Audio recordings as evidence: what they are actually worth

Published on 18 August 2026 · Author: Francesco Zinghinì

There is a recorder in every pocket, and with it two opposite misconceptions: that recording is always forbidden, and that an audio file wins the case. Both are wrong. Three questions get tangled here and are best kept apart: whether recording was lawful, what the file is worth, and how to make it hold when someone challenges it.

When recording is lawful

The governing criterion is participation. Under settled Italian case law, someone who takes part in a conversation may record it, even covertly: documenting what was said to you is not treated as an intrusion into anyone’s protected sphere. The picture changes entirely when the person recording is a stranger to the conversation: that is an interception, which outside judicially authorised cases is a criminal offence and yields no usable evidence. A further limit protects the home: capturing what happens in another person’s private premises, outside your own presence, is punishable regardless of the intended use.

Keep the courtroom and the public square distinct. Filing a recording in proceedings to assert a right is generally admitted — data-protection rules acknowledge that purpose too; posting the same file on social media or passing it to third parties is different conduct with different consequences. Lawfulness is not a property of the file: it depends each time on who recorded, where, and for what use.

What it is worth in court

In Italian civil proceedings a recording is a mechanical reproduction under article 2712 of the Civil Code: full proof of the facts it represents unless the party against whom it is produced disavows its conformity. It is the same mechanism I described for WhatsApp screenshots, with the same practical consequence: no automatic value — everything turns on the challenge. Once disavowed, the recording does not vanish from the case; it becomes an element the judge weighs freely, and the party who produced it must support it with something else: consistency with documents and testimony, and above all technical verification.

In criminal proceedings a participant’s recording enters as documentary evidence. No magic there either: the defence may dispute its genuineness, demand the original medium, or contest the identity of the speakers.

How recordings are challenged and examined

A serious challenge does not stop at “the recording is disavowed”. It states what does not add up — and on that ground a technical examination has concrete instruments:

• The container. Every recording application produces files with characteristic structure and metadata. A file presented as “recorded on the phone” that has the structure of an exported editing project tells a different story from the one declared.
• The joins. Cuts and splices leave discontinuities — in the background noise, in the room’s reverberation, in the signal’s spectrum, sometimes in the file’s internal block numbering. Not everyone who edits is careful; almost no one is careful enough.
• Re-encodings. An audio file that has passed through a messaging platform has been recompressed: normal, but it means the file produced is not the original and the metadata trail is gone. “Where is the original file?” is to audio what “where is the device?” is to chats.
• External consistency. Times, places and sounds compatible with the declared context; duration consistent with the facts; no gaps in the recorder’s file numbering.

The problem of attributing the voice

“That is not me” is the most frequent defence against a recording. Speaker identification is a discipline of its own — comparative phonic analysis — with its own methods, its own limits, and a requirement that is often missing: comparison samples of adequate quality. It should be said plainly: the outcome is probabilistic, and sensitive to audio quality, noise and the amount of usable speech. A serious expert will tell you when the material makes the examination meaningful — before charging you for it.

Attribution is often sustained contextually as well: the calling number, references to facts only that person could know, consistency with written messages. The voice is one tile, not the whole mosaic.

Preserving a recording that holds

Forensic acquisition: how I work → · Back to insights