Web copywriting and AI-generated content. Is it good for SEO?
Well, what does Google think?
Because, for most of us, only one company's opinion on AI content matters. So:
Does Google penalise AI-produced content like that created by ChatGPT?
Can it detect AI web copy at all? Does it even matter if it can?
The short answer to most of this is it depends.
If you use or are thinking about using a Large Language Model AI like ChatGPT in your content creation process, here's a dive into some things it will be useful to know:
What is AI-generated content?
AI-generated content has been created by Artificial Intelligence (AI) instead of a human.
In this post, we'll mainly be looking at AI-generated written content. Yet there are numerous tools out there, such as Midjourney and DALL-E, that can create images from human-speech prompts.
The current generation of Large Language Model (LLM) AI chatbots work in response to a user prompt. This can be delivered in normal speech.
The AI then searches its understanding of the data it has access to and creates a response. This could be anything from a single word to a full article (admittedly usually quite a bland one). All generated by the AI.
The most popular AI tool as of 2023 is undoubtedly ChatGPT, though there are now many, many others. Google itself is building its own Large Language Model AI tool called Bard.
It's worth noting that, at time of writing, ChatGPT only has access to data up to 2021. Bard will be able to search the whole of the current internet to create responses to prompts.
Google, web copywriting and AI-generated content
Google started off by stating it was not in favour of AI-generated content. Back in 2022, John Mueller (Google's search guy) said that web copy created by AI was considered spam.
This meant that AI content was officially bad for SEO.
In 2023, Google refined its opinions a little. It currently says that its real issue was marketers and businesses that used AI to quickly spam out poor-quality content.
Currently, Google does not penalise you if you use AI-generated content. So long as that content is of high quality, factually correct (we'll get to that), and relevant to your users.
Google’s helpful content update
Google's helpful content update (HCU) launched in September 2023. Like most Google content updates, the goal was to promote high-quality content and push down spammy rubbish.
(The jury is very much out on whether they've succeeded on this, with user-generated content sites like Reddit apparently winning big.)
In this case, the “helpful” in the title refers to the key quality Google thinks your content should possess.
In short, does what you've written help your user?
Well, how can you tell? In Google's eyes, the kind of content they want to reward should:
Be based on your expertise
Make sense for your business or website to write about
Be aimed at a specific audience
Target your user's search intent
Be in line with other Google guidelines
Most people seem to think the helpful content update was a direct reply to the sudden proliferation of AI-generated articles on the web.
Google says not though. While the original accompanying documentation did say good web copy had to be “written by people”, they swiftly edited this.
Who can say why. Possibly the fact that they're building their own AI now. But it could be completely unrelated.
Google's guidelines for AI-created web copy
That said, Google still has some guidelines for how you should treat content that you've used AI to create:
1) You have to say AI was involved in creating it
If you're using AI to generate content, Google wants you to be clear about it and tell people.
You can do this on a separate page on your website (and link to it). Or you can add it to the author section, byline, or add a disclaimer on each page where you've used AI to create some of the content.
2) Say which AI was used
Further, Google wants you to say which AI you used to write your content.
It also wants you to at least indicate where information about how that AI works can be found. They ask you to include:
The name of the AI tool (e.g. ChatGPT)
The data sources used to train it (again, we'll get to why this is important)
Other explanations as to how the AI generated the content
If you do have a special page explaining the AI system you use, you can use something called “markdown mermaid syntax” (a JavaScript tool) to build handy explanatory diagrams for you.
This is all because how the AI went about creating content matters.
3) Check for quality and value
Google's bottom line is essentially that as long as you end up with high-quality content that's valuable for your users, it doesn't matter who (or what) created it.
When it comes to what defines “quality”, Google often points to things like being:
Relevant to the search query
Well written and designed for the “human eye” not just search engines
Not spammy, with smart keyword use
Factually accurate and correct
Possibly including visuals to aid engagement
Can the AI tool you're using do all of that?
Most recent experiments seem to show that even the best AI the tech industry currently offers simply can't compete with human SEO copywriting services. Even human-edited AI content doesn't seem to compare favourably.
At least, not yet.
All of this may change one day. But fast as the technology has been out of the gates, it's not won the race yet.
