As technology advances, we find ourselves on the brink of a new era where Generative AI and Large Language Models (LLMs) are transforming the way in which Digital Marketing specialists approach a range of different tasks. From keyword research and content creation to overall SEO & PPC strategies, there are some big claims that tools like ChatGPT are going to take over from Digital Marketers entirely. In this blog, we’ll provide an overview of how Generative AI and LLMs work before looking at how they’re being utilised within the Digital Marketing space. Through this, we aim to provide businesses with the key information they should be aware of before implementing Generative AI within their overarching business strategies.
“Generative AI has the potential to change the world in ways that we can’t even imagine. It has the power to create new ideas, products, and services that will make our lives easier, more productive, and more creative.” – Bill Gates
How Does it Work?
LLMs like ChatGPT and Bard are used to generate predictive human-like text and content derived from enormous quantities of data that they have been fed during development. Taking ChatGPT 3.5 as an example, its training data encompasses various sources scraped from across the internet dating up until September 2021. This database includes (but is not limited to) Wikipedia entries, complete works of literature, scientific journals, and threads from social media platforms like Reddit. The data is then filtered to prioritise data points from the more ‘trustworthy’ domains included within the dataset. It is by using this data that ChatGPT 3.5 can write predictive text that gives the illusion of sentient thought.
But that all sounds a bit confusing so let’s look at an example. If prompted with the beginning of Dorothy’s famous line ‘There’s no place like home’ from The Wizard of Oz, ChatGPT assigns a probability to each token (in this case a word) that could be used to complete the phrase.
There’s no place like
· home (44%)
· the (26%)
· my (16%)
· nature (4%)
· London (1.2%)
· excellent (0.005%)
*All of these values are for illustrative purposes only
Although LLMs don’t necessarily always pick the choice with the highest probability, this approach allows for Generative AI to string together words and sentences deemed most likely to make sense based on the data it has been trained on. This process multiplied thousands of times allows these programs to meet a range of user demands from providing cake recipes to generating insults in the style of Shakespearean verse. As such, it is only natural that digital marketers have tried applying these tools their everyday tasks as well.
“Aristotle founded or discovered logic by observing the world. ChatGPT thinks logically. Why? Because it notices all the logic in the data in its training set” – Stephen Wolfram
How can this help Digital Marketers?
The most noticeable benefit of LLMs lies in the speed with which they can complete tasks. If prompted to ‘write a 2000-word article on the implications of using Generative AI for SEO’, ChatGPT 3.5 takes a grand total of 51 seconds to generate a response. Of course, the temptation I had was to simply hit copy and paste before taking the afternoon off, although as we’ll soon illustrate, there are myriad reasons why LLMs require human oversight and editing before being anywhere near publishable quality.
The capabilities of Generative AIs even extend to On-Page Optimisation tasks such as keyword insertion, wherein AI can review existing copy and adapt it based on established parameters (eg. Review this article and insert the keyword ‘Affordable Trainers’ five times). Overall, these AI tools are incredibly effective when it comes to the bulk editing of large data sets and can be utilised to complete tasks in seconds that would take hours when done manually. For more information on how Generative AI & LLMs are being used in the SEO (Search Engine Optimisation) space specifically, check out our blog detailing some of the Top Uses of AI Generated Content for SEO.
In addition, LLMs are excellent at summarising topics and answering queries providing they aren’t too technical. This makes ChatGPT an invaluable resource for beginners wishing to have the basics of a topic explained in a manner that’s concise and easy to understand. But again, there are limitations here too, given that Generative AI often fails to hold up under scrutiny. Although we’re getting ahead of ourselves a little bit.
Limitations of Generative AI
While the capabilities of Generative AI and LLMs are staggering, they’re also not without their limitations. One of the most glaring drawbacks of these tools lies in their ability, much like the best politicians, to present incorrect or misleading information with a high degree of confidence. In the technical space, this lapse in judgment is referred to as a ‘hallucination’. While the context varies, the results are often that terms, locations and facts are entirely fabricated or misinterpreted. As such it’s of the utmost importance that the results of LLMs are being reviewed by subject matter experts to prevent the presentation of inaccurate content that could compromise your website’s Search Engine Results Page (SERP) ranking. Even worse, large amounts of misleading, inaccurate information on your website will undermine users’ faith in your brand and reduce the chances of potential customers converting into sales or leads.
In addition, the training sets fuelling LLMs can be saturated with outdated, inaccurate or even discriminatory data. This can heavily influence the reliability of Generative AI and is another reason why human oversight is essential to making the most out of these programs. We cover this more extensively in our accompanying blog post looking at ‘5 Limitations of Generative AI & LLMs’ everyone should be aware of. Check it out to avoid some of the common pitfalls that come with implementing this technology in everyday business operations.
But is it Useful?
Ultimately, it’s important to understand that Generative AI is not some magic ticket to the top of SERP rankings. While these programmes are incredibly sophisticated, they are still tools that are only as valuable as the person wielding them.
“people want to use generative AI for everything… It’s like using a microscope to hammer in a nail – it might do the job but that’s not really what this tool is meant for.” – Dr. Sasha Luccioni (Machine Learning Expert)
Being able to effectively utilise LLMs is becoming a skill unto itself that Digital Marketers are learning to master. Where, if you want to get the most of programs like ChatGPT, you have to understand how to effectively write prompts that clearly outline desired parameters such as length, style and context to name a few. There have even emerged plugins such as AIPRM that allow users to access a range of targeted prompts for ChatGPT as a result.
Google has announced that AI-generated content is not going to be penalised in SERP results, although that doesn’t mean that it’s going to be rewarded either. As Search expert Danny Sullivan clarified, Google ‘are looking at the helpfulness of content, rather than how it is produced’. So, if you’re writing vague prompts to ChatGPT that you directly copy and paste to your website, don’t be surprised if that isn’t up to par with what Google deems to be ‘Helpful Content’. But with an appreciation for the limitations of this technology, Generative AI is more than capable of serving as an invaluable tool in the arsenal of SEO specialists helping inspire new ideas alongside more streamlined approaches to content creation and optimisation.
Get in Touch
Web Results Direct prides itself on staying up to date with the latest industry developments in the world of Digital Marketing. So, if you or your business have any queries or concerns about the implications of Generative AI and LLMs, feel free to send us an email or give us a call. We’d be happy to provide you with the most up-to-date information and techniques to help your business thrive in the ever-evolving digital landscape.