Should you use AI chatbots like ChatGPT for copywriting?
Throughout 2023, there has been much discussion in marketing circles of the potential uses of artificially intelligent chatbots to assist in copywriting. ChatGPT has rarely been out of the news, and one of the areas in which it has most scope to disrupt existing workflows is in content research and creation.
Arguments in favour of using AI chatbots to help with copywriting include:
- It offers significant time savings in the complete execution of copywriting projects from conception to completion
- For web content, it can suggest forms of wording for headings and subheadings that are relevant to Internet searches and help the page to attract visitors from search engines
- For any application, web or print, it can generate ideas for the content and structure of the treatment of a topic
- It helps establish relevant facts from the material it has access to, saving on initial research time and avoiding the risk of human error from inaccurate factual recall
- It gives ideas for basic intellectual arguments and analysis around any topic that requires this
- AI content does not come with any authorial ego or personal investment of time and energy attached, so if you don’t like every detail that is suggested, you can edit it or challenge it without offending your copywriter
- Although content produced by chatbots tends to have a distinctive signature or style that makes it obvious to search engines and human readers that it is not written by a person, you can consider it as a draft outline that facilitates your presentation of the topic in your own words, and it can still save you significant time even though you may need to spend an hour or so reworking it into an article in your own distinctive voice and ensuring it is on brand for you or your client(s).
Arguments against using AI chatbots in copywriting include:
- There is a risk of intellectual plagiarism by chatbots from unique sources on a topic, without your knowing their sources or how closely it copies any single source, as they generally never acknowledge them; and although chatbots will tend to express whatever information they find in the sources to which they have access in their own words, it is unethical to copy others’ ideas without extending the courtesy of a source reference, even when you put them in your own words
- Chatbots sometimes claim facts that are not true, based on their AI routines making false inferences and connections from known facts or from presumed facts implicit in the prompts that are put to them by users; this phenomenon is referred to as ‘AI hallucinations’
- Our tests indicate that chatbots like Chat GPT are generally adept at giving a fairly succinct, even tersely-worded brief encyclopaedia-entry-style summary of a topic, but much poorer at more complex intellectual analysis, so they should not be relied on to provide the arguments that should be made in a well-reasoned presentation of any intellectual question
- Because they are based on a formulaic reading and processing of existing sources by a computer program, the output of chatbots tends to lack any semblance of originality
- The output generally lacks the kind of personal voice that keeps people reading your content because they are interested in your views or like your writing style; it can therefore be bland and uninspiring to read (lacking typically human qualities that make a writer's style recognisably individual), resulting in unengaging journalistic and blog content that is unlikely to provoke a reaction
Notwithstanding the evident drawbacks of AI-generated copy, the time-saving argument is likely to be a powerful one for many creative agencies and even for professional copywriters who need to cut down on their research time for projects where they are paid by the word and not on the basis of how long the job would take to do to a high standard.
In full disclosure, all the above arguments and content were entirely the product of our own minds, and were not facilitated by any AI resource. However, in order to demonstrate its potential usefulness, we have given three examples below of how prompts in Chat GPT 4.0 can help inform articles on particular topics.
1. AI Copywriting for SEO article
Give an account of the evolution of best practices in search engine optimisation between 1998 and 2022.
The evolution of best practices in Search Engine Optimization (SEO) from 1998 to 2022 reflects the dynamic interplay between search engine algorithms and the tactics used by websites to rank higher in search results. Here's an overview of the key phases in this evolution:
1. Late 1990s: The Dawn of SEO
- Keyword Stuffing and Meta Tags: Initially, search engines relied heavily on keywords within the website's meta tags. This led to the practice of keyword stuffing, where website creators would overload these tags with relevant keywords, regardless of content quality.
2. Early 2000s: Search Engines Get Smarter
- Google's PageRank Algorithm: Google introduced the PageRank algorithm, prioritizing backlinks and viewing them as 'votes' for a page's quality.
- Focus on Quality Content: With search engines getting better at understanding content, there was a gradual shift towards producing high-quality, relevant content.
3. Mid-2000s: The Rise of Holistic SEO
- Penalties for Black Hat Techniques: Search engines began penalizing sites that engaged in deceptive practices like hidden text, cloaking, and excessive link exchanges.
- User Experience Becomes Crucial: The importance of site architecture, navigation, and a positive user experience started to gain traction in SEO practices.
4. 2010s: The Age of Mobile and User Intent
- Mobile Optimization: With the rise of smartphones, mobile-friendly websites became essential.
- Content Relevance and User Intent: Search engines started focusing on user intent and the contextual relevance of content.
- Social Signals: The influence of social media presence and sharing on SEO began to be noticed.
