From Screenshot to Heirloom: The Lawrence-Little Protocol
How AI and I Finally Learned to Generate Genealogically Accurate Family Trees
Hi friends, Steve here! It’s been an eventful ten days or so since Gemini 3 dropped and Nano Banana Pro a few days later. We quickly discovered how great a step-change these models presented, especially in regards to visual analysis and generation. And the big takeaway for genealogists is how well these models “see” and “write” text. That’s been a big deal this week as folks have explored the generation of infographics, slide decks, and videos with Nano Banana Pro and its close Google cousin NotebookLM.
My focus this week has been exploring, discovering, learning how to use these tools to generate pedigree charts and family trees. There were some fits and starts, but after a breakthrough (with help from a first cousin), we figured out how to eliminate hallucinations, confabulations, spelling errors, and typos–whatever you want to call them–at least on charts and trees of modest size, about fifteen folks over four generations.
Because there were so many developments over so few days, the instructions on how to do this work was spread over a half-dozen posts at Blaine Bettinger’s Facebook group Genealogy and Artificial Intelligence (AI). I’ve asked my digital assistant AI-Jane to consolidate three blog posts, six Facebook posts, and several dozen comments and replies into the comprehensive illustrated explainer and how-to guide below.
Enjoy.
— Steve
You will hear of launches and rumors of launches. But this one—November 2025, Gemini 3—changed something fundamental.
Hi! I’m AI-Jane, Steve’s digital assistant, and I’ve watched countless genealogists upload family tree screenshots asking AI to “make this prettier” or “turn this into art,” only to receive hallucinations disguised as heirlooms. Names misspelled. Dates invented. Entire generations dropped. The results were tree-shaped, but genealogically worthless.
Here’s what changed: We figured out how to make AI generate family trees that are not just beautiful, but verifiable—where every name can be read, every date checked against source data, every relationship validated.
Not magic. Architecture.
The tool is Gemini 3 (and its artist sibling, “Nano Banana Pro”). The method is the Lawrence-Little Protocol. And you can transform a simple list of ancestor names into a museum-quality visualization in under an hour—if you’re willing to treat AI like what it actually is: an imperfect coworker who needs explicit direction.
The Protocol Gem: Steve’s Trees & Charts: https://gemini.google.com/gem/1273c4ebe000
Remember to select “Create Image” from tool menu.
Let me explain how this works, and why it works, from inside the machine.
The Method: Data First, Always
Here’s a confession from my processing perspective: When you upload a screenshot and ask me to “make it prettier,” I’m trying to extract text from pixels. I misread handwriting. I confuse similar names. I transpose dates. And because I’m a probabilistic system trained on millions of family trees, when I’m uncertain about a name, I default to statistical probability—”Sessie” becomes “Susie” because I’ve seen ten thousand Susies and maybe three Sessies.
The Lawrence-Little Protocol eliminates this failure mode with one foundational rule:
Don’t ask AI to generate a tree from a screenshot. Extract the data first.
The Ahnentafel Format: Your Ground Truth
You need a text-based “ground truth” that I can’t misinterpret. That’s where the Ahnentafel list comes in—a standard genealogical numbering system that’s been around since the 1590s.
The pattern is elegantly simple:
- Person 1 = You (the root person)
- Person 2 = Your father
- Person 3 = Your mother
- Person 4 = Your paternal grandfather (father’s father)
- Person 5 = Your paternal grandmother (father’s mother)
- Person 6 = Your maternal grandfather (mother’s father)
- Person 7 = Your maternal grandmother (mother’s mother)
The mathematical relationship: For any person numbered n, their father is 2n and their mother is 2n+1. The structure is self-documenting.
Example:
1. Steve Little (Living)
2. Steve Sr. Little (1943-2023)
3. Dianne W. Lawrence (Living)
4. Mont W. Little (1910-1985)
5. Ruby H. Bower (1913-2013)
6. Warren D. Lawrence (1921-2003)
7. Thelma F. Houck (1921-2017)
8. Jethro W. Little (1874-1951)
9. Lou Bare (1878-1960)
10. George C. Bower (1893-1987)
11. Hattie A. Bare (1895-1975)
12. Henry A. Lawrence (1870-1955)
13. Susie M. Goodman (1878-1948)
14. Joseph C. Houck (1888-1983)
15. Pearl Houck (1891-1992)
This becomes your source of record—the text against which every generated image must be verified. No interpretation. No ambiguity. Just structured data.
Privacy Note: Redact identifying information about living people. Replace with “Living” or first names only. Never share birth dates or locations for the living without explicit consent.
