🔐 Who can use this feature?
Survey types: All except Effectiveness 360 surveys and Central Surveys.
Subscriptions: Any that include Engagement or Effectiveness
Free-text comments are often the most valuable part of a survey, and occasionally the most difficult to manage. A comment can name a colleague or repeat something that should not be read by everyone who opens a shared report. Sensitive Comment Management gathers those comments on a Sensitive comments tab in your surveys admin report, allowing you to keep track of anything that may need to be redacted.
Identifying sensitive comments is automatic and cannot be switched off, but changing or redacting an identified comment is manual and entirely your decision. Culture Amp never edits, rewrites, or removes a comment on its own, so every comment on the tab reads exactly as the person wrote it. Being flagged is an invitation to look, not a verdict.
💡 Tip: If you would like some more suggestions and advice around the best practice approach to sensitive comment management, check out our "science-behind" article for some useful tips and guidance, all backed by our people science team.
Open the Sensitive comments tab
To access the Sensitive comments tab of a survey:
Open Feedback > Surveys and click View Reports next to the relevant survey.
Open the Comments report
Click on the Sensitive comments tab
Sensitive comments sits alongside the All comments, Topics, and Saved comments tabs. It will appear on any surveys that collect free-text comments, but is not available for 360 effectiveness surveys or Central Surveys.
Who sees it: Any account or survey administrator with appropriate permissions to access the Administrator Report. Report viewers, report owners, and colleagues with read-only report access do not see the tab, which comments were flagged, or what a comment said before redaction.
No sensitive comments found: The survey was checked, and nothing was flagged, so there is nothing for you to do.
Sensitive comments not available for this survey: The comments were never checked, usually because the survey closed before the feature was released. Open a Support Conversation and request a "re-analysis of your sensitive comments". The Product Support team should be able to help.
How comments are flagged, and who decides what changes
Two separate checks populate the Sensitive comments tab, running on their own as comments come in.
A large language model reads each comment and returns a yes or no for each of eight sensitive topics. It is instructed never to quote or rewrite the comment.
Your own keywords are matched against the comment text. These are optional words or phrases your organization adds. Without them, the model does all the work.
Both feed the same list, so a comment can arrive from the model, a keyword, or both. Nobody is notified that a comment was flagged, including the person who wrote it. Neither check changes anything; redaction only happens through the export and import workflow discussed later in this article. Many administrators might review the list and decide nothing needs changing, so never feel that you are being forced to act on the flagged comments.
The eight sensitive topics Culture Amp looks for
The eight topics are the same for every account. You cannot remove or add any new topics, and a comment can carry more than one.
Topic | What it covers |
Name | The comment contains the name of a person. |
Discrimination | Unfair treatment based on a protected characteristic such as age, gender, ethnicity or race, disability, sexual orientation, religion, family situation, or socio-economic status, including treatment driven by favoritism or bias. |
Profanity | Offensive language. |
Insult | A person or group, for example "management", is the direct target of a negative label or judgment about their character. Negative language about a job, task, or situation is not an insult. |
Harassment | Unwanted or harmful conduct, such as aggressive pressure, threats to someone's employment or shifts, belittling or degrading treatment, intimidation, bullying, or humiliation. |
Self-harm | An intention or a threat to cause harm to oneself. |
Threat | An intention to cause physical harm to a person or damage to property. |
Location | A physical address or another identifying physical location. |
ℹ️ Note: These eight topics are a separate set from the comment topics in other reporting. The 28 predefined comment topics in Text Analytics in Comments Reporting describe what a comment is about, while the eight here describe why it might need your attention. The Topic column on your sensitive comments tab only ever shows these eight.
Add your own keywords before a survey launches
Keywords catch language specific to your organization that a general model would not flag, such as a site name, a product name, or an internal nickname. While yoiu cannot create your own topics, you can add keywords to any of the eight that are provided. These are set once for the whole account, not per survey.
As an Account Administrator, click the Settings icon from your Home page.
Click Surveys from the settings menu
Sensitive comment topics should be the first option you see on the survey settings page
Click Edit keywords next to the topic you would like to add to
Select Add keyword, type the word or phrase, and save.
Keywords follow a few rules:
English letters, spaces, and hyphens only. We match keywords against the English version of every comment, including the translation of one written in another language, so keywords in other languages never match.
Up to 45 characters per keyword, and a phrase of several words is fine.
