SEO Fight Club – EPISODE 162 – Google vs. AI Content
SEO Fight Club / By
Unlock Your Local SEO Power With “Confessions Of An SEO!”
– I am participating in an episode of SEO Fight Club, which is being streamed live on YouTube.
I am discussing “Google versus AI content” with Lee who prepared the topic for us.
– Terry Power and I are doing a Ranking Boot Camp at the end of this month, where attendees will gain new weapons to use in Local SEO and learn how to implement them for results.
– I also have my Tested SEO Training program starting August 3rd that covers techniques never before published and teaches attendees how Google’s algorithm works.
– Lastly we discussed Confessions Of An SEO podcast where people can ask questions about any frustrations they face between their stakeholders and an SEO as well as search it up on Alexa, Amazon Alexa or Google by its name.
==========
“Lee’s SEO Training: Quick, Repeatable Results In 5 Weeks”
– I followed Lee’s training which is a 8 week program consisting of progressive steps that teaches SEO
– Through the course, I took a website from not being indexed to number 1 on Google and Bing in 5 weeks with moderate difficulty keyword
– It was amazing to watch all the pieces fall into motion
– The technique works across local and national keywords on multiple continents
– People have come back to do it up to 3 times as they continue to get more out of it each time
– Carolyn picked a scary keyword with high search volume for her project, but despite this she achieved quick and repeatable results through her implementation of the techniques taught by Lee’s training
– It’s been eye opening for me to see how effective this method has been for so many people in such short amount of time
==========
“AI Content: Risky Business?”
– I recently saw a lot of discussion in SEO circles about AI content and how it might violate Google’s Webmaster Guidelines.
– The talk was sparked by a comment from John Mueller who said that using machine learning tools to generate content is essentially the same as “shuffling words around looking at synonyms or doing translation tricks.”
– People are worried, but Kyle Roof’s rhinoplasty Plano sites still rank for multiple cities, which proves that Google hasn’t solved this problem yet.
– Search Engine Journal doesn’t believe that Google can automatically detect content generated by language models like GPT3 and issued an article encouraging people to use them for their content.
– However, if Google finds out they might consider it to be web spam
– so there is potential risk involved.
==========
“T5: Revolutionizing Content Rewriting”
– Google recently released T5, a content rewriter that is expected to revolutionize the field.
– The source code for T5 is available on GitHub and there are articles about it online.
– Content creators may be concerned as Google’s new content rewriting could lead to their work being copied without attribution.
– Old school spinners replace synonyms manually or semi-automatically while paraphrasers do this more automatically.
– Bill Bot and GPT2/GPT3 are examples of AI type content generators which scan web pages and predict one word at a time based on what is already written.
– T5 will infer missing text, making it the newest addition to existing tools used in content rewriting.
==========
“AI Footprints: Uncovering The Trail Of Artificial Intelligence”
-AI content generators are based on probability decision trees, where each word triggers a new probability distribution and the AI picks an outcome.
-Using this system of probabilities, it is possible to create many different texts from one set of input.
-Google can easily detect AI written text due to its unique structure and common models used by AIs.
-I recently tested two AI detectors to see how well they worked for detecting GPT3 content
– one was trained on gpt2 and the other just for GPT 3 specifically.
-The results show that both were able to accurately detect GPT3 content based off the unique footprint left behind from AI generated text.
==========
“Swapping Names And AI Detection: Outsmarting Artificial Intelligence”
– I recently had an article written on Los Angeles DUI lawyer Ted, and prior to editing it included the name of another attorney so I swapped that out
– AI detection programs are used to determine if a text was written by a machine or human.
– The GPT2 detector said this article was 99.67 percent fake, while the GPT3 detector showed much different results
– When looking at word highlights in green were 10 most likely predictions; yellow indicates top 100 words, red indicates top 1,000 and purple going beyond that
– Artificial Intelligence can detect patterns such as Lorem Ipsum, but when those words are put into a different order
– even though they’re still real words
– it might not get detected as AI.