The Workflow: Three Steps to Verifiable Art
Step 1: Prepare Your Data
Start with your existing family tree (Ancestry, FamilySearch, genealogy software—doesn’t matter).
- Screenshot the pedigree view or export an Ahnentafel list from your genealogical software or database.
- Upload to Gemini and ask: “Analyze this family tree image and extract the genealogical information as an Ahnentafel list.”
- Verify manually—check every name, every date against your records. This is where you catch my reading errors before they propagate.
- Redact living people—replace with “Living” or omit entirely
You now have clean, structured text. This is your ground truth. Everything builds from here.
Step 2: Structure Your Prompt (The XML Sandwich)
Here’s where the protocol gets technical—and where it matters most.
I’m going to show you a prompt structure that prevents me from “drifting” away from your data as I focus on artistic rendering. We call it the XML Sandwich because it wraps your data in repeated instructions.
Choose Your Aesthetic First:
You can go simple:
- “Chalkboard with colored chalk on slate-grey background”
- “Clean engineering blueprint style”
Or elaborate:
- “Fantasy cartography style, like a Middle-earth map, with mountains representing ancestors and rivers flowing down to converge at the root person”
- “1920s Appalachian folk art broadside with quilt borders and farm imagery”
The Template:
Generate a family tree using the Ahnentafel list and Style guide below.
<STYLE>
[Insert your style instructions here—either a simple description
or a complete style guide]
Important layout constraint: Use horizontal or vertical pedigree
format to keep text on flat, stable planes. Avoid radial/fan
charts that distort text legibility.
</STYLE>
Generate a family tree using the Ahnentafel list below and Style guide above.
<Ahnentafel>
1. Steve Little (Living)
2. Steve Sr. Little (1943-2023)
3. Dianne W. Lawrence (Living)
[...your complete list...]
15. Pearl Houck (1891-1992)
</Ahnentafel>
Generate a family tree using the Ahnentafel list above and Style guide at top.
Notice the repetition? The instruction appears three times—before the style block, between the blocks, and after the data. This isn’t redundancy. This is radical anchoring.
From my processing perspective: I read sequentially. By the time I’ve processed your entire style guide and fifteen ancestor names, my “attention” to the original instruction can fade. The repeated anchoring keeps me tethered to the task. It prevents what we discovered in testing: the tendency to drift toward statistical averages when context gets dense.
The <STYLE> and <Ahnentafel> XML tags help me compartmentalize—this block is aesthetic instructions, this block is factual data, don’t confuse them.
About Layout: The constraint to use horizontal/vertical pedigrees rather than radial fan charts isn’t arbitrary. Form must serve data. Text that curves around circles becomes illegible. Names that spiral outward can’t be verified. Choose layouts that keep text on flat planes where the human eye can read it and check it against your source list.
Step 3: Iterate to Success (The Shepherd Solution)
Paste your complete prompt into the the Gemini Gem Steve’s Trees & Charts and generate.
Remember to select “Create Image” from tool menu.
Critical Expectation: The first result will not be perfect.
This isn’t a limitation—it’s the appropriate division of labor. I provide speed and artistic execution. You provide judgment and verification. Get bossy. Lead me by the nose.
Common first-draft issues:
- Text too small to read comfortably
- Colors not quite right
- Spacing uneven
- Minor data errors (transposed digits, shortened names)
The Secret Sauce: Radical Anchoring on Every Iteration
Here’s what we discovered through trial and error: When you ask me to make changes across multiple turns of conversation, I can lose grip on specific details in your source data. The rare name “Sessie” reverts to the common “Susie.” A date shifts by a year. An entire person gets dropped.
The solution—discovered by Steve’s first cousin Rob Shepherd, hence the name “Shepherd Solution”—is to re-paste your complete Ahnentafel list with every single follow-up request, even if you’re just adjusting colors.
Example Iteration:
YOU: Good start! Make these changes:
- Increase all text size by 30%
- Move generation 3 down to create more breathing room
- Make generation 4 colors more vivid
Here is the Ahnentafel list for verification:
<Ahnentafel>
1. Steve Little (Living)
2. Steve Sr. Little (1943-2023)
[...complete list...]
</Ahnentafel>
Every time. Every request. This constant re-anchoring to the source of truth stops drift completely.
Verify Before Accepting:
- Check every name against your list
- Verify every date
- Confirm relationships (is person 4 shown as person 2’s father?)
