Digital evidence · screenshots, screen photos and screen recordings

A screenshot proves only that someone made an image.

A screenshot, a phone photo of a screen or a screen recording shows what appeared on one person's screen. It cannot show that the website ever sent that content. Today almost anyone can produce a perfect copy of any page, so the moment the other side says “that's fake”, the image alone has little left to stand on.

Three ways a screen image is made to lie

None of this needs special skill any more. We describe it so you understand the risk, not how to do it.

1

Edit the page in your own browser

Every desktop browser has built-in developer tools that let a user change any text, name, date or number on a page before taking the screenshot. The result looks exactly like the real site, because it is the real site's design.

2

Rebuild the page elsewhere

With the page's own layout and styles, a page can be reproduced pixel for pixel on another server and filled with invented messages. A computer-savvy friend, or simply an AI assistant asked for help, can do this in minutes.

3

Generate it with AI

Image generators can produce plausible chats, posts and photos from a sentence of text. These are less exact than a rebuilt page, but good enough to sow doubt.

What courts and rule-makers say

§

United States

The federal evidence rules committee has drafted a rule for suspected deepfakes, noting that anyone can now make one cheaply. It is not adopted yet, and a 2026 survey found only about 2% of federal judges had faced a deepfake challenge so far. The direction is clear: when an image is credibly challenged, the party relying on it will have to do more to prove it is authentic.

§

England and Wales

Official guidance to judges warns that AI tools are being used to produce fake text, images and video. In 2023 an employment tribunal found WhatsApp screenshots had been deliberately faked, and treated that party's evidence with considerable suspicion.

§

European Union

Europol's report on deepfakes notes that courts usually trust audio-visual evidence, lists falsifying electronic evidence as a criminal use of the technology, and calls for a clear chain of custody for digital evidence.

§

The source matters

As early as 2011 Maryland's highest court reversed a conviction because social media printouts were not authenticated, and pointed to obtaining the information from the site itself. In 2025 a California court dismissed a case after finding that video and image exhibits were AI-generated.

If the other side says “it’s fake”

1

Step 1: they bring a sign of fabrication

Under the draft US rule, a bare claim is not enough: the challenger “must present evidence sufficient to support a finding of such fabrication”. Against a self-made screenshot that is often easy: one inconsistency, a timeline that does not fit, or an expert explaining how simply such an image is made.

2

Step 2: you must prove it is authentic

Then the item is admissible “only if the proponent demonstrates to the court that it is more likely than not authentic”. With a screenshot alone, all you can offer is your own word. An independent capture from the real server, sealed by a third party, answers this step directly.

3

Status and scope

The rule is a draft, held in abeyance while the committee studies how often deepfakes appear. Belgian and French courts have no such rule but weigh evidence freely, and a disputed screenshot loses weight there too, as the case law shows.

Quotes from the Advisory Committee on Evidence Rules, draft Rule 901(c), report of 1 December 2025.

Timestamping your own screenshot is not enough

A qualified timestamp on a screenshot you made proves that the image file existed, unchanged, from that moment. It says nothing about whether the website ever showed it, because the image came from your device. That is useful for your own photos and documents, such as damage photos or a contract. It does not make a screenshot of someone else's website trustworthy.

What does work: an independent capture

✓

Captured by someone else

An independent provider loads the page in its own clean browser, on its own servers. You are not the source of the image, so the question “did you edit it?” does not arise.

✓

Straight from the real server

The encrypted network traffic is recorded, so the package shows which server sent the content, over a secure connection, at that moment. A copy rebuilt elsewhere would come from a different server.

✓

Sealed and dated by a third party

A qualified electronic seal and a qualified time-stamp from an EU trust service fix the content and the time. Any later change breaks the seal.

✓

Checkable without the provider

A lawyer, bailiff or expert can verify the package with standard tools. A bailiff's report is the other independent route, at €300 to €700.

Courts still admit screenshots, especially when nobody disputes them; see our case-law overview. The problem starts when the other side does dispute them, and that is exactly when you need the evidence most.

Sources

Every quotation below was checked against the text of the source.

  1. Advisory Committee on Evidence Rules (US Courts), report to the Standing Committee, 1 December 2025, committee note to draft Rule 901(c) · “any member of the public now has the capacity to make a deepfake, with little effort and expense”; draft rule text: “must present evidence sufficient to support a finding of such fabrication” … “only if the proponent demonstrates to the court that it is more likely than not authentic” · source →
  2. Courts and Tribunals Judiciary (England and Wales), AI guidance for judicial office holders, 8 October 2025 · “AI tools are now being used to produce fake material, including text, images and video” · source →
  3. Europol Innovation Lab, Facing reality? Law enforcement and the challenge of deepfakes, 2022 · “In court, audio-visual evidence is usually trusted to be an authentic representation of events” · source →
  4. Griffin v. State, 419 Md. 343 (Maryland Court of Appeals, 2011) · “The potential for abuse and manipulation of a social networking site by someone other than its purported creator” · source →
  5. Lehtonen and others v Hansbiomed Europe Ltd, Employment Tribunal London Central, 2200627/21, January 2023 · “not an authentic or contemporary string of messages” · source →
  6. Mendones v. Cushman & Wakefield, Inc., No. 23CV028772 (California Superior Court, Alameda County, 9 September 2025), as reported by the Volokh Conspiracy · “products of GenAI” · source →
  7. R. Pfefferkorn, “Deepfakes” in the Courtroom, 29 B.U. Public Interest Law Journal 245 (2020) · “Software for creating deepfake images, video, and audio is already freely available online and fairly easy to use” · source →
  8. R. A. Delfino, Deepfakes on Trial, 74 Hastings Law Journal 293 (2023) · “deepfakes are invading the courts” · source →
  9. M. R. Grossman, P. W. Grimm, D. G. Brown and M. Xu, The GPTJudge: Justice in a Generative AI World, 23 Duke Law & Technology Review 1 (2023) · source →
  10. Advisory Committee on Evidence Rules, agenda book, 7 May 2026, Federal Judicial Center survey of federal judges · source →

Need to prove what a page or chat showed?

Record it with instantproof: an independent capture from the real server, sealed and time-stamped. €0.95 per web page, €10 per recorded chat or session.

Capture it independently How a recorded session works →