Ever since OpenAI’s ChatGPT was unleashed on the world in late November 2022, there’s been no end of discussion about the tool. Within little over three months, ChatGPT became the fastest growing tool ever with over 100 million users.
In the short time since its launch, there’s been a race to use ChatGPT for all manner of tasks, from people seeking to be ChatGPT gurus to writing job applications, CVs, test answers and even PhD theses. In fact Bill Gates challenged OpenAI in 2022 to create their generative pre-trained transformer to pass the Advanced Placement Biology exam, not an easy test. ChatGPT answered 59 out of 60 questions, and the results were independently marked – it scored an A or A+ and it’s follow-up answers to some of the open-ended questions were remarked as being outstanding!
With an increased investment in the tool from Microsoft, ChatGPT was then quickly integrated into Bing, providing a different type of search.
And What about Google’s Bard?
Commentators rightly pointed out that OpenAI had made a massive advanced on Google’s efforts, despite the Mountain View company’s many years of AI integration into how it handles search – remember BERT, then MUM? Behind this all is the massive strides in Natural Language Processing (NLP) and the breakthrough technology LaMDA in 2021.
ChatGPT even triggered a “code red” at Google with founders Sergei Bryn and Larry Page being called up to provide input into Google’s future chat strategy.
Late to the party, Bard was announced to journalists in March with beta list users being granted access just a week later.
So How Do ChatGPT and Bard Affect Digital Marketing?
We could discuss all these amazing stories at great length, but we’re here, as always, to provide digital marketing services for local businesses.
First things first, we had a huge backlog of digital marketing blog posts we wanted to publish for our business audience. Our first test-drive of ChatGPT was to see how it generated the two dozen stories we had lined up.
Using GPT for Content Generation
We had a wide variety of titles we wanted to write about for our blog audience and our first foray into using ChatGPT was generating these missing posts.
- ChatGPT was quick. It can take anything from an hour upwards to research and write a quality blog post, sometimes half a day to a whole day, depending on the complexity of the research and output. So, for the OpenAI tool to complete something similar in quite literally under a minute was quite astonishing.
- The output was acceptable. It was written relatively clearly and concisely and was certainly logical.
- With over twenty topics tested, we found ChatGPT’s output to be very formulaic. The resulting copy consisted of a couple of explanatory paragraphs about the topic, followed by a list of pertinent points, then a summary paragraph.
- The standard response seemed to be approximately 300 words, right on the bottom limit of what we call “thin content”.
- Lacking personality. All the writers for Clever Marketing over the years have had their own style, from the very simple to the conversational. ChatGPT was quite neutral but lacked the personal touch that our own team have brought to the table. In fact it was a little bland.
- No experience. How can a pretrained Large Language Model (LLM) inject its own experience of digital marketing or web design into the story? It has no experience and so could not embellish the text with relevant personal experiences of the industry.
- No references. Every time we discuss a topic our minds reach back into over twenty years of digital marketing experience and we seek out and link to the stories we remember to support a story. ChatGPT never once made a reference to any source of its information.
- Not up to date. The transformer’s dataset was only up to the end of 2021, so by the time it was publicly released there was zero information from the past twelve months to refer to.
- Not all there. Sometimes the topics we queried were unknown to the tool. Asking about specific but relevant entities resulted in an inability to answer.
In its defence, ChatGPT-3 is just a public beta and the subscription and API access provide far more configurations to control the output. ChatGPT-4 is out now and is superior to its predecessor.
But using ChatGPT to wite for us was enlightening. We test drove Google’s Bard this last week and had similar results.
Using ChatGPT and Bard for Keyword Research
Now that we’ve dealt in a little detail of the biggest subject, using ChatGPT and Bard for content generation, we explored more niche topics such as how could the tools be used for keyword research for instance.
Currently we use a multitude of tools for keyword analysis and research, everything from Google Search Console and Bing Webmaster Tools data, to Google Ads Keyword Planner tool. We combine our data with years of experience in SEO, PPC, and content writing to draw up strategic and tactical keyword plans.
With ChatGPT and Bard we’ve been able to prompt the tools to find:
- Top 100 keywords that businesses in [location] would use to look for [services]
- Top 100 keywords (as above) and their search volumes.
- List the boroughs of London.
- Finding longtail keywords
- Grouping or categorising keywords based on search intent.
- Similar keywords, synonyms, and long-tail variations.
- Keyword counts.
Rather than logging in to multiple tools, we’ve been able to use ChatGPT to perform a wide variety of keyword research tasks.
Bard and ChatGPT for Writing Titles and Meta Descriptions
Writing successful title tags and meta descriptions is a real artform in the world of SEO. It involves conducting keyword analysis, competitor research, and squeezing your succinct copy into a 70 or 160 character limit. Then of course you need to test whether your latest changes are impacting rank and Click Through Rate (CTR).
Both Bard and ChatGPT can assist in this area, if you’re not feeling as creative as usual. They can give you ideas that maybe you’d not thought of, or expand your vocabulary 2.
As with content development and keyword research, a higly-experienced and consummate digital marketing professional should be able to do all these tasks with ease. However, having another tool to assist is certainly helpful if you’re pressed for time or having creative block.
AI for PPC Ad Copy
One great use of Bard and ChatGPT for digital marketing is generating alternative copy for ads.
Say you’re running a Google Ads or Microsoft Ads campaign and you’ve got a perfect, keyword optimised, punchy and compelling piece of ad copy – Asking AI to come up with some alternative copy can be really useful. It can help you overcome any blocks or just highlight clever options that you may not have thought of.
The Verdict So Far
Both ChatGPT and Bard are described as Artificial Intelligence. However, they aren’t that intelligent because they’ve not been able to fulfil all of our prompts let alone to the high standards we expect. They are both artificial though, and they are actually Large Language Models (LLM) plus they are generative pre-trained transformers (GPT) so they are limited to the level of data within them. Bard seems more up-to-date. But ChatGPT has been impressive. And to get the most out of your tools, each needs to be further trained in tone of voice, your prompts need to be accurate etc.
For simplifying and augmenting our keyword research, the tools have both been quite useful.
Title and description assistance has been helpful, as has for writing multiple paid ad versions.
On the content generation side we are still very wary.
Content creation is the biggest grey area of digital marketing, with Google for instance giving mixed messages on whether AI-generated content is legitimate to use in websites. There’s still a moral and ethical dilemma and it does hark back to the time of the “content farms” over a decade ago where the world wide web was flooded with poor-quality content (And Google introduced algorithm changes to combat this).
We’re seeing the same thing happen again with some website owners generating tons and tons of fake content, and even competitors filling their sites up with pages that are very obviously created by ChatGPT (Remember where we said we thought it was formulaic? Exactly, we’re seeing just that!)
In the hands of those who are unable to do certain tasks, these tools give them a “leg up”. In the hands of seasoned professionals, these tools have even more potential. But the most important aspect is that Google’s Bard and OpenAI’s ChatGPT are just tools. They can be used badly or employed really well – and that’s where the difference may lie – the intelligent use of AI.
We originally asked AI to create this blog post but actually ended up completely ignoring the output and writing this ourselves.