AI Wills and Knowledge and Approval: Why Convenience May Come at a Cost

The rise of AI in will drafting

Artificial intelligence is rapidly embedding itself in everyday decision-making, from drafting emails to generating contracts. It is therefore no surprise that it is increasingly being used to draft wills. For a generation accustomed to instant, low-cost digital solutions, the appeal is obvious: why pay a solicitor when an AI platform can produce a will in minutes?

Research by the Association of Lifetime Lawyers, conducted by Censuswide, highlights the scale of this shift: 72% of UK adults aged 30 to 34 said they would consider using AI to draft or update a will. The attraction is easy to understand: convenience, accessibility and lower upfront cost.

However, while AI can produce a document that resembles a will, it cannot replicate the legal safeguards required to ensure that the will is valid and effective. That creates significant risk, particularly when viewed through the lens of one of the most fundamental principles of English succession law: knowledge and approval.

What is “knowledge and approval” in English law?

For a will to be valid under English law, it must satisfy both formal and substantive requirements. While the Wills Act 1837 governs execution formalities, such as signature and attestation, the doctrine of knowledge and approval requires something more: the testator must genuinely understand and approve the contents of the will they are signing and know what its effect would be.

This is distinct from testamentary capacity, which concerns whether a person is mentally capable of making a will. A testator may have full capacity yet still be unable to demonstrate knowledge and approval if they do not fully understand what the document says or the effect it will have.

In most cases where a solicitor drafts the will and explains its provisions, the court will usually infer knowledge and approval in the absence of other evidence. However, that presumption is far weaker, or may not apply at all, where the will is prepared in unusual circumstances. As we will explore, AI generated wills are more likely to fall within this second category, increasing their vulnerability to challenge.

The illusion of clarity in AI-generated content

One of the defining characteristics of AI generated text is its fluency. Outputs are often well structured and confidently expressed, using technical language to create an impression of authority and accuracy. However, this can mask deeper issues.

Terms such as “life interest”, “residuary estate” and “per stirpes” can have precise meanings that may not be intuitive. Without professional guidance, testators may misunderstand how these provisions operate in practice and how they affect the distribution of their estate.

Furthermore, AI systems have repeatedly been shown to generate inaccurate legal content, including invented authorities, flawed reasoning and clauses that appear sophisticated but are inconsistent or ineffective in practice. A clause may, for example, purport to achieve a tax-efficient result or create a trust structure that sounds familiar, but which is unsuitable for the testator’s circumstances or legally ineffective as drafted. A layperson may have no reason to question the wording and is unlikely to interrogate each clause critically, particularly when the language appears polished.

The result is a risk of only superficial understanding. A testator may feel confident in the document because it reads well, rather than because they have engaged with its legal effect. In a dispute, the court may conclude that the testator signed a document containing provisions they did not truly comprehend, thereby failing the knowledge and approval requirement.

Complex family structures and incomplete instructions

Modern estates are rarely straightforward. Blended families, cohabitation, international assets, business interests and tax considerations introduce layers of complexity that require careful planning.

AI systems are entirely dependent on the prompts they receive. If a user does not recognise the legal significance of certain facts, those facts may never be included. Unlike a solicitor, an AI system will not probe for clarification, identify missing information, or test whether a proposed arrangement truly reflects the testator’s intentions. The resulting will may therefore address only a partial picture of the testator’s circumstances.

A user may also miss valuable tax-planning opportunities, such as the residence nil-rate band, agricultural property relief and business property relief. If these issues are not raised in the prompt, they may be ignored altogether. Even where they are mentioned, they may be applied incorrectly, or without the user properly understanding the consequences.

These limitations create a significant evidential risk. If a will fails to reflect the testator’s true intentions because relevant factors were overlooked, or tax planning was applied incorrectly, challengers may argue that the testator did not understand or approve the consequences of the document they executed.