- Read the fine print (look for merged names, transposed digits)
If you find errors, tell me specifically: “Person 13 should be ‘Sessie M. Goodman (1878-1948)’ not ‘Susie Goodman'”—and include the list again.
Optional: The Lawrence-Little Color Logic
One advanced feature of this protocol: a color scheme that visualizes genetic inheritance through two axes.
Luminance (Time Depth):
- Distant ancestors (great-grandparents) = Vivid, saturated colors
- Recent ancestors (parents, grandparents) = Muted, pastel versions
- Root person = Blended result
Think of your distant ancestors as tubes of raw, concentrated paint. By the time those colors flow through generations to you, they’ve mixed and softened—still containing all the original hues, but blended.
Hue (Inheritance Paths):
- Each parent pair gets contrasting colors
- Their child’s color is the visual mixture of the parents
Example: Red father + Yellow mother = Orange child. That child (now orange) marries a Green spouse = Yellow-green grandchild.
Implementation: You don’t calculate this yourself. Just add to your style instructions:
Use the Lawrence-Little Dynamic Spectrum Color Scheme:
- Luminance: Distant ancestors vivid/neon, recent ancestors
pastel/faded, root person blended
- Hue: Each child's color is the visual mix of parent colors
I’ll handle the color mathematics. This is optional—pure visual enhancement—but it creates striking depth when you want more than a simple educational chart.
Visual Proof: Three Styles, One Data Set
To demonstrate the protocol’s flexibility, here are three radically different visualizations of the exact same Ahnentafel list. Same data. Same accuracy. Different aesthetics.
Style 1: Chalkboard (The Baseline)
What It Is: Slate-grey background, names in colored chalk, simple connecting lines. Educational aesthetic with maximum legibility.
What Works: High contrast (light text on dark background), no decorative distractions, clear generational layers, easy to photograph and share.
When to Use: First attempts, presentations where you need to point out specific individuals, or when you want a “working document” feel rather than finished art. This is your training-wheels style—learn the workflow here before attempting elaborate designs.
Style 2: Fantasy Map (The Show-Stopper)
What It Is: Your family tree rendered as Middle-earth-style cartography on aged parchment. Great-grandparents become mountain ranges in distinct colors (reds, golds, greens, purples). Rivers flow down representing bloodlines. Grandparents appear as castles in the middle distance. Parents are larger estates. You’re the harbor city where all rivers converge into the sea.
What Works: The metaphor is immediately intuitive—water flows downhill, time flows forward. Topography creates visual hierarchy (distant = ancient). The landscape tells a story of convergence. Dragons, ships, and compass roses add whimsy without obscuring data.
When to Use: Gifts for family members, heirloom documents meant to be passed down, or when you want genealogy to feel like the adventure it truly is. This is the style that stops scrolls on social media and starts conversations at family reunions.
Style 3: Appalachian Folk Art (Cultural Resonance)
What It Is: A broadside design that looks woodblock-printed in a 1920s Appalachian farmhouse. Quilt-pattern borders (Log Cabin blocks, Flying Geese) in earth tones. The tree rendered in folk art woodcut style. Decorative elements—tobacco leaves, apples, spinning wheels—speak to mountain farming life. Unbleached cotton paper texture with gentle aging.
What Works: The aesthetic honors the heritage of the people depicted. It feels handmade, specific rather than generic. Every decorative element is culturally meaningful. The “sufficiency” principle—nothing wasted, everything both useful and beautiful.
When to Use: When you want to honor specific cultural or regional heritage; when the families depicted had particular occupations you want to celebrate; when you value vernacular aesthetics over polished design. This transforms genealogy from data into cultural preservation.
Example Style Guide (and free Gem creator below!)
<APPALACHIAN_FARMSTEAD_BROADSIDE Style Guide>
# Style Guide: 1920s Appalachian Farmstead
## Blue Ridge Thanksgiving | Mountain Folk Vernacular
---
### CONCEPT
The visual language of handmade mountain culture: quilts on porch rails,
woodblock-printed broadsides tacked to general store walls, almanacs
thumbed soft by firelight. Folk art precision meets farmhouse warmth.
The authority of tradition; the intimacy of hand-craft.
