The first contract dispute I saw up close taught me an odd lesson about signatures. Two parties, one agreement, and a disagreement about payment. Both sides had a signed PDF. Nobody spent a single minute arguing about the squiggle. The entire fight was about who sent the document, when, and whether the file had changed since.
That is the part most people miss about electronic signatures. The ink, or its pixel equivalent, is the easy part. Under the US ESIGN Act, in force since 2000, a contract cannot be denied legal effect just because it was signed electronically. The EU's eIDAS regulation says the same about evidence: a signature cannot be rejected just for being electronic. Admissibility is settled. What actually gets argued in a dispute is the evidence around the mark, and if you build software that captures agreement, that evidence is your job.
What the law actually asks for
Strip away the jurisdictions and the statutes converge on the same handful of questions.
Intent. Did the person mean to sign? A name typed onto a final document after a review step shows intent. A checkbox pre-ticked by default is the kind of thing opposing counsel enjoys finding.
Consent. Did both parties agree to do business electronically? A one-line email is usually enough, and most apps capture this in the terms nobody reads.
Attribution. Can you connect the act to a person? An email thread from the signer's own account, an IP log, an authenticated session.
Integrity. Is the file presented in court the same file that was signed? A PDF that has not been modified since signing beats one that keeps getting edited.
None of these require cryptography. A drawn signature image with a good trail outperforms a cryptographic signature with no trail, the same way a photocopy of a contract you can find beats a flawless original you cannot.
The audit trail as an engineering spec
Translate those four questions into records and you get a short list of things to capture at the moment of signing, and to keep:
{
"document_sha256": "9f2c41...",
"terms_sha256": "77ab09...",
"signer_email": "dana@clientco.com",
"auth_method": "email_link",
"signed_at_utc": "2026-09-27T14:03:11Z",
"ip": "203.0.113.24",
"user_agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64)",
"rendered_for_signer": ["contract_v3.pdf", "terms_2026-08.pdf"]
}
Two hashes, not one. The document hash proves the PDF did not change after signing. The terms hash proves which version of the surrounding agreement was displayed. Version drift between what the user saw and what you claim they agreed to is a real failure mode, and it is cheap to prevent at signing time and miserable to reconstruct from server logs after the fact.
The auth method matters more than people expect. A session opened through an email link is attribution. A shared account is not, and shared accounts are where attribution goes to die: an office inbox anyone can open, a family tablet, a contractor holding the client's password. Courts know this. If your product lets five people share one login, your signature records say "somebody at that company," which is a weaker sentence than it sounds.
The failure modes I keep seeing
The common workflows break in predictable ways.
The email-back workflow. Client prints, signs, scans, emails the scan back. The signature is fine. The evidence is whatever survives in the thread, which after three forwards and one screenshot pasted into a slide deck is not much. If you are on the receiving end, save the original attachment with full headers and never re-export it.
The post-sign edit. Someone fixes a typo in the signed PDF. The edit is innocent, the file hash is not, and integrity is now a question you have to answer instead of one you get to ask. Fix typos before signing, or sign again after.
The vanishing platform. A signing service shuts down, or you lose the account, and the audit trail lives in their backend. Export the signing certificate or audit record at signing time and store it next to the document.
Where paper still wins
A short list of documents sits outside the e-signature default in most jurisdictions: wills, some family law filings, certain deed recordings and notarized documents. Remote online notarization is expanding the boundary, but for a will, use paper unless you have checked your local rule. For contracts, NDAs, quotes, and order forms, the electronic route holds up. This is general engineering information, not legal advice; a contested contract needs a lawyer in your jurisdiction.
The part worth getting right
Most of this is not legal engineering. It is ordinary record keeping, done deliberately, at the moment when it is cheap. Hash the file, note the time, tie the act to an identity, and never touch the artifact afterward.
For the fuller version of what courts examine and which mistakes weaken a signature in practice, this guide to how e-signatures hold up in court walks through attribution, integrity, and the scenarios where signatures get challenged. And when the trail is taken care of and all you need is a clean, consistent signature image for the final document, there is a free AI signature generator that works entirely in the browser.
The signature is the ceremony. The trail is the contract.
Top comments (0)