An ethical hints and tips checklist on how to effectively use AI for your marketing.
How do I teach ChatGPT to write in my tone of voice?
First define your tone of voice - this could be 1-3 words that match how you would write. For example, would you describe your writing style as balanced, articulate, confident, direct, authoritative, passionate.
Next define your brand persona - this will be more aligned with the background and career role you wish ChatGPT to emulate. For example, you could ask ChatGPT to “Think like a seasoned content creator” or “imagine you are an experienced senior marketing executive” - both of these prompts can be included as brand personas.
Now create your first prompt for ChatGPT - this will likely be the longest of the prompts you provide as it will be very specific and detailed to describe what you would like ChatGPT to write, how and in what way. It is here where you will include your tone of voice and brand persona.
Remember to add at the end output detail such as target platform, length and format. This could vary from “a short instagram post of no more than 30-40 words” or “a website blog post of 500 words or less” or “write in the style of a 200-word LinkedIn post with bullet points”.
The full format of the first prompt could therefore be as follows:
Imagine you are [insert brand persona from Step 2], create a short history of social media marketing as a guide for those in the publishing industry. Please write this in the tone of voice that is [insert your answer from Step 1] and write it in the style of [choose target platform] with a length of no more than [choose ideal word length].
So based on the examples given above, this could look like:
Imagine you are an experienced senior marketing executive, please create a short history of social media marketing as a guide for those in the publishing industry. Please write this in the tone of voice that is balanced and direct and write it in the style of a LinkedIn blog post with a length of no more than 500 words.
Once ChatGPT has created the first answer iteration you will most likely refine the output with more prompt-on-prompts. For example, you will most likely type:
“based on the previous reply, provide more detail about X” or “Now repeat your answer but with emphasis on” etc.
How do I know where a piece of content from ChatGPT is from?
The first and most important rule to remember is that ChatGPT is quite often incorrect when listing sources, and indeed when it come to supplying factual information it is rarely 100% faithful and accurate.
However, after ChatGPT has finished generating an answer, it is worth trying the following command:
“Please provide sources for the previous answer”
You can always follow this up with:
“Please provide URL sources”
Additionally, whilst issuing a request, you could add in the requirement to include the sources of the information, for example:
“Can you describe thermal dynamics and provide real-world examples, including the sources for these examples?”
How do I spot AI-generated phrases and sentences?
Here are 6 checks you can apply to content that might help you recognise AI generated text content:
1. Inconsistent tone and style - where human writing will unlikely display different writing style throughout.
2. Spelling - look out for US spelling rather than UK (for example specialised vs/ specialized).
3. Repetitive patterns - due to the underlying engine or sources, AI can sometimes inadvertently include repeated patterns which are easily picked up by the human eye.
4. Overuse of common phrases - often AI engines use overly generic or cliched words and phrases.
5. Too generic - particularly answers to questions may try to be covering both sides of a subject rather than provide a nuanced and insightful viewpoint in the way a human would, therefore not being able to provide a deep contextual understanding.
6. Unnatural phrasing - there is often an overall sense that the phrasing feels awkward and clunky even though it is grammatically correct.
How do I spot AI-generated imagery?
There are a number of indicative signs behind artificially generated imagery as opposed to digital art created without AI tools. Here are our top 4 to look out for:
1. Quite often the text used to create the image finds its way into the image itself but misspelt.
2. Look for abstract objects that are out of place and would be unlikely to be added by a human artist.
3. There are often logical errors particularly in landscape images where traditional human artistry would avoid. For example the physics behind how land is shaped or objects in and around water and how the water ‘interacts’ with those objects.
4. Some objects are repeated unnecessarily - something a human artist would avoid.
How do I spot AI-manipulated photography?
On closer inspection you may find photographs have been manipulated by AI. Here are 5 AI 'mistakes' to look out for:
Look for strange text that may have found its way into the photograph and is clearly out of place.
Unusual blurring - when photographs are ‘cleaned’ of unwanted objects often by using sampled pixels from elsewhere in the photograph, there can be distinct blurring that will look unusual.
By using pixels from elsewhere in the image, often nearby, to specifically cover parts of the photograph as part of the manipulation process, there is a risk of repeat sequences of pixels or objects which would not happen in a photo that hasn’t been manipulated.
In a similar way, objects added or copied from elsewhere in the photo can look immediately out of place.
Human features can be over-emphasised leading to an artificial or robotic look - especially facial features such as eyes or skin that looks too perfect.
Lastly, look for lighting irregularities such as light sources or shadows that do not match the reality of when and where the photo was taken.
Are images or graphics created using ChatGPT copyright-free?
Yes. Each iterative generation of an image is unique and no two output iterations will ever be the same as they are derived from parameters that you will have supplied. All images and graphics are copyright free and owned by the creator (you) without any commercial restriction.
What is the best practice around declaring the use of AI when creating content?
The best method of declaring the use of AI is to include a comment within the post description (for social media content) or at the end of a blog post (for example website content). Likewise, for any imagery, include a declaration either within the Image Description or immediately underneath the image. In all cases, a simple “This content was created with the help of AI” will suffice.
How do I effectively use AI for writing blog content?
The short answer is, you cannot use AI to replace human-writing when it comes to insightful, original and useful blog content. You can, however, use it to provide a starting structure and arguments for each blog.
Step 1: If you are stuck for a blog title, ask ChatGPT to give you 5 different title combinations
Step 2: Now ask ChatGPT to write a 1500-2000 blog post using this title and give it extra information as to which type of company you are, whether you wish opposing arguments to be considered, in what style to write it and from the viewpoint of a particular industry if required.
Step 3: Using the output as a skeleton guideline, use each paragraph as a way of inspiring ideas to write your own blog structured in the same way and using the same themes. Key phrases can still be used but by using the initial output you will find your writing productivity has been improved considerably.
Step 4: Any content that you do re-use be sure to check for American spellings such as ‘criticize’ when it should be ‘criticise’.