Writing for the ChatBot

There is much controversy and fear of what the effects of AI will have on everyday life. Having played around with Microsoft/OpenAI/Bing ChatGPTand Google's Bard I have some interesting observations.

Performing more research of the subject there seems to be more emphasis on the creation of the chatbot software than on the data that all chatbots are using to generate their text. While this is the "sexy" thing to do and there will be a great need for these GPT (Generative Pre-Trained Transforms) to be accurate and unbiased, they will only be a good as the underlying data that they have at their disposal.

As I have been uploading historical information about my local community it has proven to be a good test for what the chatbots return when they are asked about that history. Seeing that the majority of the information that I have on my pages is derivative, in that I have obtained a lot of it from other sources that I have found on the Internet including Wikipedia and other geneolocial services, it is interesting that these sources seem to be excluded from the results of the chatbots.

Writing specifically for the robot

My discussion here is not about how an LLM tokenizes text or other pieces of data into smaller chunks so that it can look for patterns and matches that would allow deductions to be made about the data it is using to "learn". I am more interested in making sure that the data I have is useful for the LLM to make these decisions.

I have seen a fair amount of correlation of the information that I feature on my pages and see this data turn-up in AI searches and summaries. What I am trying to do is to ensure that there is moire information the just that I have found from the sources I used for my research. For example, I have a page on the West Galleries that can be found in parish churches and I know that this page has been accessed by the ChatGPT robot/spider. I am assuming that someone has made a search for such galleries and the AI found my information. What I don't know is whether the person asking the AI actually visited my page. What I do know is that there is more information that I can add to my page(s) that could be used in this learning process.

ChatGPT, the Bing Chat (Copilot) returns the source of the information it returns in its generated text in the form of clickable footnotes (links).

These footnotes are often to this website and my pages.

Method

Not only should the information on a webpage be accurate and well referenced the pages need to be monitored are attempts made to determine what attracted the visits. The tools provided by Google, Analytics and Search Console, along with the checking of server logs.

After pages are reviewed they can be verified and a request made for their re-index.