When the authenticity of a signature is questioned, its appearance alone does not establish whether it is genuine. Computer-based systems can compare features in handwritten signatures, but the significance of those results depends on the system used, the available samples, and the conditions under which the method has been tested.
An automated comparison may assist an examination. It does not, by itself, establish who executed a signature or whether a questioned signature is a forgery.
What Is AI Handwriting Analysis?
AI handwriting analysis encompasses several different technologies. Handwriting-recognition systems identify written characters and convert them to text. Other systems compare features of handwriting or signatures to assess whether samples are consistent with one another.
These are different tasks. Recognizing the words in a document does not identify the writer. Similarly, finding measurable similarities between signatures is not equivalent to establishing authorship. The question in forensic document examination is what the observed characteristics, considered with the available comparison material, can support.
Can AI Detect Forged Signatures?
Automated signature verification has been studied for many years, including approaches that use machine learning to compare signature images. A study by Diaz, Ferrer, and Vessio, published in Neural Computing and Applications, evaluated an explainable automatic signature-verification method intended to support forensic handwriting examiners. The researchers tested the method using signature datasets and described its performance relative to other approaches. [3]
Such studies establish that automated signature comparison is technically possible. They do not establish that every AI application is validated for forensic casework or that a result obtained from one type of sample will apply reliably to another.
The appropriate question is not whether a program can issue a verdict. It is whether the method has been adequately tested for the particular purpose, what information it uses, and what limitations are associated with its results. A software-generated score should not be reported as conclusive proof of genuineness or forgery.
What Does a Forensic Document Examiner Evaluate?
Forensic handwriting examination considers both the similarities and the differences between questioned and known writing. It also accounts for natural variation: genuine signatures written by the same person will not be identical every time. NIST discusses the importance of multiple, suitable known samples when evaluating that variation. [1] Depending on the material available, an examination may consider:
- Letter formation and construction
- Proportions, spacing, alignment, and other writing habits
- Line quality, stroke characteristics, and indications of writing execution
- Pen-pressure patterns and their location within the writing, when these characteristics can be examined
- The range of natural variation in known handwriting
- The significance of both similarities and differences in the comparison material
No single characteristic necessarily establishes authorship. Observations must be evaluated collectively, in light of the quantity and quality of the writing available.
Why Pen Pressure Matters
Pen pressure is not simply how dark a line looks. Where pressure increases or decreases through the course of a signature, and how those changes relate to the movements of writing, may be relevant to a forensic comparison. An examiner considers these characteristics in relation to the signature as a whole and to suitable known writing. Research has examined pen pressure alongside other writing-movement characteristics. [7]
Although scans and photographs may show some visible details related to pen pressure, they cannot reproduce all the physical characteristics available for examination in an original document. Whether an image-based AI system can accurately recognize and interpret those pressure-related characteristics depends on the system and its demonstrated capabilities. Some specialized digital signing systems record pen-force data at the time of writing; that is different evidence from a scanned paper signature, and it still requires appropriate evaluation. [4, 7]
The Importance of Original Documents and Known Samples
An original document may preserve physical details that are reduced or absent in a scan or photograph. Depending on the examination, such information can be relevant to how the writing was executed. Forensic document examination standards and procedures emphasize the importance of examining original documents whenever possible, as copies can limit the physical characteristics available for examination. [4]
Known writing is equally important. Comparison samples should be suitable for the type of writing in question and, where possible, reasonably close in date. Multiple genuine samples help demonstrate the writer’s range of natural variation. [1, 2]
If the available material is insufficient, a forensic examiner may be unable to reach a conclusion about authorship. That limitation must be considered rather than disregarded.
What if a Person’s Signature Changed During Illness?
A questioned signature may have been written during serious illness, after an injury, or under other unusual physical circumstances. Research has documented changes in handwriting associated with certain neurological conditions and medication. Those changes can complicate comparison with earlier samples. [5]
Consider a signature on a will executed near the end of a person’s life. There may be no later signatures available, and earlier examples may not fully represent the writer’s condition at the time the questioned document was signed. A difference from older signatures does not, by itself, establish forgery. An image-based AI comparison cannot determine the cause of a change in writing merely from that difference. The circumstances and the available comparison material affect what an examination can support.
Can AI Replace a Forensic Document Examiner?
ANSI/ASB Standard 070, Standard for Examination of Handwritten Items, sets out procedures for the examination and comparison of handwritten material, including an evaluation of the sufficiency of questioned and known samples. It explains that the methods used depend on the nature of the available material and that written procedures do not replace education, training, skills, and experience. [2]
Computer-assisted methods may have a role in signature analysis. Their output still must be interpreted with regard to the method’s validation, the quality of the evidence, and the questions the examination is intended to answer.
This distinction matters when a general-purpose AI application is asked to compare two uploaded images and declare a signature genuine or forged. A fluent explanation or apparently precise confidence score does not establish that the application performed a validated forensic examination.
What About AI Signature Results in Court?
An AI-generated assessment is not a substitute for testimony from a qualified forensic document examiner. In federal proceedings, Rule 702 requires expert testimony to meet requirements concerning qualifications, sufficient facts or data, and reliable methods applied to the case. Whether particular evidence or testimony is admissible is determined under the applicable rules and by the court. [6]
A forensic document examiner can explain the examination performed, the comparison material considered, and the basis for an opinion. A general-purpose AI application that labels a signature “genuine” or “forged” does not, simply by producing that label, provide qualified expert testimony.
When Is a Professional Signature Examination Appropriate?
A professional examination may be appropriate when a signature or handwritten document is disputed, including signatures appearing on wills, contracts, deeds, checks, financial records, and other legal or business documents.
A forensic document examiner can review the nature of the questioned material, identify the type of known writing needed for comparison, and explain what examination is possible with the available evidence.
Findings may be documented in a report or explained by the examiner when expert testimony is requested and permitted.
What an Automated Result Can Establish
AI-based signature comparison is an active area of research, but a system’s output is only as meaningful as the evidence and validation supporting it. Neither an automated match nor an automated mismatch should be treated as proof of authorship or forgery without an appropriate assessment of the method and the questioned material.
In a disputed-signature case, the central issue is what the document evidence can support. That assessment requires attention to the nature of the writing, the quality of the available samples, and the limits of the examination.
Request a Handwriting or Signature Examination
Kathy Carlson, CFDE, CQDE, is a certified and court-qualified forensic document examiner who provides handwriting and signature examination services for attorneys, businesses, and individuals. If you are seeking assistance with a questioned signature or handwritten document, contact Western Forensic Document Examiner to discuss the materials available for review.
- Learn more about handwriting analysis and signature analysis and comparisons.
- Contact Kathy to request a consultation
