SMS as evidence: what survives, where, and for how long
Text drafted with the help of artificial intelligence tools. How I work
The text message passes for the easy kind of evidence: old, simple, sitting in the handset. In practice it is close to the opposite. Of all the conversations that reach a courtroom it is the one with the least reliable sender, the narrowest recovery window, and the largest number of misconceptions about what can be asked of the carrier. This article sets out what survives of an SMS, where it survives and for how long — three different questions that are almost always collapsed into one.
Why an SMS is not a chat
The similarities run out quickly. A conversation in a messaging application is tied to an account, lives on infrastructure that synchronises it across the devices of that account, and keeps internal identifiers for each message somewhere. An SMS has none of this: it is tied to a number, travels over the carrier's network, is delivered to a handset and stops there.
Three practical consequences follow, and they are worth stating at the outset because they shape everything else:
- There is no «delete for everyone». Deleting an SMS deletes it from your own handset. The other participant's copy stays where it is, and in practice it is the most underrated route to recovery.
- There is no account to re-download the history from. When the handset changes, whatever was not transferred is gone: there is no archive in a provider's cloud to go back to, because there is no provider.
- There is no end-to-end encryption of the content. The message travels in clear through the carrier's systems. That tells against privacy and in favour of nothing in evidential terms — and it must not be turned into the idea that the text is therefore recoverable from the carrier, which is a separate question answered further down.
The sender is not authenticated
This is the point that changes the weight of a text message in court, and the one I meet least often in pleadings. In the SMS protocol the sender field is not verified. Professional bulk messaging services — the ones a company uses to send a login code or a delivery notice — allow the sender to be set to an alphanumeric string of the customer's choosing. This is not an abuse: it is a feature, and its purpose is to show the company's name instead of a number.
The consequence is that anyone with access to such a service can land a message on a handset bearing whatever name they prefer. And here comes the detail that does the real damage: handsets group messages by sender. A message whose sender is set to a bank's name does not arrive in isolation — it is filed in the same conversation as that bank's genuine messages, below them, in the same thread, under the same heading. Anyone looking at it sees a coherent history that assembled itself.
Hence a rule of reading that applies to any SMS produced in proceedings: the presence of a message inside a conversation attributed to someone is not an element of attribution. It is precisely the appearance that the technology allows to be manufactured at almost no cost. Attribution, when it is contested, has to be sought elsewhere: in traffic data, in the putative sender's device, in the surrounding circumstances.
The converse matters just as much, lest a technical argument turn into a skeleton key: that a sender can be forged does not mean that it was. Forgery leaves traces — in the traffic data the message still has an origin, and a bulk send has characteristics of its own that a message from a handset does not have. The useful argument is conducted on those traces, not on the abstract possibility.
Not everything in «messages» is an SMS
Inside the messaging application of a modern handset several protocols coexist and look alike on screen. Telling them apart is the first step of an examination, and skipping it leads to asking the carrier for things it does not hold and to overlooking copies that exist.
- Traditional SMS. The subject of the rest of this article: no account, no end-to-end encryption, traffic data held by the carrier.
- RCS, the successor to SMS between handsets that support it. It travels over data rather than the telephony network, passes through the infrastructure of whoever provides the service and, in recent configurations, is encrypted end to end. The telephone carrier may hold no traffic record of it at all, while the conversation may have a copy where an SMS would not.
- iMessage, between Apple devices, in the same thread and wearing the same face. It is a service tied to an account, synchronised across the user's devices, potentially present in a backup: everything an SMS is not. On screen, the difference is the colour of the bubble.
A screenshot, of course, preserves none of these distinctions — one more reason not to stop there, on top of those I have set out about screenshots of a conversation. The information about which protocol carried the message sits in the data the handset keeps alongside the text, and can only be read from the device.
What the carrier has, and what it does not
This is the question I am asked most often on the subject, almost always in the wrong form. The answer in two lines: the carrier keeps traffic data, not the text of the messages.
