I found this in my email this morning:

Bots are definitely trolling websites, as this bot found my Saturday Night Genealogy Fun post about the parish of St. Demetrius, which serves the villages of Hajtovka and Udol, Slovakia.

It looks like it has collected free images available on Wikimedia taken in Hajtovka. No, I didn’t click on any links.

Notice it also says “If a grounded set like this would be worth keeping, I can have it printed and sent for $20.”

Nope, I think that offer is a hard pass.

Be aware from where and who your emails are coming!!! In this case, it wasn’t even a WHO, it’s a WHAT.

I’ve now moved into Chapter 2 of Diana Elder’s and Nicole Dyer’s workbook, trying to develop some new skills using AI.

My first roadblock – I used a sample Practice Prompt suggested in the workbook [involving data on direct line ancestors], but I was told I had reached my limit for today and to either upgrade to the paid version or come back in 24 hours (not tomorrow morning, but an entire day later.) That happened with both Perplexity and ChatGPT.

Gemini was my third attempt. It completed the task, but totally missed the “direct line ancestor” instruction and seems to have lied since I have no ancestor anywhere in my tree named George Sabo born in 1898. That was the first of many made up individuals.

I haven’t been impressed with Gemini when I’ve used it in the past and I’m even less impressed with it now. Gemini is a “no-go” for me from now on.

UPDATE: I waited a few days and went back to Perplexity to try out a suggested Practice Prompt with an ahnentafel PDF file attached. I was impressed with the answer. (I’m not sharing the prompts because they are proprietary to the authors.)

REPLY:

Below are up to 10 ancestors born between 1770 and 1850 in what is now Canada, as found in your ahnentafel PDF. Each entry includes the ancestor’s full name, approximate age at death, birth/marriage/death dates and places, plus names of parents if present.

[Ten ancestors listed.]

Observations

• The above list shows a concentration of ancestors in New Brunswick, with research potential for Loyalist, Acadian, or Scottish migration stories.
• Dates and locations are preserved as found; some further validation or deepening can be done by closely reviewing the cited ahnentafel entries.

These individuals each have ties to Canadian history during the period of 1770–1850 and should provide a rich starting point for family history research projects, especially if focusing on early communities in New Brunswick or Quebec.

I then requested three research questions based on Daniel Astle (1783-1817), one of my Canadian Loyalist family ancestors:

Here are three research questions to guide a study of Daniel Astle (1783–1817) within his historical context:

Migration and Settlement Patterns

• How did the arrival of Loyalists and other waves of immigration after the American Revolution influence daily life and land ownership in New Brunswick and Quebec during Daniel Astle’s lifetime?

Political and Social Change

• In what ways did the creation of New Brunswick (1784) and later administrative changes in Canada (like the division into Upper and Lower Canada) shape the identity, governance, and opportunities for British and French settlers like those in Daniel Astle’s community?

Impact of International Events

• How did tensions and conflicts such as the aftermath of the American Revolution, border changes, and the war preparations leading up to the War of 1812 affect the security, economy, and family life of Canadian residents in the late 18th and early 19th centuries?

These questions encourage investigation into the broader historical forces shaping Daniel Astle’s world, connecting personal biography to significant social, political, and international events of the era.

These are excellent suggestions to enable me to learn more about daily life in early New Brunswick, Canada. [The Astle family was only in Quebec for a couple of years.]

The last part of Chapter 2 was focused on writing research objectives and there was an assignment to created a research objective that I could use for the remaining chapters of Research Like a Pro with AI Genealogy Workbook.

My takeaway from Chapter 2 is the importance of writing AI prompts that are filled with as many details as possible. The fewer details, the narrower the response.

On to Chapter 3!

In Part 1 of my foray into Research Like a Pro with AI Genealogy Workbook, I outlined the format of the workbook and shared an abbreviated version of the Table of Contents.

The Introduction provided an overview of the development of AI, tools that we currently have available and how to make best use of this workbook.

