Dutch marriage annulled after ChatGPT-written vows miss required legal phrasing

A court in Zwolle, the Netherlands, has ruled that a couple’s marriage was not legally formalised after their wedding vows — written with the help of an artificial intelligence tool — failed to meet the legal requirements under Dutch law. The decision, handed down in early January 2026, highlights how automated text generation may clash with established legal standards.

The couple, who held a civil ceremony in April 2025, asked a friend acting as a one-day registrar to conduct the service. That friend used ChatGPT to draft the wedding speech and vows in a more personal tone. However, the resulting text did not include the legally mandated declarations that couples must make under Article 1:67 of the Dutch Civil Code — specifically, an explicit commitment to fulfil the legal obligations of marriage.

During the ceremony, the vows quoted by the court included informal language such as promises to “laugh together, grow together, and love each other no matter what,” and descriptions of the couple as “a crazy couple.” The court found that the statements did not satisfy the statutory requirements for a valid marriage declaration and therefore concluded that the marriage certificate in the civil registry was recorded in error.

The couple later petitioned the court to recognise the original date of their wedding as their legal marriage date, arguing that they had not intended to make an error and that the civil officer present had not corrected the proceedings. The court refused, saying it could not overlook the legal requirements even if the date held personal significance.

The case underscores a broader tension between generative AI and formal legal processes. As AI tools like ChatGPT are increasingly used to draft speeches, agreements and other text, legal experts have noted that auto-generated language may omit precise, legally necessary terminology — even if such text feels appropriate in informal settings. Cases involving contracts, regulatory filings and terms of service have previously highlighted similar risks when AI assistance is involved.