Traffic data establish that a message went from one subscriber number to another, at a given instant, through a given cell. They are a robust element, often decisive, and they answer questions of attribution and of placement in time and space that content alone does not answer. They do not say what was written, because the carrier does not have it: the service centre holds the message until it can be delivered — for a validity period of hours or days — and after delivery it does not retain the text.
On retention, the order of magnitude worth knowing before drafting an application: in Italy telephone traffic data are retained for twenty-four months and internet traffic data for twelve, with a longer regime for a narrow list of particularly serious offences. These are statutory terms, not commercial policies: once they expire the data is gone for everyone.
Who can ask for them, and how
Acquiring traffic data from a carrier belongs to criminal proceedings, for specified offences, and since 2021 it is authorised by a judge with a reasoned decree — previously a prosecutor's decree sufficed. In civil matters that door is closed: what remains is a data-subject request for one's own records and examination of the parties' devices. Building a civil case around the expectation of phone records is the commonest way of losing the months in which the data would still have existed.
Where the message lives on the handset
The text of an SMS sits in a structured store inside the handset's memory — a database, not a text file — together with information that never appears on screen: the instant of receipt recorded by the device, the one declared by the network, the read status, the protocol that carried it. Those accessory records are often more interesting than the text itself, because they are the ones a manual alteration struggles to keep consistent.
On deletion, what I have described about recovering deleted messages applies, with one peculiarity: removing a message from that store does not normally overwrite the space it occupied straight away, and for a variable period the content may still be readable as a record that is no longer referenced. It is a window, not a guarantee, and it closes as the handset is used. The more the device keeps being used, the less remains.
Two places that get forgotten, and that occasionally settle a case:
- The SIM card. It can hold a small number of saved messages, often old, often left there by a previous handset.
- The other person's phone. With no bilateral deletion, the recipient's copy is intact even when the sender's is gone. In a dispute between two parties the most useful device is frequently the one belonging to the party who did not raise the issue.
How to acquire an SMS so that it holds
The form in which a message reaches the court weighs as much as its content. In increasing order of resilience:
- A photograph of the screen. It is an image of pixels: it carries neither the protocol, nor the instants recorded by the device, nor any link to the handset it came from. One line disavows it.
- An export produced by the application. Better, because text is text and some accessory information survives; but it is a file produced by the handset itself, editable like any other, and it does not document its own origin.
- Acquisition of the device, with a copy and an integrity check. This is the form that holds, because the message arrives together with its context and the copy is bound to the original by a verification anyone can repeat — the mechanism I have described writing about forensic imaging and hashing.
Between the second and the third level lies the difference between a conversation that defends itself and one that needs to be believed. When the message is at the centre of the dispute, and especially when attribution is contested, there are no shortcuts: the device is required.
The operational corollary is the one that applies to any handset: while a question is open, the phone holding the messages must be set aside and left unused. Every day of ordinary use reduces what can be recovered, and does so silently. The other precautions for the first minutes are listed in what I have written about Android and iPhone forensics.
What to ask, and in what order
To frame a request for a technical examination of text messages — a question to an expert, an application, the first phone call — these five questions, in this order, avoid most of the dead ends:
- Which protocol is involved? SMS, RCS or iMessage change where to look and what is obtainable. If all that exists is a photograph, the answer is «we do not know yet».
- Which devices exist? The sender's, the recipient's, any replaced handsets never wiped, the SIM.
- What dates do the messages fall on? This tells you whether traffic data is still within the retention period, where the forum is a criminal one.
- Is attribution contested? If it is, content matters less than people think and the case turns on traffic data and devices.
- What is being looked for, and over what period? This is the question that makes the examination proportionate: a request circumscribed by number and by date range does not return the correspondence of strangers to the proceedings and withstands objections better.
The point in one line
Of an SMS the carrier knows that it was sent, not what it said; the handset knows what it said, but does not establish who sent it. The two halves of the proof sit in two different places, with two different expiry dates — which is why the first useful decision is to take the device and stop the clock.
Chat and messaging: how I work → · Mobile devices → · Back to insights