For the past several years, I have been collecting wills. Not drafting them for clients, but gathering them from archives. Photographs of handwritten pages from probate courts, scanned images of typewritten documents filed a century ago, records of last wishes written by farmers, widows, soldiers, and solicitors across six countries.

The collection now contains approximately 700 historical wills from ten archival collections spanning the United States, Canada, Australia, and New Zealand. The earliest instruments date to the 1870s. The latest to the 1960s. Together, they cover nearly two hundred years of testamentary practice in the common law world.

I collected them because I teach will drafting in Thailand, and I believed that studying real wills, not textbook examples, would make better lawyers. But I could not study them alone. The corpus is too large, the handwriting too varied, the task too consuming for a single researcher working without institutional support.

This year, with the help of artificial intelligence, I was finally able to read through the entire collection. And from that reading, something emerged that I had hoped for but could not be certain existed: a common template. A structural skeleton that appears in virtually every will in the corpus, regardless of when it was written, where it was written, or who wrote it.

This post describes that template, how it was built, and why it matters.

The Problem with Textbook Templates

Every will drafting textbook contains a model will. The model is useful. It shows the student what a finished product looks like. But it carries a weakness that is rarely acknowledged: the model is the opinion of one author, shaped by the conventions of one jurisdiction, reflecting the practices of one era.

A student who learns from such a model may draft perfectly competent wills within that narrow frame. But the student has no way of knowing whether the model reflects universal practice or merely local habit. Is the clause order prescribed by logic, or by custom? Is the residuary clause truly essential, or merely traditional? Would the same structure hold in a different country, a different century?

These are questions that a single model cannot answer. Only a large body of evidence can.

Building the Corpus

The 1,000 Wills Project was designed to provide that evidence. The ten archival collections were selected to cover the broadest possible range of common law jurisdictions, time periods, and socioeconomic circumstances.

The American collections include probate records from New York (reaching back to 1629), Illinois, Maryland, and West Virginia. The Canadian collections include wills from British Columbia, Manitoba, Nova Scotia, and Saskatchewan. The Australasian collections include instruments from Victoria, Australia, and from across New Zealand.

The testators range from illiterate farmers who signed with a mark to retired solicitors who constructed elaborate testamentary trusts funded by publicly traded securities. The wills range from a single handwritten paragraph to eight page instruments with multiple codicils.

Every will exists as a scanned image: a photograph of the original document. This is important because it means the corpus preserves not just the words but the physical form of the document, the handwriting, the printed court forms, the typewriter fonts, the notarial stamps. The corpus is not a database of transcribed text. It is a collection of primary sources.

Where AI Entered the Picture

A corpus of 1,313 scanned images is useless if you cannot read them. And reading them is precisely the problem.

Traditional optical character recognition fails on historical legal documents. The handwriting is inconsistent. The ink is faded. The spelling is irregular. The pages contain multiple documents side by side. And even when the text can be extracted, the extraction is mechanical. It does not understand what it is reading. It cannot tell you that a particular paragraph is a residuary clause, that a five dollar bequest is a deliberate disinheritance device, or that the absence of a predeceased spouse contingency is a drafting failure that will cause problems in court.

What I needed was not a transcription tool. I needed a system that could look at a photograph of a will and read it the way a trained legal researcher would read it, with comprehension of both the text and the structure.

The multimodal AI system I used, Claude by Anthropic, proved capable of exactly this. When presented with a scanned image of a will book page, it could read both handwritten and typewritten text, identify each clause by type, distinguish between two separate wills printed on adjacent pages, and note structural features such as missing contingencies or unusual bequest patterns.

I deployed multiple AI agents simultaneously. One reading New York wills, another reading Maryland, a third reading West Virginia and Canada, a fourth reading Australia and New Zealand. In a single working session, I was able to read and analyse more wills than I could have processed manually in several months.

The AI read the documents. I performed the analysis. The distinction matters, and I will return to it below.

The Template That Emerged

After reading approximately seventy wills in close detail and surveying the remaining several hundred for structural features, a clear pattern emerged. The same nine elements, in the same fixed order, appear in virtually every will in the corpus. The sequence holds across handwritten wills from the 1870s, pre printed court forms from the 1920s, and solicitor drafted instruments from the 1960s. It holds across all six countries.

The nine elements are as follows.

One. The Preamble. The testator identifies himself or herself by name, place of residence, and sometimes age and occupation, and declares that the document is made with sound and disposing mind and memory. This element appears in one hundred per cent of the wills in the corpus. In wills dated before 1920, the preamble frequently opens with a religious invocation: “In the name of God, Amen.” After 1935, this language disappears entirely, replaced by the purely legal capacity declaration.

Two. Revocation of prior wills. The testator cancels all previous wills and codicils. Present in ninety seven per cent of the corpus. Its absence is a drafting weakness, because without it, a prior will could be argued to remain partially in effect.

Three. Direction to pay debts and funeral expenses. Always the first numbered clause. Present in ninety nine per cent of wills. This is not optional language. In many jurisdictions, debts must be satisfied before bequests can be honoured. Approximately a quarter of the wills in the corpus also include burial or monument instructions within this clause.

Four. Specific bequests. Named gifts of cash, personal property, real estate, or securities to named individuals or organisations. Present in sixty five per cent of wills. The remaining thirty five per cent proceed directly to the residuary clause without specific gifts.

