SOC 2 evidence screenshots: what a usable one shows
Updated
Auditors accept screenshots as SOC 2 evidence when the image ties itself to a system, a person and a date inside the period. Strike Graph's help center puts the date rule plainly: "Your auditor will not accept screenshots without a date & time stamp". A usable capture shows the system clock, the logged-in user, the URL or tenant, any filter beside the result, and the whole window, and it is saved as a named, dated file. Where the system can export the same data, send the export instead.
What the four sources say
Four pages were fetched for this guide. All four come from companies that sell compliance services or software, three of them platforms that collect evidence automatically, so read their advice against manual screenshots with that in mind. No accounting firm's page on screenshots was found, and TCSA notes that "No AT-C section specifies screenshot contents." The list below is practice, not a rule in the standard.
| Source | Who they are | What it covers |
|---|---|---|
| TCSA (Surendra Pal Singh; last reviewed August 2026) | Consultancy that prepares companies for the examination | Why a screenshot is weak evidence, the six things a usable one shows, Type 1 against Type 2 |
| Strike Graph help center (Stephanie Lorraine; undated) | Compliance platform vendor | The date and source rule, and effective dates inside the monitoring period |
| Bastion (Robin Coste, co-founder; updated November 2026) | Compliance platform vendor | Manual collection habits, exports over screenshots, redaction |
| Secureframe help center (Brady Price, 13 July 2026) | Compliance platform vendor | Configuration screenshots that may need to be observed again during a Type 2 period |
Sources: (TCSA) (Strike Graph) (Bastion) (Secureframe).
The checklist: what a usable screenshot shows
TCSA lists six things it says audit firms consistently ask for. This is one consultancy's list, put in this page's words (TCSA):
- The system clock in the frame. The operating system clock or a timestamp inside the application, not the date in the file's properties.
- The logged-in user, visible in the interface, so the capture can be tied to an account with real access.
- The full URL or a tenant identifier, which is how a reviewer tells production from a test environment.
- The whole window, uncropped. TCSA: "Cropping is the fastest route to a request for the export."
- Any filter or date range in the same frame as the result. TCSA: "A result without its query is an assertion."
- A named, dated file in the evidence folder, not an image pasted into a document.
The other sources point the same way with less detail. Strike Graph says the screenshot should show its source, "(i.e. - AWS, Jira, etc)" (Strike Graph), and Bastion's tips for manual collection include "Include timestamps in screenshots" and "Document the source system", with full context meaning "Include system name, date, user" (Bastion). The SOC 2 evidence checklist lists undated screenshots among the common gaps; this page is the longer version of that line.
When to export instead
A screenshot is an image of a system, framed and saved by the company being audited. TCSA explains the ranking with the evidence standard the auditor works under: documents beat oral answers, what the auditor obtains directly beats what is inferred, and originals beat copies. In TCSA's words, "A screenshot is a transformation." Bastion's short form is "System exports > screenshots when possible" (TCSA) (Bastion).
Screenshots still have a place. TCSA names the cases where a system has no export: a setting buried in a console, a vendor portal with no reports, a physical access panel, and a queue that has to be shown to be empty. Everywhere else, three habits from TCSA, in this page's words:
- Export, do not re-save. Opening a CSV in a spreadsheet application and saving it again changes the file, and an auditor who sees a modified time after the run date will ask what changed.
- Send the export with its context: the source system, the exact query or report name, the filter and date range, who ran it and when, and a record count that matches the count in the console.
- Read settings through the system's own records where you can. TCSA's example is branch protection read through the source control API rather than photographed: "A settings screenshot shows today only."
The generated Access Control Policy already works this way for one recurring control: the quarterly access review starts from exported lists, not from screenshots of admin pages:
The quarterly access review (Access Control Policy, 5.3)
Quarterly access review. Within the first two weeks of each quarter the Security Owner exports user and role lists from Okta, the identity services of AWS and Vercel, GitHub, Supabase, Stripe, Datadog and Slack and production databases, reconciles them against the personnel list from People Operations, and sends each system owner their list for attestation. Owners respond within ten business days, removals are completed within five business days of the response, and the signed-off review is retained as evidence.
The access review tool takes user exports from Okta, Google Workspace, Entra ID, GitHub and AWS IAM, flags leavers, inactive and admin accounts, and builds a review workbook in your browser, so the review starts from the export rather than from images.
The period problem
A screenshot shows one moment. TCSA: "A screenshot taken in November tells an auditor about November." How much that matters depends on the report (TCSA):
| Report | What it looks at | What a screenshot can do |
|---|---|---|
| Type 1 | The design of controls on one date | TCSA says a careful screenshot taken on or near that date is often the right evidence |
| Type 2 | Whether controls operated across a window, often months long | One capture covers one day. The auditor needs dated records across the window: an export, the system's change history, or a series of dated captures |
Strike Graph says the same from the platform side: "All evidence must be effective during your monitoring period, including evidence such as populations and samples that are requested after your monitoring period has ended." (Strike Graph) And Secureframe warns that configuration settings proven by an uploaded screenshot, its example being password settings, "may need to be re-observed by the auditor with a time stamp dated within the Type 2 period" (Secureframe). So a screenshot of a setting does not carry over from last year, and one taken during fieldwork covers only the day it was taken.
