The big topic for SEO right now is Chat GPT and how to use AI to generate content. We’ve been saying “content is king” for several years, but now we have to include a caveat: the right type of content is king.
There are hundreds of uses for Chat GPT and OpenAI – title tags, on-page content, meta descriptions, marketing emails, chat bots, etc. Basically, any text can be automated to a certain extent.
Google’s Official Guidelines
Officially this is a big no from Google. It’s considered spam in general, but the latest evolutions of AI may be progressing faster than Google can keep up with in terms of flagging within its algorithm.
This was much easier to check over in the past when autogenerated content could be spotted a mile off. Thin content, bad spelling and grammar, and general non-sensical sentences were all the rage for link building 10+ years ago. Still, if Google knows you are using autogenerated content, it can be marked as spam.
What this means is up for interpretation though. If you have a few articles which are AI generated, then you’ll probably be alright and might just have a few pages ranking lower. If you use it across the majority of your site, then you could run the risk of penalty actions or being de-indexed all together.
Of course, this is assuming they can tell what is AI generated or not, and Google use chat bots and other forms of AI themselves, so it’s also quite hypocritical of them!
Update: Google announced further info on the 8th February. This says that “Google prioritizes high quality content, regardless of whether humans of machines generate it.” and “Using automation, including AI, to generate content with the primary purpose of manipulating ranking in search results violates Google’s spam policies.”
How To Check For AI Content”
The data and content from Open AI and Chat GPT is all saved, but with it’s huge surge in popularity, this may change.
So theoretically the content can be matched up, but this is assuming the content is shared with search engines – which seems unlikely.
Open AI launched their AI Text Classifier to help check the probability of AI generated content, but this is by no means fool-proof.
We ran some quick tests on this and found that some of our content is unlikely AI-generated, but some blogs we know are human written are flagged as “possibly” AI-generated. So we either have secret robots on our team, or the tool isn’t good enough yet. It would be incredibly harsh of Google to penalize a site which gets flagged falsely.
Presumably Google are actively working on methods to locate AI-generated content, but nothing official has been said in relation to this latest increase in use. So for now, we are OK to use it, but it might be worthwhile keeping a record of what it’s used for, particularly if it’s your content on high ranking pages.
Just FYI, I’ve put in the content above into the tool and it’s “very unlikely”, so I’ve passed the Voight-Kampff test for now!
How To Use AI Content
As mentioned, we have used these AI tools, much like every other digital marketing agency probably has!
The prompts given to it are the key to getting good results, so experimentation is needed. Stipulating styles and lengths are good for meta titles and descriptions.
It can be done by using the URL as ‘seed’ information, but it works much better by crawling and extracting existing content for optimisations.
Of course, you might not have this and you will be using AI to create the content on the page in the first place. If this is the case, make sure you stipulate offers, delivery limitations and other specifics so that it doesn’t just return erroneous delivery and offer information.
You will invariably need to re-write, edit and double check everything. It will be a good time saver, but unless you are very, very specific on your prompts, you probably won’t be able to cut and paste directly.