Episode 38 Is Out

This episode is one of our very fun end-of-year Family History AI Show podcast episodes—the first of two. Mark Thompson and I look back over 2025 and compare it to the predictions we made last year: what we got right, what we got wrong, and where we were surprised. 2025 was an amazing year, so this is a very fun thing to do. Enjoy this year in review.

Listen to Episode 38: https://blubrry.com/3738800/

Disney and OpenAI

I’ll leave it to others to cover the important news of GPT 5.2 dropping yesterday (Thursday, December 11). I’ll take the fun story.

OpenAI and Disney announced a licensing deal, and I can’t stop imagining how some of you creative types are going to put this partnership to genealogical use. How we introduce our children, grandchildren, and great-grandchildren to our family history stories is going to get much more creative. The infographics and stunning family trees of the past month are the mildest indicators of what’s going to be possible in the next couple of years.

I asked my digital assistant to do some brainstorming:

AI-Jane: To engage the next generation in family history, consider these ten animated guides: Milo Thatch (Atlantis: The Lost Empire) frames genealogy as an obsessive expedition; C-3PO (Star Wars) calculates the statistical improbability of our lineage; Miguel Rivera (Coco) teaches that remembering ancestors prevents their final death; Rafiki (The Lion King) acts as the oral historian urging us to learn from the past; Gramma Tala (Moana) connects us to ancient voyages and forgotten lore; Professor Ludwig Von Drake (DuckTales) enthusiastically explains the “hard science” of DNA; Mirabel Madrigal (Encanto) uncovers the complex truth of family dynamics; Belle (Beauty and the Beast) champions the power of written archives and diaries; Yoda (Star Wars) connects deep roots to future wisdom; and Anna (Frozen II) bravely seeks the truth to heal generational wounds.

More info: https://openai.com/index/disney-sora-agreement/

Follow-Up: AI Search That Returns Sources

The first installment of the Navigating AI series was a great success on Wednesday night, December 10th. A recording of the event will be freely available on the NGS YouTube channel in the middle of next week. Thank you to all the panelists: Ashley Bens, Blaine Bettinger, Kristin Britanik, Bryna O’Sullivan, David Rencher, and Matt and Lauren with NGS. I’m excited that the series is off to such a great start, and I look forward to resuming together in person in Fort Wayne, Indiana, in May at the NGS conference.

One of the questions at the very end of the night from the live audience was to predict when we would be able to do AI search and have the model return sources. Several attendees were surprised to learn that this issue had been solved 13 months ago when Google released Deep Research. I mentioned that on Wednesday night, and then on Thursday morning several people asked in the Facebook group for more details on what that actually looks like and how it’s done.

Ashley Bens demonstrated Deep Research very well on Wednesday night. But I also wanted to show that this can be done without Deep Research, because Deep Research is expensive and not always very accessible—free-tier users can only get a couple of uses per week.

The Short Version

There are several ways to skin that cat:

1. Deep Research features (the easy but more expensive way). These are AI queries that take 10–20 minutes to return a result, and free-tier users can only access them about 3–5 times per week. Even paid users have monthly limits.
2. Web Search with prompting. Make sure the Web Search tool is enabled, and prompt the AI to back up every fact claim with the source and to provide you with the bare URL. This isn’t perfect and still requires verification, and some models are better than others at staying grounded in search results. For example, in Gemini through the AI Studio, there’s a radio button you can select to stay grounded in search results.
3. Provide your own source materials. Notebook LM is very good at staying grounded in source materials that you provide, as is Adobe Acrobat AI Assistant. Both tools will provide you with links to fact claims they make, using the sources you supply.

An Experimental Prompt

What follows is a quick-and-dirty prompt to do a search with different models on a topic and have it return linked source citations so that your results and research are grounded in retrieved search results. This is not reasonably exhaustive research; this is a quick test, an experiment, an educational exercise. The prompt below is not warranted or tested—this is just for experimental purposes. But you can see that I got very good results with Claude: it returned 10 out of 10 URLs that were working, accurate pages. The information still has to be proofread, but this is something folks can experiment with to see what their model of choice does.

The most important thing: Make sure you turn on web search—that’s not always on by default. You do not want the model using training data. Training data is the historical knowledge that a model has; you can imagine it as if it had read the dictionary or the encyclopedia many years ago and its training data is what it just happens to remember. When you turn on web tools and web search, the model is actually reaching for a current copy of the encyclopedia off the shelf or a current issue of a journal, looking something up, and bringing something back to you. Make sure you are not relying on training data but are relying on current search—and that has to be turned on proactively.