Updating a solicitor-drafted will with AI

A more realistic use of AI may be to alter an existing solicitor-drafted will either via amending the document itself or by drafting a separate codicil to effect the change requested by the testator. 

Where a codicil is drafted, a myriad of issues could arise. The AI may not, for example, inform the testator that executing the codicil re-publishes the will, or it may use definitions inconsistent with the will adding confusion or affecting the dispositions being made. To illustrate, a will which leaves a gift to a spouse by name, who is subsequently divorced (thus voiding the gift), may cause the gift to take effect once more when the codicil is executed. If the AI is not probed on the implications of the codicil, the testator may have no knowledge as to this effect.

At first glance, it may therefore seem less risky to amend the will itself because the overall structure is already in place and the testator may only want to make a small change. In practice, however, even a limited amendment can have significant legal and tax consequences.

Take, for example, a testator who changes their will so that a long-held family home passes to their brother rather than to their children. An AI platform may make that amendment without hesitation or explanation as to any consequences, and without probing further. Yet the change would mean that the residence nil rate band is no longer available which could in turn increase the inheritance tax burden and, if the property is the main asset in the estate, may mean that it has to be sold to pay the tax. The amendment may reflect the testator’s immediate instruction, but what if their intention was for the brother to manage the property after their death but for the benefit to pass to their children, in a form of trust structure? Failing to appreciate the financial and legal consequences of this change raises suspicion as to whether the testator understood and approved its true effect.

Suspicious circumstances and increased scrutiny

If a will is disputed, certain features are known to excite the suspicion of the court, including wills drafted without solicitor involvement, unexpected dispositions and significant departures from previous testamentary arrangements. AI-generated wills are more likely to exhibit these characteristics.

For example, an AI tool may produce a generic structure that inadvertently excludes close family members or creates an outcome that appears irrational or inconsistent with the testator’s known intentions. Where this suspicion is aroused, the propounder of the will has the burden of proving, on the balance of probabilities, that the testator satisfied the requirement for knowledge and approval. In the absence of professional involvement, they may struggle to produce sufficient evidence.

Lack of professional oversight and evidential support

A critical advantage of professionally drafted wills lies not only in their quality, but also in the evidential framework that surrounds them. Solicitors maintain detailed attendance notes, record instructions and ensure that clients understand the implications of their decisions before signing.

These records play a vital role in defending the validity of a will if it is later challenged. They provide contemporaneous evidence that the testator knew and approved the contents.

AI-generated wills lack that protective framework. There is rarely a reliable record of what the testator understood, what explanations, if any, were given, or why particular decisions were made. Questions may also be raised over whether a beneficiary helped prompt the creation of the document. In contentious probate proceedings, that absence of contemporaneous evidence can be decisive.

Implications for private client lawyers

The increasing use of AI in will drafting presents practical challenges for private client practitioners. During a client’s lifetime, lawyers may be asked to review AI-generated documents that require substantial reworking to correct legal defects, clarify intentions and repair structural problems.

After death, contentious probate teams are likely to see a rise in disputes involving AI drafted wills. Lack of knowledge and approval will be a natural ground of challenge, often combined with issues of construction, rectification or testamentary intention.

This environment reinforces the importance of professional advice. The value of a solicitor lies not only in drafting a legally sound document, but in guiding the client through a process that demonstrates understanding and informed decision making.

A false economy in estate planning

While AI offers speed and apparent cost savings, those benefits may prove illusory in the context of wills. The true test of a will arises only after death, often in contentious and emotionally charged circumstances, when the testator is no longer available to explain their intentions.

Where an AI generated will fails to meet the threshold for knowledge and approval, the consequences can be significant. Disputes may lead to delays, increased costs and, ultimately, outcomes that do not reflect the testator’s intentions.

For now, the law remains firmly rooted in human understanding and intention. Until AI can replicate not just the form of a will, but also the process of ensuring genuine comprehension, its role in this area will remain inherently limited.