While Google’s Fred algorithm update helped push higher quality content up in the SERP rankings in 2017, it might get a little tougher for SEO copywriters to be able to distinguish their content in the coming future. There’s a new competitor that is going to have the ability to create volumes of content while you’re still deciding on which keywords to tackle first.
While it may be replaced by updated versions, this competitor is currently called GPT3. Ever since OpenAI announced the release of GPT3 in 2020, AI generated content has become an increasingly common sighting.
What is GPT3?
GPT3 stands for Generative Pre-trained Transformer 3. Simply put, it is an autoregressive language model that has the ability to produce human-like text. This means that you can give it a 2 sentence prompt and have it return an entire article based on the subject of that prompt.
Andrew Tate’s blog post on animalz.co mentions that GPT3 has read about 45 terabytes of text which includes every bit of content that existed on the web between 2016 and 2019. It has analyzed every single line of that text, which enables it to come up with the most fitting word that should come next for any prompt. You simply need to specify how many words you require and GPT3 will keep adding the ‘most fitting’ words based on its algorithm.
Tate’s article gives some examples of text generated by GPT3 and I doubt that I would be able to distinguish it from human written content.
Issues with GPT3
If you’re already prepared to use GPT3 in order to write up something important like a college thesis, you might want to take another look at how it generates content and what you can expect from it.
Tate mentions that, unlike humans, GPT3 lacks cognitive flexibility and the ability to see reason. Its goal is simply to find the most fitting word that comes next in any example of written text. This tells us that the content created by GPT3 has a tendency to beat around the bush and say things without a purpose instead of leading to a thought out conclusion.
Three of the biggest flaws in GPT3 generated content (when it comes to publishing a longer article) were:
- It lacked a narrative.
- It made claims on inconclusive evidence.
- It lacked the ability to convey new and meaningful information.
After testing it out on various types on content, Tate found that GPT3 most often gave the best results when producing short and creative content (as opposed to long and factual content).
Why GPT3 is threatening to content creators
Ryan Law says that there is an upcoming calamity of a vast amount of ‘copycat content’ that will swarm the internet and calls it The Search Singularity.
There is no doubt that when a lot of companies get their hands on a tool like GPT3, they will begin publishing AI generated content at an exponential rate. It makes sense because, while a lot of top ranking sites are SEO friendly, there is a huge amount of low quality human generated content that ranks as well.
Content creators simply can’t do the bare minimum anymore in order to compete. Since quantity is not an option, the only way to beat AI is by making higher quality content. Of course, as Law mentions, the best strategy going forward is probably going to be a combination of AI generated and human edited (or vice versa) content.
How will Google’s Algorithm react to GPT3 generated content?
We can only speculate as to what kind of updates Google will come out with once AI generated content floods the internet. While Google’s E-A-T (Expertise, Authoritativeness, and Trustworthiness) is not a ranking factor (according to the Search Engine Journal), it is easily possible that Google might consider AI generated content to be spreading misinformation and implement E-A-T inspired updates across the entire search engine. This would mean that the ‘Medic Update’ from 2018 could be applied to a broader range of subjects.
Another possibility, suggested by Takeshi Young, Global SEO Lead at DiDi, could be that Google’s MUM will be used to generate it’s own version of AI content. By doing this, Google can get an idea of what keyword targeted AI generated content looks like and use that data as a comparison tool while crawling a website’s data. If that happens, sites found using purely AI generated content will undoubtedly be penalized.
In case you haven’t come across it yet, MUM (Multitask Unified Model) is an improvement of Google’s search engine that was introduced in May 2021. It is a language model, similar to GPT3, and is capable of generating content. Its purpose, however, is to transform queries into conversational inquiries, eliminating the need to optimize for keywords (but let’s not get into that right now).
SEOs often have to speculate on how the industry will change and implement new tactics as soon as it does. I think it’s safe to say that GPT3 will hugely impact SERPs and the way we compete to rank higher. It’s never too early to get started on a change that can only have a positive impact. These are my final suggestions in order to prepare for the search singularity:
- Work on improving your website’s E-A-T.
- Get access to GPT3 or an alternative and test it out for various use-cases.
- Use GPT3 in order to get topic ideas.
- Create outlines for your content and use AI to fill in the blanks.
This technology is still relatively fresh and, if you’re a content writer, you might want to learn how it might benefit you sooner rather than later.