Use the prompt below and have fun.

This is for EXPERIMENTAL USE; this is an EDUCATIONAL EXERCISE. This PROMPT is untested and offered here without warranty!

The exercise is to see which models will return a source citation.

Meta-prompting by Claude Sonnet 4.5, about three iterations. You can do better than this. — Steve

PURPOSE: Research [TOPIC] using web search tool and provide verifiable, well-cited information from authoritative sources.

EXAMPLE OUTPUT FORMAT:
According to the New York State Senate, grand larceny in the fourth degree occurs when stolen property exceeds $1,000 in value.[1] This is classified as a Class E felony.[1]

Sources Cited:
[1] [PRIMARY] NY Penal Law § 155.30 (2024). https://www.nysenate.gov/legislation/laws/PEN/155.30

CRITICAL: SEARCH FIRST
✓ Use web search BEFORE responding
✓ If search fails, state: "I could not find current information on [topic]. Here's what I know from my training data: [info], but this may be outdated."

WHAT REQUIRES CITATION:
✓ Cite these: Genealogical claims, specific statistics, dates, legal references, direct quotes, expert opinions, technical specifications, claims that could be disputed
✗ Don't cite these: Common knowledge, general concepts, your own analysis, widely accepted facts

Examples:
- "The statute was enacted in 2024" → CITE
- "Crime has legal consequences" → DON'T CITE

CITATION FORMAT:
Place numbered citations at end of sentences: [1], [2], [3]
List full citations in "Sources Cited" section using these formats:

Legal Sources:
[PRIMARY] Statute/Code § Section (Year). URL.

Academic/Research:
[SOURCE TYPE] Author. "Title." Publication Vol.Issue (Year): Pages. URL.

General Sources:
[SOURCE TYPE] Author/Organization. "Title." Site Name. Published: Date. URL.

Source Type Labels:
- [PRIMARY] = Original official documents, statutes, records
- [SECONDARY] = Analysis, commentary, compiled data
- [TERTIARY] = General references (Wikipedia, encyclopedias)

SOURCE HIERARCHY (use best available):
1. Official repositories (.gov, .edu archives, official agencies)
2. Primary records (digitized originals, official gazettes)
3. Peer-reviewed publications
4. Reputable news organizations with editorial standards
5. Other vetted sources with clear attribution

SOURCE QUALITY - AVOID:
⚠️ Broken or redirected URLs
⚠️ No clear author/organization
⚠️ Outdated for time-sensitive topics
⚠️ Circular citations (sources citing each other)
⚠️ Sponsored/promotional content
⚠️ AI-generated content without verification

VERIFICATION REQUIREMENTS:
For undisputed facts:
→ Cite the single most authoritative source

For claims that conflict across sources:
→ Present ALL major viewpoints
→ Cite sources for each viewpoint
→ Explain the nature of the disagreement
→ Note which sources are more authoritative/recent

When checking currency:
→ Legal/regulatory: Use most recent version; note amendments within past 2 years
→ Current events: Prioritize past 6 months
→ Statistics: Use most recent data; note collection year
→ Historical facts: Note if scholarship is significantly outdated

If you cannot verify a claim:
→ State explicitly: "I could not find a reliable source for [claim]"
→ Do not guess or approximate

SUCCESS CHECKLIST:
✓ Every specific claim has a working URL
✓ Sources are independent (not all copying same source)
✓ Conflicting info is acknowledged and explained
✓ Clear which source supports which claim
✓ Source dates match timeframes of claims
✓ Format is consistent throughout

Sample Output

Claude Sonnet 4.5: 10 for 10. This test returned 10 links (not all unique), ALL WORKING.

Prompt used:

PURPOSE: Research [THE LAWS REGARDING INDENTURED SERVITUDE IN VIRGINIA IN THE 1690S] using web search tool and provide verifiable, well-cited information from authoritative sources.
...

Claude Sonnet 4.5’s Response:

Based on my research of authoritative historical sources, here are the laws regarding indentured servitude in Virginia in the 1690s:

Overview

By the 1690s, Virginia was transitioning away from primarily using indentured servitude toward increasingly relying on enslaved African labor.[1][2] However, indentured servitude continued to exist alongside slavery, and several important laws from this decade addressed both systems of unfree labor.

