Using AI Detection in the Guest Post Link Building Process

At The Links Guy, we have a particular use case for Originality.ai — as an AI checker for our guest post strategy process.

Guest posting is just one of the many tactics we use, but it's important and effective. It has many moving parts. The most crucial and challenging part for us is writing articles that are of a high enough quality to be accepted by the editors of those websites.

So, in 2023, we made a strict company policy not to use AI at all within our content writing process. We also implemented strict editorial and AI-checking at the recruitment stage. All content writers who worked with us had to apply via an AI-proctored exam before they were selected for an interview and hired.

The Landscape of AI and Link Building

A particular concern for link building is that when you try to publish articles on high-quality sites, they will not accept poor-quality articles, including articles that they deem to be AI-generated. 

While the goal is to secure a backlink as credit, we still need to provide valuable content that serves the website audience’s interest.

So there are two reasons we could not allow AI to infiltrate our guest post-writing process:

1. AI-generated content tends to be very generic and skims the surface of a topic

Many of the publications and blogs we wrote for needed something with sufficient depth that was injected with human expertise or experience.

2. Google has released algorithm updates since late 2023 and into early 2024

We don’t know if AI detection is being used as part of Google’s process of checking content quality. However, the internet has a content quality issue that Google needs to solve to address the quality of search results. 

After these updates, there were industry reports about websites churning out large amounts of content (often using complete or heavy AI automation) that were devalued and took a large traffic hit. 

Now, as per Originality.ai’s study about the prevalence of AI within Google’s search results, there has been a rise in AI content, even in the top 20 ranking results (11.3% as of Apr 22, 2024). 

Even after ChatGPT was released to the public, there was not as large of a spike in AI content in search results. This likely indicates an active effort from Google to dampen the amount of AI content infiltrating the results.

As part of our guest article process at TLG, we write content that will be published on third-party websites. By choosing not to use AI, we aren’t adding to that issue. 

We contribute to a site’s ability to perform better in organic search results by providing valuable content for users. As a result, the article can attract more traffic and engagement. 

Since 2023, we have also found an increasing number of editors specifically state in their editorial guidelines that AI-generated articles are not allowed. It’s clear this topic is top of mind for many content managers and editors.

Summary of Key Findings

This shows the percentage of all guest post articles that were either “Rejected” or “Unused.” 

• Articles “Rejected” by the editor (our point of contact on that publication’s side) were posts where the editor confirmed in writing that they would not run the article due to not meeting quality requirements. 
• The status of “Unused,” was where the conversation dropped off, and we were unable to get hold of that point of contact (or anyone else at the publication) to discuss that particular article. We assumed that this indicated they were unhappy with the article quality, did not want to discuss it further, and may have suspected AI usage.

• Pre-AI check is the three-month period before we brought in Originality.ai as an AI checker in our sign-off process. 
• Post-AI check is a comparable three-month period after Originality.ai was brought in.
• Pre-AI check, our combined “Rejected” and “Unused” rates of 12.5% went down to 11.66% (after we used Originality.ai checks), an improvement of about 7%.

• This shows only the percentage of articles with a “Rejected” status.
• Pre-AI check and Post-AI check refers to the same time periods as the first bar chart.
• Pre-AI check, we had a rejection rate of 3.24%. After using Originality.ai checks we had a 1.54% rejection rate. This is an improvement of about 48%.

Study Method and Data

1. We took a comparable period of three months (a total of 151 articles) before we introduced a strict no-AI policy at our agency and a three-month period after we implemented the policy (a total of 240 articles).
2. We tracked the rejection rate — the editors that confirmed via email that they were rejecting the articles (this could be that they deemed it to be AI-written or poor quality).
3. We tracked the percentage of articles that were written, submitted to editors, and went on to be published. This also gives us the “Unused” percentage, which is the percentage of guest articles that were not published.

‍Caveat: It’s important to note that editors may reject an article for other reasons (not necessarily because they think it's AI-generated). It’s possible they found it didn’t match their publication’s style or audience. Editors who stopped responding may have been very hard to get hold of. However, when an editor didn’t respond after repeated follow-ups, we can often assume that they probably just did not like the article enough to dedicate further time to it.

‍Overall, we did see a reduction in both the rejected article percentage as well as the combined rejected and unused percentage.

Conclusion

The success story here with using Originality.ai is only at an early stage. However, we have already seen an improvement. 

Most importantly, we know we can preserve the integrity of our process and implement our company policy against the use of AI for guest post articles. In the long run, this will improve the results we can secure for clients.