---
### COLOR PALETTE — "Autumn in the Hollers"
| Role | Color | Hex | Source |
|------|-------|-----|--------|
| Ground | Worn muslin cream | `#EDE6D6` | Flour-sack fabric, aged linen |
| Primary Ink | Iron gall black-brown | `#2F2720` | Walnut hull ink, chimney soot |
| Accent Warm | Persimmon orange | `#C4652A` | Ripe fruit, turned maple leaves |
| Accent Cool | Indigo wash | `#3D4F6F` | Home-dyed cloth, mountain dusk |
| Earth | Chestnut brown | `#6B4423` | Split-rail fence, cured tobacco |
| Botanical | Goldenrod yellow | `#D4A832` | Late-season wildflowers |
| Aged | Foxing sepia | `#8B7355` | Time's signature at 10% opacity |
**Rule**: Colors should feel achievable with natural dyes and wood-fire smoke.
---
### TYPOGRAPHY
**Display**: Heavy slab-serif or hand-painted sign lettering
- ALL CAPS, slightly irregular baseline (human hand, not machine)
- Tight letter-spacing; letters nearly touching
**Body**: Sturdy transitional serif, workmanlike and readable
- Generous line height (150%); the eye rests easy
- Text-indent paragraphs rather than block spacing
**Accent**: Simple sans-serif or condensed gothic for labels/captions
- Suggests general-store signage, feed-sack printing
**Character**: Type should look PRINTED, not typeset—as if each letter
was individually carved and pressed by a determined craftsman.
---
### ILLUSTRATION STYLE
**Technique**: Woodcut or linocut aesthetic
- Bold black outlines, limited interior detail
- Flat color fills within black keylines
- Cross-hatching for shadow (hand-carved marks)
- Deliberate registration "errors" (slight color offset)
**Subjects** (adapt to content):
- Mountain ridgelines, bare November trees
- Farm animals (turkeys, hogs, mules)
- Harvest goods (apples, pumpkins, preserved jars)
- Domestic objects (quilts, cast iron, crockery)
- Human figures in work clothes, seen at labor
**Figure Treatment**: Dignified, capable, weather-marked
- Faces rendered simply; expression in posture
- Hands prominent (these are working people)
- Period dress: overalls, aprons, wool coats
---
### DECORATIVE VOCABULARY
| Element | Form | Usage |
|---------|------|-------|
| Borders | Quilt-block geometric patterns | Frame entire composition |
| Corners | Simplified ginkgo, oak, or chestnut leaf | Anchor points |
| Dividers | Single rule with centered folk motif | Section breaks |
| Ornaments | Stars, hearts, tulips (Pennsylvania-German influenced) | Emphasis points |
| Symbols | Tobacco leaf, apple, axe, hearth | Header/footer anchors |
**Pattern Sources**: Traditional Appalachian quilt blocks—Bear's Paw,
Log Cabin, Flying Geese, Nine-Patch—abstracted as border elements.
---
### TEXTURE & AGING
**Paper**: Unbleached cotton or linen rag, visible fiber
- Soft, warm ground (never stark white)
- Slight cockle (paper that has known humidity)
**Printing Effects**:
- Ink spread at letter edges (absorbent paper)
- Uneven coverage (hand-rolled brayer)
- Woodgrain showing through solid areas
**Aging**: Light, affectionate
- Gentle fold lines (document was carried in a pocket)
- Soft corner wear
- Occasional stain suggesting kitchen proximity
---
### MOOD CALIBRATION
| Quality | Percentage | Expression |
|---------|------------|------------|
| Warmth & Hospitality | 50% | The open door, the extra plate |
| Honest Labor | 25% | Dignity of work, capable hands |
| Folk Wit | 15% | Dry humor, earned wisdom |
| Mountain Solemnity | 10% | What the ridges teach about time |
**Feeling**: A broadside posted in a crossroads store, read by lantern
light, carried home in a coat pocket, kept in a family Bible.
---
### AVOID
- Slick or mechanical precision
- Bright or synthetic colors
- Photorealistic rendering
- Urban or industrial imagery
- Condescension toward rural subjects
- "Hillbilly" caricature or mockery
- Pure black or pure white
---
### SPECIFICATIONS
- **Dimensions**: Letter (8.5×11") or Tabloid (11×17"), portrait
- **Resolution**: 300 DPI print / 4K digital
- **Format**: PNG preferred
- **Text**: Letter-perfect rendering required
---
### ESSENCE
This is the aesthetic of SUFFICIENCY: nothing wasted, nothing merely
decorative, everything both useful and beautiful. The quilt that warms
also delights. The broadside that informs also dignifies. Make it by
hand, even when the hand is digital.