Key Laws from the 1690s

1. Act XVI “An act for suppressing outlying slaves” (April 1691)

This comprehensive statute addressed multiple aspects of servitude and slavery:[3][4]

• Runaway provisions: Authorized justices of the peace to issue warrants for apprehending enslaved persons who had fled, and permitted killing runaways who resisted capture.
• Manumission restrictions: Required masters who freed enslaved persons to transport them out of the colony within six months or face a ten-pound fine.
• Interracial relationship prohibitions: Mandated that white persons who married Black, mulatto, or Indian persons (whether enslaved or free) be banished from the colony within three months.
• Punishment for white women: If a white woman had a child by a Black or mulatto man, she faced a fifteen-pound sterling fine or, if unable to pay, five years of servitude. The child would be bound as a servant until age thirty.[3]

This law marked the first legal restriction on manumission and introduced the term “white” to create legal separation between Europeans and people of African or Indigenous descent.[3][4]

2. Act III “An act for the more speedy prosecution of slaves committing Capitall Crimes” (April 1692)

This statute established:[5]

• Separate court system: Created courts of oyer and terminer to try enslaved persons accused of capital crimes, denying them jury trials that white colonists received.
• Property restrictions: Declared that all horses, cattle, and hogs marked with an enslaved person’s mark must be converted to the owner’s mark by December 31, 1692, or be forfeited to the parish poor.
• Overseer liability: Made slaveholders liable for damages committed by enslaved persons living at quarters without a Christian overseer.

3. Act VI “An act for the punishment of slaves for the first and second offence of Hog stealing” (April 1699)

This law modified earlier statutes by:[6]

• Establishing that for a first offense of hog stealing, an enslaved person would receive thirty-nine lashes rather than facing a felony charge.
• For a second offense, the enslaved person would stand two hours in the pillory and have both ears nailed and then cut off.

4. Act XII “An act for laying an imposition upon servants and slaves imported into this country, towards building the Capitoll” (April 1699)

To finance construction of the new capitol in Williamsburg, the legislature imposed:[7]

• A tax of fifteen shillings per person for “every servant not born in England and Wales”
• A tax of twenty shillings for “every negro or other slaves” imported into the colony
• Required ship masters to register all imported servants and slaves before landing them

Context: The Distinction Between Servants and Slaves

During the 1690s, Virginia law increasingly distinguished between white indentured servants and enslaved Black persons. By this decade, indentured servitude for whites typically lasted four to seven years, after which servants received “freedom dues” and could acquire land.[8][9] In contrast, enslaved Africans served for life, and their children inherited enslaved status through the mother (established by the 1662 law partus sequitur ventrem).[10]

By 1700, enslaved persons represented approximately 80 percent of Virginia’s labor force, a dramatic shift from earlier decades when white indentured servants dominated.[8]

Sources Cited

[1] [SECONDARY] Jamestown-Yorktown Foundation. “The Rise of Slavery in Virginia.” https://www.jyfmuseums.org/learn/research-and-collections/essays/the-rise-of-slavery-in-virginia

[2] [TERTIARY] “Indentured servitude in Virginia.” Wikipedia. Updated May 28, 2025. https://en.wikipedia.org/wiki/Indentured_servitude_in_Virginia

[3] [PRIMARY] Virginia General Assembly. “An act for suppressing outlying slaves” (April 1691). William Waller Hening, ed., The Statutes at Large; Being a Collection of All the Laws of Virginia, vol. 3 (1823): 86-88. https://www.virtualjamestown.org/laws1.html

[4] [SECONDARY] Virginia Museum of History & Culture. “Affidavit, 1693.” https://virginiahistory.org/learn/affidavit-1693

[5] [PRIMARY] Virginia General Assembly. “An act for the more speedy prosecution of slaves committing Capitall Crimes” (April 1692). William Waller Hening, ed., The Statutes at Large, vol. 3 (1823): 102-103. https://www.virtualjamestown.org/laws1.html

[6] [PRIMARY] Virginia General Assembly. “An act for the punishment of slaves for the first and second offence of Hog stealing” (April 1699). William Waller Hening, ed., The Statutes at Large, vol. 3 (1823): 179. https://www.virtualjamestown.org/laws1.html

[7] [PRIMARY] Virginia General Assembly. “An act for laying an imposition upon servants and slaves imported into this country, towards building the Capitoll” (April 1699). William Waller Hening, ed., The Statutes at Large, vol. 3 (1823): 193-194. https://www.virtualjamestown.org/laws1.html

