Treat the post and image as connected evidence
The post is the published communication; the screenshot is an image you created of one interface state. It may preserve an edit label, response, engagement count or layout that later disappears, but it can also crop context. Cite the direct post where available and record precisely when and how the image was taken.
A deleted post or private account should not automatically be treated as an ordinary public publication. Examine research purpose, participant expectations, platform terms, privacy, copyright and ethics approval. For vulnerable communities or sensitive subjects, searchable wording may identify a person even after the handle is blurred.
Record identity and time precisely
Keep the display name and unique handle because either can change. Record platform, visible posting time, time zone and direct URL. State whether it is an edited post, quote, repost or reply. Retain the full wording only to the extent necessary for evidence.
Likes, views and replies change continuously. If they are data, give an observation time and definition. A screen may show rounded totals or personalised ordering. For quantitative claims, prefer an authorised export or official API where feasible and explain collection limitations.
Read the thread and platform grammar
A sentence may answer an earlier post, use irony or form part of a longer explanation. Capture the relevant sequence and explain it. A repost does not transfer authorship. Embedded photographs, videos and articles remain separate works with their own provenance.
Algorithms may reorder replies and recommendations. Mark the boundary between the post and your personalised interface. Crop adverts and unrelated accounts from the published figure where appropriate, while preserving a protected original if lawful and ethically justified.
Separate attribution, legal basis and research ethics
Citation establishes provenance. Whether text, profile image, username and embedded media may be reproduced is another question. Copyright exceptions depend on jurisdiction, purpose and proportion. Decorative reproduction differs from analysis of the interface itself. Obtain institutional or legal advice for difficult cases.
Publicly readable personal information can remain personal data. Consider lawful basis, minimisation, notice and the expectations of the group studied. Paraphrase or controlled access may reduce risk in qualitative research but may also change the object. Record that trade-off.
Build the reference and figure caption together
A general reference structure is: “Display name [@handle] (date, time), opening words or descriptive title [Social media post], Platform, direct URL (accessed date).” The figure caption can add: “Screenshot by the author, captured [time]; replies cropped; third-party handles redacted.” Apply the required style.
If the post is gone, use an ethically and lawfully available archive record. Do not publish a private capture without review. The guide to citing software and webpage screenshots explains general provenance fields.
Store original, research copy and publication image separately
Give the capture an ID and keep an edit log. Access to the original may need tighter controls than the redacted thesis figure. Before submission, check whether the post was edited, deleted or annotated by the platform. Do not rewrite the historical screenshot; add a dated status note and distinguish the former communication from later developments.
Where many posts are studied, create consistent collection rules and record failed or unavailable captures. Avoid selecting only striking examples after seeing the outcome. A sampling log makes the visual evidence part of a method rather than an anecdotal decoration.
Run a second-person evidence check before publication
Ask a colleague who did not collect the post to compare the stored capture, platform record, caption and analysis. The check should cover the exact account identifier, post URL, displayed time, wording, redactions and any quoted reply. Similar display names, parody accounts and cropped reply chains are common sources of mistaken attribution.
Record the outcome against the capture ID. Correct the publication derivative and metadata when necessary, but preserve the original observation and edit history. If the live page now differs, report that as a later state rather than altering the dated capture. A short verification note gives the reader a clearer boundary between what was observed, what was inferred and what subsequently changed.