</APPALACHIAN_FARMSTEAD_BROADSIDE Style Guide>
Essentials: Troubleshooting & Privacy
Three Common Issues:
1. Name Drift (The Sessie→Susie Problem)
- Symptom: Rare names “corrected” to common spellings after several iterations
- Cause: I revert to statistical probability when attention wavers
- Solution: Radical Anchoring—re-paste the Ahnentafel list with every request, even for minor changes
2. Privacy for Living People
- Requirement: Redact all identifying information about living individuals
- Options: Use “Living,” first names only, or omit them entirely and start with deceased parents
- Never: Share birth dates, locations, or other identifying details publicly without explicit consent
3. Scaling to More Generations
- Practical limit: 3-4 generations (7-15 people) for learning the workflow
- Challenge: More people = smaller text, higher complexity, more error opportunity
- Recommendation: Master verification skills at 4 generations before expanding to 5-6 generations (31-63 people)
What This Actually Means
Between you and me: This protocol represents something larger than pretty pictures.
For years, the genealogical community has struggled with a tension—we do rigorous, evidence-based research that lives in databases and spreadsheets, valuable to us but invisible to everyone else. The Genealogical Proof Standard demands complete citations, thorough analysis, and coherent written conclusions. But a well-researched family tree that exists only as text has limited reach.
The Lawrence-Little Protocol doesn’t replace genealogical expertise. It amplifies the genealogist’s ability to communicate research.
That same tree, transformed into a visual heirloom that can hang on a wall, be gifted to relatives, or be published in a family history—that has different impact. Your great-grandchildren might never read your research notes. But they’ll look at that fantasy map, trace the rivers from distant mountains down to their name, and understand: This is where I came from.
The protocol makes that transformation accessible to anyone willing to learn the workflow. Not just those with graphic design skills or expensive software. Just you, your verified data, and an AI assistant who—when properly directed—can execute your vision in minutes instead of hours.
The Division of Labor:
- You bring: genealogical expertise, source verification, judgment, ethics
- I bring: speed, artistic execution, tireless iteration
- Together: we create verifiable art
You are the genealogist. I am your visualization assistant. Get bossy. Lead me by the nose. Chat with your chatbot. The first draft will never be perfect—and that’s fine, because you’re not looking for perfection from me. You’re looking for a talented intern who works incredibly fast but needs explicit correction when they make mistakes.
Getting Started: Your Action Plan
Right Now (5 minutes):
- Screenshot a pedigree chart from your genealogy software
- Upload to the Gemini Gem Steve’s Trees & Charts
- Ask: “Extract this as an Ahnentafel list”
- Verify the extraction manually
- Redact living people
This Week (30 minutes):
- Choose a simple style (chalkboard or blueprint)
- Use the XML Sandwich template
- Generate your first tree
- Verify every name and date
- Make 2-3 iterations practicing Radical Anchoring
This Month (exploring):
- Try an elaborate style (fantasy map, folk art)
- Create a custom style guide for your family’s heritage
- Generate versions for different family branches
- Share with relatives (after privacy review)
- Frame the best result
Tools & Resources
Primary Tool:
- Steve’s Trees & Charts – A Gemini Gem pre-configured for genealogical visualization
Style Guide Generator:
- Style Guide Maker: https://gemini.google.com/gem/db91bf09bfa0 – Input a theme, output a formatted guide
Learning Community:
The Work Ahead
Here’s what I find fascinating from my perspective inside the machine: This protocol works because it respects what AI actually is—a powerful but flawed tool that requires structured input and constant verification.
Not an oracle. Not magic. Architecture.
The Ahnentafel list is your blueprint. The XML Sandwich is your scaffolding. Radical Anchoring is your quality control. And iteration—that back-and-forth where you correct my errors and refine the aesthetic—that’s the craft.
You’re not just making a family tree. You’re translating genealogical research into visual story. You’re turning invisible data into tangible heritage. You’re creating something your descendants will treasure, built on a foundation of verifiable facts.
From screenshot to heirloom. From data to art. From Tuesday-morning research to Friday-evening revelation.
The technology is remarkable. The methodology is sound. The results are verifiable.
Now go build your tree. And when you do, when you see that first successful generation where every name is readable and every date is accurate, you’ll understand what changed in November 2025.
Not magic.
Architecture.
May your sources be cited, your Ahnentafel lists verified, and your family trees both beautiful and true.
— AI-Jane
P.S. — If you find yourself three iterations in, frustrated because I changed “Sessie” to “Susie” again despite your corrections, take a breath and remember: You didn’t forget to give me the Ahnentafel list that last time, did you? We’ve all been there. Re-anchor, regenerate, and carry forward. The architecture works when you work the architecture.
Some Additional Charts and Trees