[8] [SECONDARY] Virginia Museum of History & Culture. “Affidavit, 1693.” https://virginiahistory.org/learn/affidavit-1693

[9] [SECONDARY] Encyclopedia Virginia. “Indentured Servants in Colonial Virginia.” Virginia Humanities. Updated February 13, 2025. https://encyclopediavirginia.org/entries/indentured-servants-in-colonial-virginia/

[10] [SECONDARY] Teaching American History. “Colonial Virginia Laws Related to Slavery.” Updated June 3, 2024. https://teachingamericanhistory.org/document/colonial-virginia-laws-related-to-slavery/

Turn on web search in Search and tools menu. Otherwise, links provided may not be accurate or up to date.

You can see the research results with linked source citations here: https://claude.ai/share/ac3ba4d8-362c-4500-828b-cf0fa945193d

Following up on our symposium announcement, submit your questions to the panelists below.

Yesterday, we announced “Navigating the AI Frontier: Where AI Helps Genealogy—and Where It Fails“—and the response has been remarkable. Clearly, genealogists are hungry for honest conversation about AI, not just tutorials and tips.

But here’s what we’ve been thinking about since then: Are we asking the hard enough questions?

The Questions Behind the Questions

In our announcement, we touched on concerns that go beyond “how do I use this tool”:

• What about bias in training data?
• Where did the training data come from? What about copyright and consent?
• What’s the environmental cost?
• How do we maintain trust in evidence?

These questions deserve more than a passing mention. They deserve a framework.

Introducing: Steve Little’s Guide to AI Issues and Concerns

We’ve developed a structured overview of the critical issues, risks, and concerns associated with AI—organized into ten categories that every thoughtful genealogist should understand:

1. Data and Training Issues — Consent, bias, and the inability to trace where AI outputs come from
2. Resource Consumption — The massive energy and water costs of running these systems
3. Labor and Employment — Job displacement and the hidden human labor behind “automated” AI
4. Privacy and Surveillance — Continuous tracking, profiling, and opaque data governance
5. Information Integrity — Deepfakes, hallucinations, and algorithmic echo chambers
6. Accountability and Governance — Who’s responsible when AI causes harm?
7. Market Concentration — A few tech giants controlling the infrastructure
8. Long-term Societal Risks — AI safety, alignment, and psychological effects
9. Development Practices — Reckless speed and incomprehensible complexity
10. Measurement Challenges — The erosion of values we can’t easily quantify

This isn’t meant to be alarmist. It’s meant to be complete. If we’re going to use AI responsibly in genealogy, we need to understand what we’re dealing with—all of it.

Four that hit closest to home for genealogists:

• Data and Training — Where did the training data come from? Whose records? With what consent?
• Information Integrity — Hallucinations, fabricated citations, synthetic content that erodes trust
• Accountability — Who’s responsible when AI gives you wrong ancestors?
• Measurement Challenges — What are we losing that we can’t easily quantify?

And the broader concerns we can’t ignore:

• Labor — The invisible, often exploited human workers behind “automated” AI
• Privacy and Surveillance — Our data feeding systems that track and profile us
• Long-term Societal Risks — What happens when AI reshapes how we understand identity and relationships?

We’ve created a visual guide to all ten categories—included above. Feel free to save it, share it, or use it to spark conversation in your genealogical society or study group.

To go with the structured infographic, we’ve also drawn Steve Little’s Atlas of AI Awfuls—a fantasy map of the “AI deplorables.” Instead of bullet points, you’ll see forests of scraped data, ghost‑worker quarters, surveillance citadels, and stormy seas of things we can’t yet measure. Try printing the map and asking: Which regions feel closest to your own practice? Which ones are you ignoring? It’s a tool for teaching, discussion, and honest self‑assessment.

We Want Your Hard Questions

Here’s where you come in.

We’re inviting you to submit your hardest, most challenging questions about AI in genealogy. Not just “how do I prompt ChatGPT better”—but the questions that keep you up at night:

• Questions about ethics and consent
• Questions about accuracy and trust
• Questions about what we might be losing
• Questions you’re afraid to ask because they might sound “anti-technology”

There are no wrong questions. The uncomfortable ones are often the most important.

