01 · The Question
If you remove the username, is an online quotation anonymous?
A researcher quotes a social media post but removes the author's username. In the paper, the source becomes “User 12,” “Forum participant,” or perhaps no attribution at all.
At first glance, the person appears anonymous.
There is a problem. Unlike many interview transcripts, online text may already exist in a searchable database. A reader can copy a distinctive sentence from the paper, paste it into a search engine or the platform's own search function, and potentially locate the original post. From there, the reader may find the author's username, profile, photographs, location, other posts, and considerably more information than the researcher ever published.
In online research, the quotation itself can function as an identifier.
03 · What You Need to Know
Online text can behave like an indirect identifier
Traditional anonymization assumptions do not always work online
Researchers commonly protect participants by replacing names with pseudonyms and removing obvious identifying details. That approach can work well for interview transcripts that exist only within the research team's files.
Online material is different because the source text may already be publicly searchable.
UKRI's guidance on internet-mediated research identifies potential identifiability as a specific ethical concern and advises researchers to protect individuals' identities wherever possible unless disclosure is critical to the research or permission has been obtained. It also cautions that the public nature of online information should be critically examined rather than assumed.
Earlier ESRC work on social media research makes the quotation problem especially explicit: direct quotations used in research can be traced to their original authors through search engines, potentially compromising anonymity.
The quotation itself can become identifying information
Researchers often think of identifiers as names, email addresses, usernames, photographs, or locations. Online research adds another possibility: a distinctive sequence of words.
Direct identifier
Information such as a real name, recognizable photograph, account handle, or other element that directly points to a person or account.
Search-based identifier
A distinctive passage, phrase, or combination of details that can be used to locate the original online source and thereby reconnect the research material to its author.
Removing the first kind does not necessarily remove the second.
Searchability creates a reverse lookup problem
Consider how a reader encounters a quotation in a research article. They already know that the words came from somewhere online. They may know the platform, approximate date, topic, community, or hashtag because the methods section provides those details.
A distinctive quotation can then function as a search query.
Once the original post is found, the reader may gain access to much more than the quotation itself. They may discover a persistent username, photographs, biography, workplace, location, relationships, other sensitive posts, or links to additional accounts.
The research article has therefore provided a bridge between an ostensibly anonymized research subject and an identifiable digital identity.
Not every quotation is equally searchable
Searchability depends partly on textual uniqueness.
| Type of quotation |
Likely traceability |
Why |
| “I agree.” |
Usually low |
The wording is extremely common and likely appears in many unrelated sources. |
| A distinctive multi-sentence personal narrative |
Potentially high |
Longer unusual combinations of words can function almost like textual fingerprints. |
| A quotation containing a rare hashtag or username mention |
Potentially high |
The additional searchable elements substantially narrow possible sources. |
| A common statement plus exact date and community name |
Potentially moderate to high |
Context supplied by the paper can narrow the search even when the wording alone is common. |
| A faithfully paraphrased statement |
Often lower |
The exact searchable wording has been altered, although contextual details may still identify the source. |
Researchers should therefore assess quotations individually rather than assuming all verbatim text carries the same identification risk.
Contextual details can make a quotation easier to trace
The quotation does not operate alone.
A paper might report that the post came from a particular subreddit, forum, Facebook group, hashtag, month, geographical area, profession, or event. Each detail can narrow the search space.
Even if the exact quotation produces many results, combining it with contextual clues may make identification straightforward.
This means researchers should assess the whole published representation, not simply whether the quotation contains a name.
Pseudonymous accounts can still represent identifiable people
Finding the original account does not always reveal a person's legal identity. But that does not necessarily eliminate ethical concern.
A persistent pseudonym may itself carry reputation, relationships, community membership, posting history, or social significance. Revealing that a particular pseudonymous account authored a sensitive statement can therefore matter even when the research reader never learns the person's government name.
Moreover, many pseudonymous accounts contain enough biographical information to permit further identification.
This is one reason researchers should treat the broader question of whether internet users are participants or data sources separately from whether their identities can be reconstructed.
