Start free trial Book a demo
Webinar on Using search analytics for improving content - June 26, 2024 | 2:30PM UTC - Register Now!
writing content for chatbots

Best practices in writing content for Generative AI-based chatbots

Category: Customer Support

The emergence of Generative technologies such as ChatGPT has changed the game in which technical writers produce content. Large Language Models are text hungry and need more text to understand the context better. Thus, technical writers need to cater to both human readers and the Generative AI-based chatbots that consume the content! Writing characteristics for human readers and Generative AI models are very different. In this blog, we shall explore some emerging best practices in producing content for Generative AI-based chatbots.

New content creation

When creating new content from scratch, it is essential to adopt emerging best practices such that new content is positioned for the ChatGPT-based chatbot. Knowledge base articles need to be positioned such that new content adheres to the following principles

  • Clarity through elaborate explanations
  • Conversational tone through narratives
  • Consistency through business glossary

It will be easy for ChatGPT-based agents to ingest all the knowledge base articles and provide accurate answers based on the questions submitted by your readers.

Clarity through elaborate explanations

Content brevity is more suited for human readers, whereby technical writers understand the limitations of the human attention span and their rational behaviors. However, Generative AI based agents are text hungry and need more content to better understand the context of the content. Thus, it is imperative that technical writers create elaborate content by following practices such as

Simple language

Use simple language to create content and avoid ambiguous words to ensure clarity. Instead of making the content simple, use simple language to add more content.

For example, if you are explaining the functionality of a product feature,

Simple content: “This feature helps to save time in analyzing feedback. This feature can be configured by navigating to the Settings menu and enabling the toggle button ‘Use assist intelligence’”

Traditional content: “This feature can be configured with enabling ‘use assist intelligence’ under the setting menu.”

Another example of old practices is “contact us Sales: +1 548 78945, Support: +1 789 124580”

The new practice is “contact our sales department using the phone number +1 548 78945 and contact support department using the phone number +1 789 124580”

Semantic content structure

Structuring the content using H1 – H6 headings provide a way to understand the importance and hierarchy of content. When human readers search for answers, this semantic structure helps get the right and accurate content for answering questions.

Images and videos

For human readers, it is very important to include images, screenshots, animated GIFs, and videos in the content for easy comprehension. However, these artifacts need to be annotated with more metadata for ChatGPT-based agents to understand.

Find out how Document360 can help you to reduce support tickets!

Book A Demo
Document360

Conversational tone through narratives

A conversational tone is an important aspect, such that we can pick up the most frequently asked questions and provide answers right away. To build new conversational content, technical writers need to collaborate extensively with the customer support team to curate frequently asked questions.

Create FAQs (Frequently Asked Questions)

Create at least 4–5 FAQ-type questions for each knowledge base article. These FAQs can be troubleshooting questions, known product bugs, limitations of product features, and so on. Once this content is ingested, it is easy for ChatGPT-like agents to respond to customer queries quickly.

Personas

It is also important to establish personas in your content such that ChatGPT-like agents can imitate those personas. For example, you have a few roles in your product features such as admin, owner, user, and so on. In that case, making a reference to these personas in each knowledge base article helps with narration in a more natural way.

In fact, you can force ChatGPT-like agents to act as an admin, and then questions can be answered using that persona.

Inclusivity

The content needs to be inclusive such that it spreads harmony among human readers. The content should be accessible to all human readers and be human-friendly

Consistency through Business Glossary

Using consistent terms across your knowledge base is important in avoiding ambiguity. This sets a precedence for a consistent business glossary to be established and used consistently. Technical writers need to be trained on the corporate business glossary during the onboarding process. A consistent business glossary also avoids causing confusion in the answers.

For example, take this content “ the users of the software must configure this feature. It is important for the staff to make sure that product is configured for providing better security”. Here are the terms users and staff will confuse human readers.

Migrating the Old Content

The world is moving from “searching for information using keywords” to “needing accurate answers to their questions”. Thus, migrating old content to this new paradigm is important to future-proof your knowledge base. Establishing a change management plan through which old content is revised based on the emerging best practices to produce content for Generative AI-based agents is crucial. When reviewing the old content, care should be taken to

  • Balance the content to be more suitable for human readers and Generative AI-based agent; Tools such as Article summarizers helps human readers understand elaborate content faster
  • Adopting consistent business terms
  • Reducing the extensive usage of images and videos
  • Annotation of images and videos
  • Adding more FAQ-style content

Conclusion

ChatGPT has changed the way human readers discover new information. Gone are the days when we search using keywords and browse to different URLs to seek the needed information. Nowadays, human readers demand accurate answers to their questions. The knowledge base must have clear content such that ChatGPT-like agents can ingest the content and be equipped to answer questions from human readers. Following the emerging best practices to produce Generative AI – friendly content would help technical writers adapt to the newer breed of content consumers!

An intuitive knowledge base software to easily add your content and integrate it with any application. Give Document360 a try!

GET STARTED
Document360
Using search analytics for improving content                            
                       
                     
Get the latest customer support stories & observations

By signing up, you agree to our Terms, Policy and GDPR

Related Articles