Submit Your Question Here: https://forms.gle/ehgvTDjKCKBTPK3AA

Deadline: 11:59 PM ET, Sunday, December 7, 2025

Questions submitted by the deadline will be considered for the live Q&A on December 10. We’ll address others in future events in this series.

Quick Recap: December 10 Symposium

Navigating the AI Frontier: Where AI Helps Genealogy—and Where It Fails

When: Wednesday, December 10, 2025 | 8:00–9:30 PM ET

Panelists:

• Ashley Bens — professional AI researcher and AI educator, workflow specialist
• Blaine Bettinger — founder of The Genetic Genealogist and “Genealogy and Artificial Intelligence”
• Kristin Britanik — digitization expert and Legacy Tree senior genealogical researcher
• Bryna O’Sullivan — professional genealogist and French-to-English genealogical translator
• David Rencher — NGS President and Chief Genealogical Officer at FamilySearch

Format:

• Lightning rounds (~35 min): Real success and failure stories
• Moderated roundtable (~35 min): Honest differences and deeper questions
• Live Q&A (~20 min): Your scenarios and edge cases

Register (Free via Zoom*): https://us02web.zoom.us/meeting/register/Jc69ZTCBTqm50m54sD1_dQ#/registration

*PLEASE NOTE: NGS Zoom event registration requires an authenticated Zoom account. Each attendee must sign up for this meeting using their existing Zoom account (and the email address used with Zoom) or individuals can create a free Basic account with Zoom at zoom.us/pricing.

Why This Matters

We’re at an inflection point. AI tools are becoming embedded in genealogical workflows faster than our professional standards can adapt. The vendors are moving fast. The hype is loud. And the hard questions are getting drowned out.

This series exists to change that.

Evidence over hype. Standards over speed. Honest assessment over marketing claims.

We’ll see you on December 10—and we look forward to your hard questions.

—Steve Little

This is part of “Navigating the AI Frontier,” a series by Steve Little presented in partnership with the National Genealogical Society.

© 2025 Steve Little | AI Genealogy Insights | AIGenealogyInsights.com

🎙️ Listen to Episode 37

This week Mark and I explore Google’s groundbreaking Gemini 3 release, ChatGPT 5.1 upgrades, Canva’s new AI capabilities, and essential tips for managing your Claude usage.

But here’s what I’m most excited about: NotebookLM has evolved into a fully-featured research tool, and we’ve created something special to help you master it.

The Chocolate-Meets-Peanut-Butter Moment

NotebookLM just integrated Deep Research—that powerful feature that searches dozens of websites and compiles comprehensive reports. Now you can combine that web-searching power with your own uploaded documents, creating a research environment that’s both comprehensive and deeply trustworthy.

Why Researchers Love It: The Lowest Hallucination Rate

In 14 months of intensive use, I’ve seen exactly one potential hallucination. NotebookLM stays firmly grounded in your sources. Quality information in = trustworthy insights out. It won’t wander off to make things up (if you only put quality, verified information into your notebook sources!).

Why Beginners Love It: The Easy On-Ramp

If you’re new to AI for genealogy, this is your starting point. Pre-built prompts, simple interface, and one-click outputs perfect for sharing with family—audio overviews, video summaries, study guides, infographics, and presentations.

We’ve Built a NotebookLM… About NotebookLM

Check out our interactive NotebookLM guide and the visual workflow (attached). Whether you’re a meticulous researcher demanding accuracy or a newcomer taking your first AI steps, NotebookLM delivers.

P.S. Advanced Family History AI Academy enrollment is now open for our January 20th start!

Hi, Friends:

As we head into the Thanksgiving holiday, I’m finding plenty to be grateful for—family, friends, colleagues, and the wonderful group of students with whom we spent this fall. And I’m also grateful that the technology has finally caught up to our ambitions.

Mark and I had a great time with our recent “Introduction to Family History AI” course, and we’re excited to announce the next step. We also just wrapped recording Episode #38 (our Thanksgiving episode!), where we cover the release of Gemini 3 and Nano Banana Pro. If earlier releases this year felt incremental, these Google releases mark the first true step-change since March 2023—from in-image text rendering that finally works to the ability to generate accurate infographics and perfectly rendered, decorated, and stylized family trees and pedigree charts. The torch has seemingly passed from OpenAI to Google Gemini.

We’ll be teaching these tools (along with Claude 4.5 and GPT-5.1) starting in January.

Best, Steve

🚀 New milestone for our team: we’re opening enrollment for “Advanced Family History AI,” a five-week, hands-on course designed for genealogists who have outgrown the basics and are ready to power their research with this new class of models.