Chapter 1 has its focus on learning a bit about the four main AI tools in use today – ChatGPT, Claude, Gemini and Perplexity – and a format to follow to craft effective efficient prompts for genealogical research.

I’ve used all four platforms in the past – only the free versions – and learned something interesting with the genealogy group that I teach. There were six of us, all using my house internet. I gave the exact same prompt to everyone and we shared out results. Curiously, NO TWO of us had the exact same responses – they varied slightly. However, one person had a response that wasn’t anything like what the rest of us had even though she used the exact same prompt wording.

This time around, I didn’t try giving the same prompt to the same platform more than once, and I already had a couple of favorite platforms, but I did want to see the replies from each bot to the same prompt:

Role: Professional genealogist specializing in early New Brunswick, Canada research. Task: Search the web for information about primary records pertaining to Loyalists who settled in New Brunswick, Canada before 1790. Provide details about the repositories with primary record collections and books written about Loyalists in Parrtown before 1790. Create a list of the resources, repositories and websites found.

I will state, right from the start, that Gemini is my least favorite bot and it provided the worst results – only about 2 pages of information – when the other three bots provided 4-5 pages of details.

Of the remaining three – ChatGPT, Claude and Perplexity – ChatGPT was the second weakest, although I was surprised to see that only ChatGPT (5) included mentions of Black Loyalists or the New York Public Library Manuscripts Division or the New Brunswick Genealogical Society.

Yes, the focus is on New Brunswick records, but the prompt doesn’t specify only New Brunswick resources. Many Loyalists sailed to Canada from New York and there are primary records to be found there.

Aside from that, ChatpGPT 5’s font was much larger than that of Claude and perplexity plus the last page was just a summary of the items on the first 3 pages.

Let’s say I was not impressed with ChatGPT 5’s performance. On top of that, I wanted to print out hard copies of each of the responses so I could go through and compare repeated resources and those unique to one response. Apparently, to print is an upgrade on ChatGPT so I had to copy and paste into Word. Doable, but kind of a nuisance.

Claude and Perplexity both produced well-done responses and both had easy-to-use download/print buttons.

CLAUDE grouped resources by Major Repositories and Archives (University of New Brunswick, Provincial Archives of new Brunswick, Library and Archives Canada and FamilySearch collections), Primary Record Collections (Land, Military, Court and Vital Records), Specific Parrtown/Saint John Records (Key Collections and Important Documents), Published Works and Books (Only one book mentioned though, which is a big weakness here), and Special Research Notes. There were a few online links, but, for the most part, it will take a bit more digging to land on the websites. Its font size was on the big side so its 4 pages would equal about 3 1/2 compared to Perplexity’s font size.

PERPLEXITY, overall, had the most thorough response, with five pages of detailed information using the smallest font. It covered Major Repositories and Websites, Books and Major Collections on Loyalists in Parrtown Before 1790, Key Primary Record Sets and Finding Aids, Summary List of Top Resources, Historical Overview of Parrtown, Major books, Major Archival Manuscript Collections, Principal Repositories and Websites and closed with 18 source citation links.

As a retired teacher, in my mind, I graded the responses for each. Perplexity was clearly an A, Claude a B, mostly for the lack of book titles provided, ChatGPT got a C- and Gemini a D.

The remainder of Chapter 1 involved using a format for creating effective prompts, techniques for creating more advanced prompts [five steps which I will be following], ways that bots can transform information and ethical and legal considerations when using AI.

As I am completing Chapter 1, I want to say upfront because of my own personal bias, I adjusted the tools I used to get set up and ready to begin this workbook. Google Docs and Airtable are both highly recommended, but I am choosing not to use either for one simple reason. I don’t like my original work to be sitting in the cloud and much prefer to have it saved on my own computer and will be using Word and Excel instead. As stated, that’s just a personal opinion. The cloud is fine for backups, but not original work.

The Introduction and Chapter 1 practices were interesting. Since I’ve used AI tools already, I think the most useful tip I’ve learned is from the Crafting Effective Prompts section using the five detailed components for prompts.