5. Mid-2010s: Technical SEO and Algorithm Updates
- Frequent Algorithm Updates: Google's algorithms, like Panda and Penguin, were regularly updated, penalizing low-quality content and unnatural link profiles.
- Technical SEO: Aspects like site speed, responsive design, and structured data gained importance.
6. Late 2010s to 2022: The Era of AI and Machine Learning
- AI and Machine Learning: Google's RankBrain and other AI systems further refined search results based on user interaction patterns.
- Voice Search Optimization: With the popularity of digital assistants, optimizing for voice search became a focus.
- E-A-T Principle: Google emphasized Expertise, Authoritativeness, and Trustworthiness as key factors in content ranking.
- Featured Snippets and SERP Features: SEO strategies started targeting position zero (featured snippets) and other SERP features like local packs and knowledge graphs.
7. 2022 and Beyond: Looking Ahead
- Core Web Vitals: Google's Core Web Vitals, focusing on user experience metrics like loading performance, interactivity, and visual stability, became ranking factors.
- Semantic Search and NLP: Continued advances in Natural Language Processing and semantic search to better understand and match user queries with content.
- Increased Focus on Privacy and User Data: Shifts in how user data is collected and processed impacted personalized search and SEO strategies.
Throughout these periods, the overarching trend in SEO has been a consistent move away from manipulative tactics towards a focus on high-quality, user-centered content, and a seamless user experience. The evolution of SEO is a testament to the internet's maturation and the increasingly sophisticated nature of search engine algorithms.
We feel that Chat GPT’s response has given a very useful structural outline to help inform an article on the evolution of SEO over a 25-year period. It would not make for a good article as it stands, because it is little more than a bulleted summary of developments in SEO during different periods of the past quarter of a century. Many of the summary points made lack sufficient definition to be informative. For example, 'Shifts in how user data is collected and processed impacted personalized search and SEO strategies' is too vague to tell the reader much at all. Shifts in what direction? Impacts of what kind? These details are needed to make for a useful resource.
Still, each of the points it raises could be considered a useful starting point for a human copywriter’s account of what occurred at these times, in their own words and in fuller sentences, and preferably with the help of independent fact-checking. This is therefore a good example of how a chatbot can save time in gathering ideas for content and structure, although it has not created anything that would make for appealing reading in its raw form, and nor does it give enough detail on each point for it to be fully adequate or educational as a history.
It would also be possible to prompt ChatGPT to approach this task in a different way, so it provides more detailed answers - we explore a strategy for this in the prompt below.
2. AI Copywriting for Conveyancing article
Write a good article for a professional lawyer's website on the differences between England, Wales and Scotland in the procedures and charges associated with house sales and purchases.
Understanding the Varied Landscape of House Sales and Purchases in England, Wales, and Scotland
When navigating the property market in the United Kingdom, it's crucial for prospective buyers and sellers to understand that the legal systems and processes for house sales and purchases differ significantly across England, Wales, and Scotland. Each country has its unique set of procedures, charges, and legal nuances, making it essential for legal professionals and clients to be well-informed. This article aims to elucidate these differences, offering a comprehensive guide for those involved in the property market.
England and Wales: A Shared System with Distinct Practices
England and Wales largely share the same legal system, including the process for buying and selling property. However, there are subtle differences in practice and execution, especially regarding charges and fees.
In both countries, the conveyancing process begins with the acceptance of an offer and progresses through several stages: exchanging contracts, completing the sale, and post-completion formalities. The involvement of solicitors or licensed conveyancers is standard practice to handle the legal aspects.
Fees and Charges
- Stamp Duty Land Tax (SDLT): In England and Wales, buyers must pay SDLT, with rates varying based on the property's value and whether it's the buyer's first home.
- Conveyancing Fees: These vary widely and can be influenced by the property's location, complexity of the transaction, and the solicitor's expertise.
- Search Fees: These are mandatory and cover local authority searches, environmental searches, and other necessary investigations to ensure no legal impediments.
Scotland: A Distinct Legal Tradition
Scotland's legal system, including property law, is markedly different from that of England and Wales, reflecting its unique legal heritage.
The 'Missives' System
One of the key differences in Scotland is the use of 'missives' – a series of letters between the buyer's and seller's solicitors. The conclusion of the missives marks a legally binding contract, unlike in England and Wales, where this occurs at the exchange of contracts.
Fees and Charges
- Land and Buildings Transaction Tax (LBTT): Replacing SDLT in Scotland, LBTT is a progressive tax with rates that depend on the property price.
- Conveyancing Fees: As in England and Wales, these fees vary. However, the inclusion of the missives process can influence the overall cost.
- Home Report: Unique to Scotland, this report provided by the seller includes a property survey, an energy report, and a property questionnaire, adding an extra layer to the transaction.
