About a year or two ago, I asked ChatGPT to convert a metes and bounds land description into a land plat drawing. It was not able to do so correctly. Yetserday, I tried again and the results were astounding!
I compared three scenarios:
- ChatGPT 5.2 Auto with a short, simple prompt
- ChatGPT 5.2 Auto with a longer meta and chain-of-thought prompt
- ChatGPT 5.2 Thinking with a short, simple prompt
I used a deed I had already platted with DeedMapper, a year or two ago. This helped me compare the AI plat drawing to a correct version. Here’s the deed transcription:
Arnold-West Deed 25 Nov 1788
Fauquier County, Virginia, Deeds 10:144-145, Humphrey and Ann Arnold to John West, 25 Nov 1788, recorded 22 Jun 1789; image online, FamilySearch (https://www.familysearch.org/ark:/61903/3:1:3Q9M-CS4C-M998-K : accessed 20 Apr 2022), digital film 007894522, image 104 of 591.
This indenture made this the 25th day of November in the year of our lord one thousand seven hundred and eighty eight Between Humphrey Arnold of the County of Fauquier and State of Virginia and Ann his wife of the one part and John West of the County and state aforesaid of the other part. Witnesseth that for and in consideration of the sum of sixty pounds current money of Virginia in hand paid by the said John West at or before the sealing and delivering of these presents the receipt whereof they doth hereby acknowledge and thereof acquit and discharged the said John West his heirs exctrs and admors and by the presents the said Humphrey Arnold and Ann his wife have granted, bargained, sold, aliened released and confirmed and by these presents do grant bargain, sell, alien, release and confirm unto the said John West his heirs and assigns forever all that tract or parcel of land situate lying and being in the said County of Fauquier and parish of Hamilton and bounded as followeth, viz Beginning at a white oak in the line of Richard H. Lees land running thence so 45 E 59 poles to a large poplar standing on the east branch of Licking run corner to Samuel Porter thence down said branch with said Porter So 6 E 7 poles to an Ash thence with the said Porter No 66 E 106 poles to a stake on a stony nole in an old field thence no 33 W 86 poles to an hole oak in a thickett then west 94 poles to a stake in the aforesaid Lees line thence with the said Lees line to the beginning containing seventy acres to the same more or less it being apart of a larger tract of land formerly by the property of Thomas Bunbery and Elizbaeth Bunberry with all houses holdings orchards woods underwoods, trees, ways, water and water courses, …
Signed, sealed and delivered Humphrey Arnold LS
In presence of Ann her X mark Arnold LS
William Russell Wm Bryan
Henry Jones Seemore Arnold
Alex Brown John Knowling
William Bryan John Cooke
At a court held for Fauquier County the 22nd day of June 1789
This Indenture was acknowledged by the said Humprhey Arnold to be his act and deed and ordered to be recorded. Teste H. Brooke CC
DeedMapper Plat Drawing
Here’s the land plat drawing I made by entering the calls into DeedMapper 4.2:
ChatGPT 5.2 Auto with Short Prompt
To see if AI could map the deed correctly, I pasted the entire transcription of the deed into ChatGPT 5.2 Auto. This model uses the prompt to decide of the user needs a fast vs. a thinking model, and other features. It chose the fast model, and quickly used its image generator to draw this:
The drawing is incorrect and doesn’t match the correct land plat drawing I made in DeedMapper. It apparently used its image generator to create something like a land plat drawing mixed with art. A more detailed and thorough prompt would probably have helped.
ChatGPT 5.2 Auto With a Longer Prompt
Knowing that this task was challenging, I decided to try a couple prompt strategies to get better results.
Chain-of-Thought Prompting
The first is chain-of-thought prompting, where the user gives the chatbot questions and prompts leading up to the challenging task to get it to “think” about what it needs to know before doing the task.
Meta Prompting
The other strategy is called meta prompting, where a user asks the chatbot to write a detailed prompt describing what needs to be done for the challenging task. Then the user gives that prompt to the AI chatbot and asks it to do the task.
Results
Here are the series of prompts I gave to ChatGPT 5.2 Auto to incorporate both of these strategies:
- What information would a genealogist and expert in land records in Virginia need to know about metes and bounds and land platting from a description in a deed to be able to draw a land plat correctly?
- Here’s a deed – extract what you would need to draw the land plat. [pasted entire deed transcriptoin]
- Write a detailed prompt for how to draw this land plat.
- Now write the prompt for an AI to draw the land plat.
- Now follow this prompt and generate an image of the land plat drawing.
The resulting image was still incorrect, though it did look much closer:
Results with Python Script
Throughout the conversation, ChatGPT suggested different things it could do with the data it extracted from the deed. One of them was to “Write a Python plotting script.” Since generating an image wasn’t producing a technically correct plan drawing, I thought the Python code would work better. My next prompts included:
- Write a Python plotting script
- Run the script and create the drawing
- Add full beraing labels
The resulting land plat drawing was an accurate representation of the land description and matched my DeedMapper drawing! Here’s the image file I exported from the conversation:
ChatGPT 5.2 Thinking
After this success, I wondered if the thinking model would be able to do this chain-of-thought thinking process itself, without my guidance. After all, I barely did anything in the chain-of-thought – simply directed it to think about drawing a deed before it drew the deed.
I gave the same prompt as I did before to GPT 5.2 Auto: Draw a land plat drawing of this deed: [pasted deed transcription].
The results were very good!
Takeaways
I’m no expert in deed mapping. I’ve dabbled in DeedMapper and I’ve transcribed many deeds from the southern state land states. It would be challenging to know if AI gets the plat drawing right or not if you don’t have a lot of experience. My suggestion is to learn to use DeedMapper, take a class on mapping deeds, then see how AI can be your assistant with this task.
For those who ahve tried it, what have your experiences been with AI deed mapping?
Leave a Reply
Thanks for the note!