How AI large language models work and content quality
Large Language Models (LLMs) use statistical modelling to analyse massive sets of data. They then use that analysis to generate text that appears similar to what a human would create.
This kind of AI isn't “intelligent” in the way that a human is intelligent. But they are designed to work somewhat analogously to how the neurons in a human brain do.
When using AI essentially as a copywriter for websites, it is very important to remember that:
1) Training data is vital
LLMs are “trained” on sets of data. These models “tokenise” the words in the data and analyse the relationships between them to predict what the next “token” should be.
The simplest possible way to imagine this is similar to how the predictive text function on your phone works. LLMs work very differently, but the general idea is roughly the same.
(Here's a great visual explainer on AI chatbots from The Guardian.)
With a bit of training, an LLM can start to understand things like context, meaning, and relationships between words. But they only have the data on which they've been trained to go on.
It's also not impossible for them to misinterpret the data and output something that isn't “faithful” to what they've been trained on.
2) Always check for hallucinations
A “hallucination” is the name given to AI output that's factually incorrect or nonsensical but is written in a way that makes it sound like an established fact.
Checking your AI-generated content for hallucinations – i.e. factual accuracy – is absolutely vital.
LLMs are nowhere near the stage yet where you can employ one to write web content and expect it to be perfect without some human attention.
In fact, one LLM's inability to correctly say which telescope took the first picture of an extrasolar planet caused a huge hit to the developing company's stock value.
That fact itself is pretty mundane (unless you're a huge Sci-Fi nerd like me). But it's a good indication of the need to fact-check every statement in AI-produced content.
(The telescope was the scientifically named Very Large Telescope in Chile – here's the NASA website on the fact. I mean, if you can't trust NASA, who can you trust?)
3) Emotion does not compute (yet)
Understanding human psychology makes you a better writer. This is important because most LLMs don't do humour and persuasiveness very well.
There's also a risk that they don't understand the meaning of, for example, hate speech they've been trained on. They just calculate what token should come next.
Writing effective web copy for sales requires addressing things like pain points and understanding your ideal customer.
You might be able to get some okay sales copy from your favourite AI chatbot if you're really good at creating prompts.
But if you want high-converting content, “okay” probably isn't good enough.
Can Google detect AI-generated content?
Google is getting pretty good at detecting AI content.
There are also numerous AI-detection tools out there (my favourite is named ZeroGPT, purely because of the name).
Google's current guidelines say you must disclose that AI was used to create the content.
But they don't actively penalise you for doing so. Again, it's the quality of your content that Google cares about.
Why is my writing being detected as AI?
The increasing use of AI content detection tools has been a problem for some writers.
These tools aren't perfect. They're in their infancy themselves. Some are dialled far too loosely to be of any use. Others give weirdly low scores to all content. Just to be on the safe side.
If your writing is being detected as AI, it may be because you are using overusing:
Cliches
Very common phrases
Repetitive language
This also gives you some idea of the potential pitfalls of using AI to create content.
Without human input, editing, and proofreading, it's not uncommon for it to create content that features all of the above.
Create quality content to appeal to Google
Does Google think you can farm out your web copywriting to AI? Is it good for SEO? In short, Google says "it's up to you" with a "but".
AI chatbots are still pretty brand-spanking new as a technology. The developers haven't quite gotten all the kinks worked out just yet.
You absolutely cannot just use an AI chatbot to produce content and publish it without checking it for factual accuracy. Doing so will almost certainly be bad for your SEO (and maybe your reputation too).
Plus, not all Large Language Models are created equal. ChatGPT can create an article for you. It probably won't be that good for SEO or your human readers. You can't guarantee it will be accurate. But it will be an article.
Yet many spin-offs and copies of ChatGPT exist out there that are barely functional. They're "jump on the bandwagon" efforts that create terrible copy.
If you're trying to trick Google by spamming out large quantities of rubbish content, it's not going to work. But that doesn't mean Google penalises AI content – just poor content in general.
Want to create high-quality content your audience will value? How about copy that sells?
Let's talk it over. The Maiden Standard has been providing SEO copywriting services to businesses in Bristol and further afield for over a decade.
Let me know your requirements or arrange a chat at benjamindmaiden@gmail.com or by filling in the form below.