TCSA also warns about the move from a first Type 1 to a Type 2: a team that got through the Type 1 on screenshots keeps the habit and finds late in the window that the logs behind the earlier months were never kept. Which report to start with is in SOC 2 Type 1 or Type 2 first, and how the window is chosen in the SOC 2 observation period.
File names and folders
This section is the page's own suggestion, not a rule from a source. Bastion's first tip for manual collection is "Use consistent naming conventions" (Bastion), and a pattern that carries the date and the control makes each file readable without opening it:
evidence/
CC6.1-logical-access/
2026-03-14_okta_mfa-policy_screenshot.png
2026-03-14_okta_users-export.csv
CC8.1-change-management/
2026-04-02_github_branch-protection_api.json- Date first, in year-month-day order, so a folder sorts by time and the gaps show.
- Then the system, then what it shows, then whether it is a screenshot, an export or an API response.
- One folder per criterion or control, matching the control matrix, so the auditor's request list maps to a folder.
- Never rename an export after the fact in a way that changes its date; the date in the name should be the run date.
The generated Information Security Policy names the control matrix this layout follows:
Control evidence (Information Security Policy, 5.5)
Control evidence. The Security Owner maintains a control matrix mapping each criterion in scope (Security and Availability) to the policy statements and evidence artifacts that satisfy it. Evidence is collected on the cadence in the matrix and stored where the CPA firm can be given read access.
The text above is rendered from the policy templates for the Northwind Cloud Inc sample intake, not typed by hand, so it changes when a template does. Policyseed sells a kit, so read it as one example of a written rule.
Redaction without losing the evidence
Bastion lists over-redaction as one of five common evidence pitfalls: "Evidence is too redacted to be useful". Its prevention steps are "Consult auditor on redaction needs", "Preserve essential information" and "Offer unredacted viewing if needed" (Bastion). In practice, keep the four things the checklist relies on (clock, user, system and filter) visible, and cover what the request did not ask for.
Redaction is also a policy question. If your company adopted the generated set, its Acceptable Use Policy names screenshots as a place customer data does not go:
Where customer data stays (Acceptable Use Policy, 4.5)
Customer data and Confidential or Restricted information, as defined in the Data Classification and Handling Policy, stay inside approved systems and are never copied to personal email, personal cloud storage, removable media, screenshots, chat messages or unapproved tools.
So a screenshot that shows customer records should stay in the approved evidence store, with customer data covered or filtered out before capture, and be shared with the auditor through the access item 5.5 above describes rather than by email or chat.
Before you send a screenshot
A last check, built from the sections above. It is the page's own list, not a source's.
- Could the system export this instead? If yes, send the export.
- Are the clock, the user and the URL or tenant all in the frame?
- Is the filter or date range visible beside the result?
- Is the date inside the observation period, and is it the date the control ran?
- Is the window whole, and is the redaction limited to what the request did not need?
- Is the file named with its date and saved in the control's folder?
To rehearse the questions your evidence has to answer, practice with the mock audit: an AI auditor asks questions from your own policies and says whether each answer matches them, which shows which records you will be asked for. What the requests look like during the audit itself is in SOC 2 fieldwork, and collecting evidence by hand from start to finish in SOC 2 without a compliance platform.
Frequently asked questions
- Do SOC 2 auditors accept screenshots?
- Yes, when the screenshot can be tied to a system, a person and a date inside the period. Strike Graph's help center says "Your auditor will not accept screenshots without a date & time stamp" and that the capture should show its source, such as AWS or Jira. TCSA, a consultancy, says auditors accept a careful screenshot routinely where a system has no export, and ranks a system export above it. The CPA firm doing your audit makes the final call.
- Does the file's created date count as the timestamp?
- No. TCSA's list asks for the system clock in the frame, meaning the operating system clock or a timestamp inside the application, "not the file's metadata date". A file date changes when the file is copied or saved again, and it is not part of the image the auditor reads.
- Can I use my phone to photograph the screen?
- No fetched source talks about phone photos. Judged against TCSA's list, a photo of a monitor tends to lose what the auditor checks: it crops the window, the address bar and the clock are hard to read, and it is one more step away from the system. Use the computer's own screenshot tool and keep the full window. For a physical device with no screen export, such as a door access panel, a photo may be the only option; ask your CPA firm before relying on one.
- Can I retake a screenshot after the observation period ends?
- Not as evidence for the period. TCSA's point is that evidence put together after the fact describes the present, not the period, and Strike Graph says all evidence must be effective during the monitoring period, including populations and samples requested after the period has ended. A new screenshot shows how a setting stands today. What you can do is go back to the source system: TCSA's example is a missing March record: noticed in April, it can often still be recovered from the source system; noticed in October, it cannot.
- Is one screenshot enough for a Type 2 report?
- Rarely. TCSA writes that "A screenshot taken in November tells an auditor about November." A Type 2 report covers a window, so a setting needs a dated record across that window: an export, the system's change history, or a series of dated captures. For a Type 1 report, which looks at a single date, TCSA says a screenshot taken on or near that date is often the right evidence.
- How much should I redact?
- Only what the auditor does not need. Bastion lists over-redaction among five evidence pitfalls and says "Auditors need to verify evidence". Its prevention steps are to ask the auditor what they need redacted, keep the essential information, and offer an unredacted view if needed. Keep the clock, the user, the system name and the filter visible; cover customer data you were not asked for.
Policyseed provides governance policy templates and AI tailoring. It is not legal advice and not a compliance guarantee. Management adopts the policies; the CPA firm performs the SOC 2 examination.