AI content is growing: You are probably reading ChatGPT3+ content several times daily

ChatGPT and AI are everywhere all at once!

No, you are not imagining it; if you have read the news or been on social media in the last year, you have seen it. ChatGPT can write code, poems, and essays – all with a single prompt from a human.

“AI and, in particular, natural language models that generate content (generative AI) have already permanently altered content marketing. AI models can’t do everything; they can only write about what has already been written.

The models also rely on older-style” spinning” tactics that have existed on the web for years; this spinning is where machine learning looks for words to substitute with similar words to drive up the overall uniqueness of the content.”AI and, in particular, natural language models that generate content (generative AI) have already permanently altered content marketing.

AI models can’t do everything; in particular, they can only write about what has already been written. The models also rely on older-style “spinning” tactics that have existed on the web for years; this spinning is where machine learning looks for words to substitute with analogous words to drive up the overall uniqueness of the content.

However, these models dramatically drop the cost of written content production on the web and have changed how the web is publishing dramatically over the last year. There were several reports over the previous year of websites like CNET and Bankrate using AI to produce SEO content at scale. We only know of these content cases because they were disclosed. Both websites paused their production of AI content after it garnered public attention.

Large publishers and writers are flocking to use AI tools because the time savings are actual. The simplest way to explain it is as if you have an unlimited staff of high school-aged to college freshmen writers. You can outline a piece of content, prompt your AI, and edit the work to match your standards at about 4X the speed as if you had to do it all from scratch.

Pre-ChatGPT3+ early movers

GPT-3, the receptive language model that powers ChatGPT, became available as a beta in 2020. By 2021 natural language models started to hit the mainstream for consumers in beta forms, the largest being Jasper.ai, formerly Jarvis. Jasper.ai says their model is trained on 10% of the web and customer feedback.

Jasper.ai has several work modes, but like other AI models, it takes a user prompt and spits out a response. In the tool itself, users can rate if an answer is good or not – with a small refund issued if the response is rated as not good. This feedback loop is likely a critical piece of Jasper’s model and is one reason being an early mover in this field is a huge advantage.

Jasper.ai has positioned itself towards marketers and content creators as a time saver. At the end of 2022, they raised 125M at a 1.7B valuation (Techcrunch) and have generally benefited from the hype around AI. At this time, ChatGPT introduced its extremely popular beta; according to ARS Technica, it had the fastest-growing user base ever.

Post ChatGPT3+

ChatGPT3+ hit the mainstream beyond digital natives and specialists in the field. The success came as a little bit of a surprise to me because while ChatGPT3+ is a leap forward – we had seen similar tools released in the prior year broadly. But, whether through a combination of their marketing or tooling – ChatGPT exploded, and so did the concern over what exactly the impacts of AI would be.

The explosion in interest from ChatGPT3+’s successful launch and potential use cases has spurred even more interest in AI than there already was. Overnight seemingly, companies like Microsoft, Quora, Amazon, Hubspot, and Google were all promising to release their versions of the technology. Of course, one reason they could so quickly respond is the technology is both open source and has been around in these private companies for well over a decade – at least, this is likely for Google and Microsoft.  

AI’s Impact on web content

AI tools being open to everyone is increasing spam and misinformation noticeably. I believe one of the reasons Microsoft and Google were not first movers – even though they understand this advantage better than everyone, is that natural language AI is a significant legal grey area.

There are 2 principal issues I see that slow them down:

1. AI can spit out straight misinformation – it’s trained on the web; think Reddit.

2. Many consider it tantamount to plagiarism, and it is unclear what standard the courts in the US will apply to this technology.

In the short term, it’s unlikely burdensome regulation will be enacted in the US because of the issue’s complexity. However, some countries, like Italy, moved quickly and outright banned ChatGPT.

Misinformation

Natural language AI, like anything, has good and bad uses. In the case of natural language AI, the inadequate services are pretty astounding. Natural language is already writing most of the spam emails you are getting or at least helping. No one wants to write a cold email – writers are making machines do it, and spammers are using it to scale – whether email or PPC ads.

I wrote a deep dive piece on misinformation related to Amazon A9 search that spread through top influencers on LinkedIn unchecked for quite a while.

You are reading AI content if you read the web.

People are using the tools to write social media posts, and whether you know it or not, you are reading the outputs of ChatGPT several times a day. Large companies, including CNET, Bankrate and openly adopted it to generate entire pages from scratch. Writers use it in their freelance work, and websites are both knowingly and unknowingly publishing this content daily.

As of 2022, Jasper AI, just one of several hundred companies, had over 100,000 paying users using their tools. So if you regularly read web content, you read generative AI content.