written or AI-written articles. We also applied the same guidelines to both types of content to avoid factors that could affect the performance of certain content.
1. Selecting the Website to Use
In this experiment, we selected six websites that each already have content assets and organic traffic with varying amounts. The six websites are tiketresmi.com , museumnusantara.com , inklusiuang.com , animalogi.com , teknohacks.com , and bundatimes.com .
In the future, each website will publish five SEO articles written by humans and five SEO articles written by AI with topics that have been tailored to the website’s niche .
The SEO team uses third-party SEO tools , such as Google Keyword Planner, Semrush, and Sequence Keyword Research Tool to conduct keyword research . This stage is a crucial stage because each keyword has its own level of competition that can affect its ranking on Google.
Therefore, we apply the following guidelines to ensure that each keyword, whether used in AI articles or human-written articles, has the same level of competition:
The presence of ChatGPT by OpenAI
Has attracted the attention of the wider community. Although ChatGPT is not the first type of AI in the world, since the release of this chatbot on November 30, 2022, it has succeeded in making various platforms compete to provide features supported by ChatGPT or based on AI.
We chose to use ChatGPT through a Chrome extension called AIPRM for ChatGPT. We also used several template prompts available in this extension. The goal is to help us ensure that all articles written by AI are of the same quality, thus preventing bias and inequality between AI articles that can affect their ranking performance on Google.
Previously, we also tested the template prompt to produce an article using the keyword ” informational ” and compared it with an article using the same keyword in the SERP ( search engine result page ).
Although Sarah had always wanted to start her whatsapp number database own bakery, she found it difficult to stand out in a crowded market. She made the decision to invest in a tailored WhatsApp number database aimed at local food fans after doing some research. She composed personalised notes introducing herself and asking them to a free sample of her pastries, all while feeling a little nervous.
Content Writing and Publication
The goal of this experiment was to see how AI-written and human-written articles performed in SERPs. So, we avoided or minimized the human touch in AI-produced articles as much as possible. However, we still made sure that all articles followed the same guidelines to be optimized evenly.
Following are the guidelines for writing the entire article:
Article writing uses the template prompt. Human written | 100% unique | seo optimized article” (including writing introductions , headings , and closings ).
Writing meta information using the template prompt “title + meta description + 10 hn tags”.
Furthermore, for the needs of cover images in articles. Both in human-written and ai-written articles, we use licensed images that we download from stock image provider sites .
After 30 articles by ai and 30 articles by. The content writer team have been completed. The articles are published on each website.
Creating a Dashboard in Looker Studio
The Data Team utilizes Looker Studio to How AirBnb and Uber help analyze the performance development of each article (based on URL) in SERP. Some of the matrices that will be analyzed are clicks ( organic traffic ) and impressions . This data will later be used as supporting data for the final results of the experiment.
6. Analyze and Track Keyword Rankings
The process of analyzing and tracking the ranking of each keyword was carried out by the SEO and Data teams for one month. As previously mentioned, the Data team used Looker Studio to obtain the necessary data.
Meanwhile, the SEO team uses Sequence Stats to track the keyword positions of each article. This keyword ranking tracking activity is carried out every day since the first day after the article is published (H+1 article is published). The average keyword rank of each view will also be automatically visible .
The following is a view to track keyword
Rankings from articles written by humans on the museumnusantara.com domain. We took this data on October 12, 202334%. This is a positive canada email lead thing because domainnusantara.com managed to occupy the third highest position of all competitors competing for market share for the same keywords.
Figure 3: View of Sequence Stats used to track keyword rankings on the museumnusantara.com domain. Where the average keyword rank of human-written articles is at 6.8 and the average keyword rank of AI-written articles is at 21.2.
Then, here is the view to track the keyword ranking of the AI article on the tiketresmi.com domain. We also took this data on October 12, 2023.
Figure 4: View of Sequence
Stats used to track keyword rankings on the tiketresmi.com domain. Where the average keyword rank of human-written articles is at 12.6 and the average keyword rank of AI-written articles is at 40.8.
AI vs Human Experiment Results
Figure 5: Results from an AI vs. human article writing experiment showing ranking changes over one month.
For 30 days, there are visible rank changes. That occur in each keyword.where on all five websites. Keywords from articles written by humans have a higher average rank than keywords from articles written by ai.
Here is the average rank of each keyword on. The six websites along with supporting data ( clicks and impressions ) on the last day of the experiment.
Then, from the data above, there is another
Thing that also caught our attention, namely the accumulation of User Engagement () which got positive results. Where human-written articles got more clicks (1302) than AI articles (233).
Human-written articles also managed to achieve higher total impressions than AI-written articles on four out of six websites.
During these 30 days, human-written a
Rticles also showed an increase in performance on the clicks and impressions matrix over time. Meanwhile, AI-written articles tended to stagnate or experience a decrease in performance.
Figure 7: Data from Looker Studio showing the performance of AI and human-written articles on the tiketresmi.com website. Where the performance of human-written articles tends to increase over time.