Key Differences at a Glance
|England & Wales
|Shared but with local practices
|Exchange of contracts
|Conclusion of missives
|No equivalent to Home Report
|Home Report required
|Essential but varies
|Central, including missives
Conclusion: Navigating the Varied Terrain
The differences in property transaction procedures and charges across England, Wales, and Scotland underscore the importance of specialized legal advice. For solicitors and conveyancers, an understanding of these distinctions is crucial to provide accurate guidance to clients. For buyers and sellers, seeking professional legal assistance attuned to the specificities of each jurisdiction is key to a smooth property transaction. As the property market continues to evolve, staying informed and adaptable to these regional differences remains a paramount consideration for all parties involved.
In this case, our prompt went beyond asking Chat GPT to answer a simple factual question. We asked it to process the answers to that question into an article suited to the website of a professional lawyer. This has clearly resulted in changes to the presentation of its responses as compared with the ones it gave to our question about the evolution of SEO practices. Chat GPT still uses bullet points in parts of its article, but they no longer dominate. Instead, we find a fairly well-structured template for an article on the topic we requested, with headings that are better-suited to a journalistic end-product, and substantial sections of free paragraph text that are not reduced to bullet points.
It has generally brought out the differences between conveyancing procedures in England and Wales and those in Scotland fairly well. However we think it is still not good enough to publish on a lawyer’s website, for three main reasons.
First, it has made a critical factual error in failing to distinguish between the taxes due in England and those due in Wales, incorrectly stating that Stamp Duty Land Tax is due on property purchases in both regions, whereas in fact Land Transaction Tax is due on purchases in Wales, and Stamp Duty Land Tax only on purchases in England. Besides, it is more than just a difference in name: the rates and property value thresholds for LTT bands, being set independently by the devolved Welsh Administration, are not the same as those for SDLT.
Secondly, no details have been given at any point in the article of how the charges are calculated. Beside the omission of mention of LTT charges from the article, no information is given on how the charges for SDLT and the Scottish LBTT differ in material terms. This makes the article as a whole inadequate in its coverage of the topic of charges on house purchases. It is not useful to simply name the different charging systems without explaining how they work from the buyer's point of view.
Thirdly, there are elements of the style of presentation that are not suited to the purpose of a professional lawyer’s website, although this was stated in the prompt as being the target. It reads more as a general journalistic article than it does as one suited to the target readership of a conveyancing lawyer’s prospective clients.
The article should be addressing the fact-finding needs of those prospective clients. It is not necessary or even desirable to tell them, as Chat GPT’s articles does, that it is essential for legal professionals to be well-informed. It should be taken for granted that the lawyers whose website the article is to appear on are well-informed, and it should be apparent that they are well-enough informed for their treatment of the topic to be authoritative. Furthermore, vague statements, such as the one made by Chat GPT, in reference to the differences between the systems in England and Wales, that 'there are subtle differences in practice and execution, especially regarding charges and fees', are not useful without supporting information, in this case on what those differences actually are. The quoted statement cannot sensibly serve as more than an introduction to a more precise presentation of the differences. Yet none is given.
Overall, Chat GPT has again provided some good ideas for the structuring of an article and key points of difference between the legal and charging systems to be covered, but the voice is too detached and journalistic, and in our view does not address the needs of the law firm's clients. So the article needs to be rewritten in a much more client-focused style that makes the expertise of the lawyers on whose site it is to appear self-evident. This stylistic reworking is required in addition to the correction of the factual inaccuracy regarding SDLT and the provision of details lacking from Chat GPT’s draft of how much the charges actually are in each case.
3. AI Copywriting for creative storyline
Tell an original fairy tale incorporating a kitten, a mole, a postbox, a vagabond and a princess.
The Princess, the Vagabond, and the Postbox
Once upon a time, in a sprawling kingdom, there lived a young and beautiful Princess named Elara. She lived a sheltered life within the castle walls, her days filled with royal duties. But deep inside, Princess Elara longed for adventure and the freedom to explore her realm.
One day, while walking in the castle garden, she spotted a tiny, playful kitten with soft, glossy fur and sparkling green eyes. The kitten was adventurous and fearless, and Princess Elara, charmed by its spirit, named the kitten Whiskers.
Not far from the castle, there lived a humble vagabond named Hugo. He had no riches, but his heart was filled with kindness. One day, while wandering through the woods, Hugo came across a little mole who was lost and frightened. Hugo comforted the mole and decided to help him find his home.
Meanwhile, Princess Elara, yearning for a taste of adventure, began to write letters detailing her dreams and desires. She would drop these letters in an old, unused postbox that stood at the castle's edge, hoping that they might somehow find their way to someone who understood her longing for freedom and adventure.