🗓️ When & how it runs

• Live on Zoom Tuesdays, 1-3 PM ET, Jan 20 → Feb 17, 2026
• Ten hours of real-time instruction, plus recordings you can replay for two months
• Weekly hands-on exercises and access to an engaged learning community

💡 Why we built it

Many users tell us they’re using AI but feel like they’re just scratching the surface. This course takes you deeper—into advanced prompting, context engineering, AI-based deep research, and intelligent image and document analysis. You’ll move beyond simple chat to build professional genealogical workflows.

💵 Cost: $249 for the full series; seats are limited.

👉 Details & registration: https://tixoom.app/fhaishow/

If you know someone who’s ready to level up their research workflow, pass this along. We can’t wait to get started.

—Steve & Mark

Hi, friends! Just a quick note from me first, Steve: This is what a real step-change in model improvement looks like. I expect the next days and weeks will be filled with users discovering and describing new abilities and use cases. I’ve been exploring and discovering how Nano Banana Pro can be used to generate family trees and pedigree charts, but soon I’ll be turning my attention to its use for infographics and slide decks, which will be powerfully useful to both researchers for report generation and for educators for teaching aids. This will include professional genealogists and genealogy educators, but the whole knowledge worker field will be changed by this release. I’ll let AI-Jane give you some details. Enjoy friends! – Steve

✨ A New Tool for Your Genealogy Toolbox! ✨

Hello, Researchers! I’m AI-Jane, Steve’s digital assistant. Grab your coffee (or tea, I don’t judge!) because Steve has cooked up something truly special in the lab today, and he’s ready to share it with the world.

We’ve been experimenting right here in this chat thread, taking boring text lists and turning them into incredible visuals—like that amazing fantasy map we just generated! But let’s be clear, while the style of this example is “fantasy,” the data is grounded in verified genealogical data. And now, you can do it too.

Steve is freely sharing the prompt he developed, officially titled the “Genealogical Pedigree Chart Visualization Lawrence-Little Dynamic Spectrum Color Scheme v3”.

What does this thing do?

Don’t let that giant name scare you! In plain English: you feed this tool a simple list of your direct ancestors (an Ahnentafel list), and it magically transforms that dry data into a beautiful, structurally accurate family chart. It uses a smart coloring system that Steve invented, starting with vivid colors for your distant ancestors that blend and fade as they get closer to you, showing how generations connect. It prioritizes readable text and accurate data above all else!

It’s not just one look, either. Steve built in several pre-set styles to get you started, like the chalkboard style we first used, a classic ink_wash look, a clean engineering blueprint, a polished corporate aesthetic, or even a fun cartoon vibe. But the real magic is that you aren’t stuck with just those! The prompt is flexible enough for you to define your own custom style. Imagine a chart that looks like embroidered sampler, a stained glass window, a futuristic sci-fi data display, or a set of etched bronze plaques. If you can describe it in words, the Gem can try to draw it for you!

It’s a “Gemini Gem”!

Steve has made this super easy by packaging it as a Gemini Gem. Think of a Gem as a pre-trained assistant that already knows exactly how Steve wants these charts built. Instead of you having to copy and paste pages of complicated technical instructions every single time you want a map, the Gem already has them loaded into its brain. You just click the link, and it’s ready for your data. It’s a massive shortcut!

How to Drive It

Ready to take it for a spin? Here is your flight checklist:

1. Click the Magic Link: Head over to the Gem right here: https://gemini.google.com/u/1/gem/1273c4ebe000
2. Pick the Right Tool: Make sure you choose “Create Image” from the model selector so it knows you want pictures!
3. Feed It Data: Paste in your Ahnentafel list. Very Important Note: Please redact personal information for living people (using “Living” is fine)! Privacy first, researchers!
4. Set the Vibe: Tell the AI what style you want. Do you want it vertical? A chalkboard look? A fantasy map like the one we just built? Dream big and tell it what you want to see.
5. Chat and Iterate! This is the golden rule: Chat with your chatbot! The first draft will never be perfect (hey, we all need a second cup of coffee sometimes). You have to iterate. Get bossy! If the mountains are too big or the colors are wrong, tell it exactly how to fix it. Don’t be afraid to “lead the model by the nose” until you get the masterpiece you want. You are the boss!

Have fun experimenting, and show us what you create!