The truth about AI in SEO
Wednesday, January 17, 2024
The rise of AI
In 2023 AI hit the digital industry in a big way. November 2022 saw the launch of ChatGPT and it quickly picked up users. According to Exploding Topics, it took just 5 days for the platform to reach the 1 million users mark - a feat that surpassed platforms like Facebook, Instagram and Spotify.
The launch of this platform had a huge impact on the SEO industry. Suddenly the barriers to entry for AI had been obliterated and the floodgates opened. People flocked to the tool in a bid to create content at a ridiculous speed. We also saw the rise of ‘Prompt Engineers’ on platforms like LinkedIn as people learned how to craft prompts to get the maximum output.
It’s no surprise, then, that Google released an article that provided guidance on the usage of AI-generated content in February 2023. Contrary to what we were all expecting, Google actually explains that their problem isn’t with AI-generated content - it’s with content that has the “primary purpose of manipulating ranking in search results”. This has always been the case in search - Google’s mission has always been to promote the most relevant, high quality content for users.
So why, then, have we seen a rise in AI-generated content being demoted by Google in their search results?
The simple answer is that the content that they’ve demoted isn’t satisfying user intent and is low in quality.
An example of this is the ‘SEO Heist’ that we saw hit the headlines back in December 2023. A marketer used AI to essentially copy a competitor’s entire sitemap and use that to create new content and outrank them.
This is the site that the content was ‘stolen’ from:
And here’s a screenshot from the site that leveraged AI to ‘steal’ this traffic:
As you can see from the above, this worked for a while. The sceptic in me thinks that the success of the offending website could have continued for longer, but they made a mistake - the architect of this strategy decided to share their success and process on X. The post quickly picked up a lot of views (7.6M as of writing this) and I suspect that it also caught the attention of the Google Search Quality team, hence the steep decline in organic traffic.
The nuances of AI content
So why am I telling you this?
The first reason is that I wanted to dispel some of the misinformation and stonewalling around the use of AI for content creation purposes.
Ultimately, Google doesn’t really care where your content comes from (and arguably they’re still not perfect when it comes to differentiating AI from human content) - they just want quality content.
Serving high quality, relevant content to users makes Google the primary choice for information retrieval. This then makes it more likely that users will interact with ads and therefore generate revenue. Basically, their business model will always rely on the Search Quality and Google Ads teams ensuring that users get the best experience possible.
But here’s the nuance:
If you’ve only read up to here, you might think that I’m advocating for the use of AI content.
And I am.
I see a lot of people saying that they absolutely won’t use AI generated content because it’s against Google’s guidelines, or that it’s poor quality.
But the reality is that it’s not against Google’s guidelines, and whilst the quality of the content can vary based on the information requested and the prompt used, it does do a lot of the leg work in terms of pulling vast amounts of information into one place.
However for all of this ability, there are a lot of things that AI-generated content can’t accomplish:
It can’t fact check content (this is a huge issue for anyone who has to worry about compliance)
It can’t accurately replicate a brand voice
It doesn’t have depth of brand and product knowledge
It can’t easily convey ‘Experience’ (as well as the wider scope of EEAT)
Something that we should also all bear in mind is the true nature of AI-generated content. When you use something like ChatGPT or similar, it will use a GPT to create the output.
The full name for GPT is Generative Pre-Trained Transformer, which is a type of large language model. I’m not going to LLMs in this article but if you’re interested in getting under the hood of language and text generation then Google actually offers a course that covers this.
What I am going to provide is this - the ‘pre-trained’ part of GPT means that it is trained based on an existing dataset. On existing information.
That means that it doesn’t matter how much the content is spun by the LLM, it’s ultimately nothing new. It’s content that has been created based on an existing corpus which is readily available. Part of that corpus will be high quality, unique, relevant content, but there might also be a part that contains misinformation and low quality content. This will all be used and amalgamated to generate the response for your output. There are two major issues with this:
The obvious - you could quite quickly end up with incorrect, low quality content which wouldn’t do you any good in search - it could even harm your site if you’re just blatantly regurgitating content. It’d probably also give your users a bad experience.
The not so obvious - if you’re covering the same thing as every other site, you’re not adding any value (or information gain) over any other site and you may struggle to rank your content.
In either of the above scenarios, you could create content faster but it’d still be a waste of time as this wouldn’t perform.
How to use AI intelligently in your content strategy
If you’re set on using AI to help with content creation then there are a few tips that I’d like to leave you with to help you strike that balance between creating quickly and maintaining a level of quality to meet user needs:
Fact check everything. Misinformation could easily find its way into the dataset for an LLM and there may be elements of hallucination that could also produce false information. It’s crucial that you catch this.
Make it your own - there’s nothing worse than reading a piece of content that has zero personality. A lot of work goes into constructing a brand and brand voice - don’t let AI take that away.
Focus on information gain - what could be added to the content to add value that doesn’t exist on sites that are frequently surfaced for the same topic in search?
Ensure that you’re displaying EEAT in your content (and on your wider site). Edit everything to include details of real world experience where you can. Audit your site as a whole to make sure that you’re sending the right signals when it comes to EEAT
Ultimately if you’re utilising AI-generated content, you need to be prepared to still do a lot of legwork in terms of editing and building upon the base of content you receive.