One day, while helping the mole search for his burrow, Hugo came across the postbox. Curious, he opened the box and found the princess's letters. As he read through them, he felt a connection with the princess's longing for adventure. He decided to write back, offering the princess encouragement and sharing his own humble adventures.
Weeks passed, and through exchanging letters, the princess and the vagabond formed a close bond. They would share their daily happenings, with the princess's tales filled with royal duties and court intrigues, and Hugo's filled with his simple but fulfilling life, his adventures with the little mole, and his admiration for the kitten's bravery and spirit.
One day, the princess shared her desire to step outside the castle and go on an adventure with Whiskers. Inspired by her courage, Hugo wrote back with a plan for the princess to sneak out of the castle.
The next day, under the cover of darkness, Princess Elara managed to leave the castle grounds with Whiskers nestled in her cloak. Following Hugo's instructions, she found the vagabond and the little mole waiting for her by the edge of the forest.
For the first time, Princess Elara felt the thrill of freedom. They spent the day exploring the woods, playing with Whiskers and the mole, and sharing stories. The princess couldn't remember ever being so happy.
After that day, the princess, the vagabond, the kitten, and the mole became inseparable. They continued to go on little adventures, and each day, the princess would drop a letter into the postbox, detailing her latest adventure.
The princess realized that wealth or status didn't define one's life, but the courage to follow one's heart did. She felt gratitude for her new friends, for the little postbox, and for the adventures that had changed her life forever. The vagabond, the kitten, the mole, and even the postbox, held a special place in her heart, for they had given her the greatest treasure - the joy of freedom and true friendship.
And so, they lived their days filled with adventure, friendship, and laughter. And they all lived happily ever after, cherishing the special bond they shared.
This time, we put Chat GPT’s abilities to help with ideas and structure for a creative storyline to the test, by feeding it an inventory of five disparate characters and objects to be included in the storyline and the stylistic brief that they should be processed into a fairy tale.
Chat GPT’s machine learning has understood the stylistic requirements of fairy tales remarkably well. It has contextualised the princess as the primary character within a romantic setting and storyline that is a good match for traditional fairly tales, emphasising her dreams and their ultimate fulfilment. The language throughout Chat GPT’s story has an old-fashioned, romantic feel to it that is appropriate to the genre, from the use of the classic genre-specific stock phrases ‘Once upon a time’ and ‘They all lived happily ever after’ to the description of the kitten’s ‘soft, glossy fur’. Even the suggested title is a good match for the style of traditional fairy tale titles (think Beauty and the Beast).
The storyline has made good use of the contrast between the backgrounds of the Princess and the vagabond and brought them together in a way that emphasises the triumph of the human spirit over seemingly insurmountable differences in social class. It differs from classic fairy tales in that it concludes with the two main characters becoming lifelong friends, and not marrying each other, as might have been the classic dénouement for a fairy tale of this kind.
For a creative storyteller or a website hosting creative stories, the one created by Chat GPT from a rudimentary list of elements fed to it has significant time-saving potential. It is too short a story in its present form to make for a satisfactory complete fairy tale, except possibly for very small children with limited attention span being read to in class by a teacher. But it could be developed into a wonderful short story of a more satisfactory length by a creative story-teller from the starting point of the inspiration served by Chat GPT.
From the examples above, we have seen how the use of AI chatbots such as Chat GPT can help in the initial stages of developing content ideas, structures and headings for a wide variety of copywriting applications. They do not currently serve as a satisfactory complete replacement for the input of a skilled human copywriter, or anything even close to this, and the use of their output in unvarnished, unedited and unprocessed form would lead to low-quality content.
However, they can be a time-saving research aid and a valuable creative assistant, and we fully acknowledge their potential as tools assisting in the copywriting process.
For those looking to outsource their copywriting more completely to AI routines, there are already commercial offerings in the marketplace that go beyond the process we have described, and claim to offer more sophisticated tailored content than Chat GPT.
A case in point is Surfer AI. At the time of writing, the subscription-based pricing for this starts at US $139 a month for five AI-generated articles, equivalent to $27.80 per article. Its makers claim its copy is much harder than the output of other chatbots for AI content-detection routines to automatically identify as AI-generated, which will increase its effectiveness for mass content generation for SEO purposes, without incurring a penalty from search engines such as Google.
While it is probable that Google will continue to improve its algorithms for detecting AI-generated content, this and similar services may offer a shortcut for generating content that passes muster as far as the search engines are concerned.
We think it will remain important to personalise and rewrite the output of these tools to ensure the best results, but this is certainly an area where the software is evolving rapidly and the quality of its output is improving.