The consequences of identification matter as much as its probability
A searchable quotation about enthusiasm for a television series presents a different risk profile from a searchable quotation describing abuse, addiction, illegal behavior, financial hardship, sexual experiences, mental health difficulties, workplace misconduct, or political activity in a dangerous context.
UKRI's current ethics guidance emphasizes that research can affect participants through reputational risks, privacy loss, employment consequences, family relationships, and disclosure of sensitive behavior. It also identifies social media research involving potentially identifiable quotations and sensitive discussion as a situation that may warrant full ethics review.
A useful risk assessment therefore asks two questions: how easily could the person be identified, and what could happen if they were?
Public source material can still create an anonymity problem
Researchers sometimes respond that no privacy has been lost because the original post was already public.
Publicity is relevant. It may substantially affect whether public social media material can be used as research data. But it does not make every subsequent disclosure equivalent.
A post may have been visible to an unrestricted audience yet obscure in practice. A research article can give it renewed visibility, connect it to a sensitive analytical category, preserve it after deletion, or direct readers toward the author's broader profile.
The distinction between public accessibility and ethically appropriate research use therefore remains relevant at the reporting stage.
Attribution may sometimes be preferable to false anonymity
Not every online source should be anonymized.
If researchers analyze an official statement deliberately issued by an organization, a public statement by an identifiable official, or another source where authorship is integral to the analysis, transparent attribution may be appropriate. UKRI guidance recognizes that protecting identity may not be appropriate when disclosure is critical to the research, including some statements by public officials.
The problem arises when researchers promise or imply anonymity while publishing material that can readily reverse that protection.
Searchability should be assessed before publication
Researchers using online quotations should consider testing proposed quotations through ordinary search methods during analysis or manuscript preparation.
This is not a guarantee of anonymity. Search indexes change, platforms have different search functions, and future technologies may make linkage easier. Still, testing can reveal obvious identification pathways before publication.
Watch Out
A quotation that does not reveal an author today is not guaranteed to remain untraceable indefinitely. Anonymization should therefore be understood as risk reduction rather than an absolute promise that identification is technically impossible.
Changing the quotation can reduce risk, but it changes the evidence
One possible response is to alter or paraphrase the wording so that exact-string searches no longer locate the source.
That can be appropriate when the analysis concerns meaning rather than exact language. It becomes more difficult when wording itself is evidence.
The decision about whether researchers should paraphrase online posts to reduce searchability therefore requires a trade-off between privacy protection and analytical fidelity.
04 · A Practical Example
A quotation can undo anonymization in seconds
Hypothetical Example
A pseudonymous post about workplace discrimination
A researcher analyzes public online discussions about workplace discrimination. One contributor uses a pseudonym and writes a distinctive paragraph describing an incident involving a supervisor. The researcher removes the username and labels the author “User 27.”
Initial assumption
Because the username, employer, location, and names have been removed, the researcher initially considers the quotation anonymous.
Search test
The researcher searches one distinctive sentence from the proposed quotation using an ordinary search engine.
Result
The original post appears among the first results. The user's profile contains additional posts revealing an occupation, approximate location, employer characteristics, and personal photograph.
Risk
Publishing the quotation would provide readers with a straightforward route from the research article to considerably more information about the person.
Decision
Because the study analyzes experiences rather than exact linguistic features, the researcher considers faithful paraphrasing and removes unnecessary contextual details while preserving the substantive analytical point.
The researcher had removed the obvious identifier. The search engine found the less obvious one.
07 · A Quick Checklist
Before publishing an online quotation, test whether it identifies the source
Before treating a quotation as anonymous, check:
Remove direct identifiers unless attribution is ethically appropriate and methodologically necessary.
Search distinctive portions of the proposed quotation using ordinary search tools to identify obvious traceability.
Check whether platform names, community names, dates, hashtags, locations, or demographic descriptions narrow the search substantially.
Examine what additional personal or sensitive information becomes visible if a reader finds the original account.
Assess both the likelihood of identification and the consequences if identification occurs.
Determine whether exact wording is scientifically necessary or whether a faithful paraphrase can support the analysis.
Do not describe material as anonymous merely because usernames have been removed.
Ensure the reporting strategy is consistent with the study's ethics approval, consent arrangements, confidentiality commitments, and applicable legal requirements.