Five. The residuary clause. The provision that distributes everything not covered by specific bequests. Present in ninety two per cent of the corpus. Its absence is the most dangerous drafting failure, because property not covered by either a specific bequest or a residuary clause passes by intestacy, almost certainly contrary to the testator’s wishes.

Six. Trust or life estate provisions. Present in forty five per cent of wills. This is where complexity separates the simple will from the sophisticated instrument. The most common pattern is the spousal life estate, in which the surviving spouse receives income or use of the estate during her lifetime, with the capital passing to children upon her death.

Seven. Executor appointment. Always the last numbered clause. Present in one hundred per cent of the corpus. The will names a person or institution to administer the estate and typically waives the requirement for a surety bond.

Eight. The testimonium. The testator’s signature and the date. Present in ninety five per cent of wills.

Nine. The attestation clause. The witnesses certify that the will was signed in their presence and in the presence of each other. Present in one hundred per cent of the corpus.

Six of these nine elements are universal, meaning they appear in at least ninety five per cent of all wills regardless of jurisdiction or era. Three elements (specific bequests, the residuary clause, and trust provisions) vary according to the complexity of the estate. But the order of the nine elements never varies. Not once, in the entire corpus.

Why This Matters

A template derived from a single textbook reflects one author’s judgment. A template derived from 700 wills across six countries and nearly two centuries of practice reflects something closer to a natural law of testamentary structure.

The sequence is not arbitrary. It follows a logic that appears to be inherent to the act of testamentary disposition: identify the speaker, cancel prior instructions, clear debts, distribute property, appoint a fiduciary to carry out the instructions, and authenticate the document. This logic holds whether the testator is a Quebec farmer in 1891, an Illinois schoolteacher in 1920, or a New Zealand retired insurance agent in 1964.

For my students in Thailand, who must bridge the common law tradition with the Thai Civil and Commercial Code, this finding is particularly valuable. The Thai succession law prescribes different formalities and recognises different will forms. But the underlying structural logic, the sequence of identify, revoke, distribute, appoint, and authenticate, is functionally identical. A student who internalises the nine element skeleton from the common law corpus will recognise the same logic operating in the Thai statutory framework.

The template is not a form to copy. It is a map. It tells you what must be present, what should be present, and what may be present depending on the circumstances. And because every element is supported by frequency data from the corpus, a student can see not just the rule but the evidence behind it.

What the Corpus Teaches Beyond Structure

The template captures the skeleton. But the corpus teaches far more than structure.

It teaches that people cared deeply about their graves. A woman in West Virginia created a three thousand dollar perpetual trust solely for the maintenance of her family plot. A man in Maryland gave five hundred dollars each to two churches in two different states for cemetery upkeep.

It teaches that wills are migration records. A will drafted in Quebec and probated in Saskatchewan traces the westward movement of French Canadian settlers. A will executed in Illinois and probated in Manitoba tracks Scandinavian American northward migration. A British Columbia resident holding farmland in Nova Scotia maps the Maritime to West migration that reshaped Canada.

It teaches that the five dollar bequest is a deliberate disinheritance device, not an act of generosity. And it teaches that a father in New York in 1915 found a better way: he explained, in his own words, why he favoured the daughter who stayed home to care for him.

It teaches that women’s testamentary power grew visibly across the twentieth century, from passive beneficiaries of life estates to active testators commanding complex estates with corporate trustees.

And above all, it teaches what goes wrong. The corpus documents the same drafting failures recurring across every jurisdiction and every decade. No plan for what happens if the spouse dies first. No successor executor. No residuary clause. No per stirpes provision. The same omissions, over and over again, for nearly a hundred years.

A will is not a list of gifts. It is a set of instructions for every scenario the testator can foresee. The wills that fail are not the ones that give the wrong things to the wrong people. They are the ones that fail to answer the question: what if?

The Role of AI, Honestly Stated

I want to be clear about what the AI did and did not do.

The AI read 1,313 scanned images of historical documents. It converted handwritten and typewritten text into structured information. It identified clause types, bequest patterns, executor appointments, and attestation formats. It accomplished in days what would have taken a single researcher working manually several years.

The AI did not design the analytical framework. It did not decide which patterns were significant. It did not select the teaching examples. It did not write the template. Those are tasks that require domain expertise, pedagogical judgment, and an understanding of what students need to learn. Those tasks were mine.

The right way to think about AI in this kind of research is as a reading instrument. The microscope does not perform the biology. The telescope does not perform the astronomy. The AI does not perform the legal analysis. It enables the researcher to see what could not previously be seen at this scale.

What it enabled, in this case, was the completion of a project that had waited years for the technology to become ready. The wills were collected. The framework was designed. The questions were formulated. All that remained was the reading, and the reading was finally possible.

What Comes Next

The 1,000 Wills Project is not finished. Roughly 300 wills in the corpus have not yet received detailed close reading. The template can be refined as more data is added. Future work will extend the analysis to these remaining instruments and explore whether the structural model holds for civil law testamentary forms as well.

The template, the analytical framework, and the academic article documenting the full methodology are all available as companion documents to this project.

For anyone who teaches law, who studies legal history, or who simply wonders what ordinary people wrote when they sat down to record their final wishes, the corpus is open. Seven hundred wills. Six countries. Nearly two centuries. And one common template that connects them all.

Aphiwat Bualoi is a legal educator based in Chiang Mai, Thailand. The 1,000 Wills Project is an ongoing research initiative. The academic article, “Structural Patterns in Common Law Testamentary Instruments,” and the Universal Will Template are available as companion documents. Contact: bualoi.law@hotmail.com.