Every day, I am asked about preparing content for generative AI (GenAI) and the role of structured content. People want to know if their painstaking efforts to structure their content are all for naught now that AI seems to have arrived on the scene. And for folks who have not yet moved to structure, they want to know if they can simply skip that step and have AI take care of everything.
While you can skip structuring your content, I don’t think you should. Your GenAI output is going to be only as good as the content you use to train it. And if you curate and structure your content, making sure you have a good reuse strategy, you greatly increase your chances of success.
Nothing is worse than having the AI select the wrong information from conflicting content that it ingested. Or, just as bad, having the AI hallucinate because it is unsure of the answer and decides to make something up.
The conversation about structuring content focuses on the content you provide for the AI to ingest. Far less time has been spent exploring the content that GenAI produces. And I think it is time to give the output some consideration.
ChatGPT and other GenAI engines were designed to be chatbots. After content is ingested, the GenAI chatbot is ready to answer questions. You give it a prompt and it generates a response. In fact, prompt development is a burgeoning field in tech comm and beyond.
READ MORE: How to improve your AI performance
Two Types of AI-Generated Output Content
GenAI was designed to create new content based on the content it has ingested. That is what makes it generative. In fact, most GenAI systems generate new content whenever you provide a prompt. Even if you prompt the AI the same way, over and over. There are times when, in response to a prompt, newly generated content is acceptable or even desired. However, more often than not, we need content to be persistent rather than transitory.
What is Persistent Content?
Persistent content is content that, once written and approved, should not be changed. Its role is to serve as “the” answer. It is content that we expect to be the same each and every time we look it up. Most of the content we are accustomed to creating is persistent content. Some examples of persistent content include:
- Technical and scientific documentation
- Training manuals
- Online help systems
- Whitepapers
- eBooks
- Reference tables
The list is seemingly endless. Almost everything we write has to be reviewed by a subject matter expert. Once approved, the content can be published.
DOWNLOAD OUR FREE E-BOOK: AI In Your Pocket
What is Transitory Content?
Transitory content is content that changes each time it is published. While the meaning and intent of the content should not change, the words are not necessarily the same each and every time the answer is provided.
The current free GenAI chatbots respond to prompts using transitory content. You can ask ChatGPT the same question over and over again. You are likely to get different wording each time you ask. Its responses usually mean the same thing, but the words are almost always different.
Transitory content breaks one of the primary best practices of technical writing. Technical content should be standardized. In fact, technical content should be standardized along five dimensions:
- Words (terminology)
- Sentences (grammar and style)
- Paragraphs (voice and tone)
- Components
- Outputs
When content is standardized, it is also FAIR: findable, accessible, interoperable, and reusable. Transitory content that changes words and sentences does not follow best practices and cannot be considered FAIR.
The Negative Effects of Transitory Content
The negative effects of transitory content are many. Here are some instances where transitory content is problematic:
- People who have English as a second language
- Translation
- Regulatory content
- Content that is critical for health or safety
- Legal contracts
- Certification content
People who have English as a second language (ESL) find it much easier to comprehend information when that information is always stated the same way, using the exact same words. With transitory content, the words are always changing. This makes it difficult to understand.
If content is going to be translated, the best practice is to say the same thing, the same way, every time you say it. As long as you use the exact same words, in the exact same order, the content does not need to be translated, saving you quite a bit of time and money. In a traditional setting, transitory content will kick off a new translation project because a change in the wording will be detected.
This is why most of the content that we produce cannot be transitory. It must be persistent. We cannot afford, in either cost or understanding, to have the words be different each time we search for an answer.
Where Transitory Content Fits
Transitory content does have its place. There are lots of situations where generated text can be fluid without having an impact on comprehension. For example, if I am interacting with a chatbot on a shopping site because I want to initiate a return of merchandise, it doesn’t matter if the chatbot says, “Why are you returning the item” or “Why is the item not satisfactory” or “Please state your reason for the return.”
All of those responses are understandable. They mean the same thing. And it doesn’t matter which response I get. I can still answer the question.
I need persistent content when I look up the contraindications of a drug product. In this case, it is much more important to have a persistent answer that is validated by a subject matter expert and never changes.
Can a GenAI Chatbot Provide Persistent Content?
There are GenAI chatbots that allow you to either approve a generated response to a question (sometimes called “verified” or “curated”) or write the response that you want the GenAI to provide whenever the question is asked. In this way, the generated transitory content becomes persistent content. You can then have the persistent content reviewed by subject matter experts and guarantee it will always be used.
READ MORE: The best way to train your LLM as you move to Generative AI
But Why Deploy GenAI?
It does beg the question, though, if you are going to write or approve every response that will be provided by a GenAI, why go through all the trouble of deploying GenAI to begin with? You might as well write your answers, have them reviewed, and then use standard technology to interface with your users. You will likely get better written sentences than GenAI produces. You will be able to guarantee accuracy. It will likely be cheaper and easier to deploy.
Just because AI can generate content doesn’t mean you want it to. I think many (if not most) cases where we use persistent content are better served in the old-fashioned way: write content, have it reviewed, and publish it.
Are you a Content Rules newsletter subscriber? Sign up now and get expert tips right in your inbox.
Val is on the Advisory Board for the Technical Communications Program at the University of North Texas. When not working with customers or students, Val can be found sitting behind her sewing machine working on her latest quilt. She also makes a mean hummus.
- How to Make Conferences More Inclusive for the Hard of Hearing Community - December 2, 2024
- Preparing Content for AI: 6 Reasons Why You’re Not Ready - August 29, 2024
- How to Be Inclusive in the Workplace: My Experience as a Hard of Hearing Person - August 12, 2024