One keyword per field. There is no bulk upload, and duplicates are rejected when you save.
Whole words and phrases, not fragments. Matching ignores capitalization and punctuation, so "idiot" matches "idiot!" and "IDIOT". It will not match inside an unrelated word, so "kill" does not flag "skills".
Common English variations also match. "manager" also flags "managers" and "managing". This is deliberate, and the likeliest reason a comment appears against a keyword you did not expect.
A survey copies your keyword list at launch and uses that copy for the rest of its life, so adding a keyword later has no effect on a survey already launched. Review your list before launch, not after results arrive.
ℹ️ Note: If you have added new keywords while the survey is live, and would like these to be considered by the sensitive comments analysis, you can open a Support Conversation and request a "keyword re-sync for our sensitive comments". The support team should be able to help.
Understand the Sensitive comments table
Each row you see on the Sensitive comments tab is one comment, with six columns helping to explain why it has been flagged:
Comment
Date submitted
Status
Topic
Keyword
Sentiment.
Keyword is empty for any comments the model found on its own rather than by using one of your keywords. Work from Status, which tells you what has already been decided and what still needs review.
Status | What it means |
Needs review | Nobody has looked at this comment yet. Every comment starts here. |
Not redacted | Somebody reviewed it and left the wording as written. |
Redacted | The wording was replaced with text an administrator supplied. |
Status is worked out for you, not chosen. On import (explained in the next section), Culture Amp compares your edited text with the comment as it was originally written. Identical text records 'Not redacted', while different text records 'Redacted'.
A few controls help when the list is long:
Language: the Comment column shows the English translation by default, and the Translate comments to original language toggle switches it back.
Comment attributes filters: status, topic, keyword, sentiment, and date. Use these to filter the list as you need
Demographics filters: the demographic filter set up for your Comments Report, explained in Set Up a Filter for the Comments Report.
Comment text search: looks for words inside the comments themselves.
Sentiment: each comment carries Positive, Negative, Neutral, or No sentiment.
Redact comments with the export and import workflow
Redacting sensitive comments happens in a spreadsheet rather than on screen, so you can work at your own pace, share the file with whoever needs to approve the wording, and apply every decision in one pass.
Step 1: Export the sensitive comments
From the Sensitive comments tab, select Export sensitive comments and stay on the page while the file is prepared. A notification saying Ready to download will appear when the file is ready.
What the file holds:
Every sensitive comment in the survey, including ones already dealt with. On-screen filters are ignored.
One sheet named 'Sensitive comments', newest comments first.
Columns: Original comment text, Date submitted, Status, Topics, Keywords, Sentiment, Updated comment text, Comment ID, then any demographic columns that the survey allowed as comment filters.
Do not rename, reorder, or delete columns, and never edit Original comment text or Comment ID, which match each row to the correct comment.
Status and Sentiment as capitalized values such as NEEDS_REVIEW, not the friendly on-screen labels.
Topics match the tab exactly.
Updated comment text is empty for comments with NEEDS_REVIEW status, but pre-filled with the current wording for those already decided.
Step 2: Decide what each comment should say
For any comments you've decided to redact, put the final wording you want in the Updated comment text cell of every row, including rows you decided to leave alone. A single empty cell stops the whole file from importing.
There are 3 valid options:
Replace the whole comment. If it cannot be shared at all, write something that shows a change was made, such as
[Comment redacted by the survey administrator]. The comment records as Redacted.Change only part of it. Often the kindest option, it keeps the feedback and removes the identifying detail. "Priya in payroll" becomes "a colleague in payroll", which also records as Redacted.
Leave it exactly as it is. Copy the text from Original comment text into Updated comment text unchanged. The comment records as Not redacted, which marks it as reviewed and keeps the employee's words intact.
A stray space at either end of Updated comment text, or a difference in how line breaks were saved, is not read as a change.
Step 3: Import the file
Back on the Sensitive comments tab, select Import to redact comments.
Select file. Drag it onto the page or browse for it from the files on your device. It will accept one .xlsx file of up to 128MB
Validate file. This checks every row before anything changes. If the file is clean, you are told there are no errors and that every row will be imported.
If anything is wrong, nothing at all is applied, and you will receive an error and a downloadable file showing guidance on how to fix the sheet before trying again.
Import. Finalize the import of the validated file to apply the changes
The import may take 25 to 30 minutes, and you will see a progress bar on the tab to show it is still loading.
⚠️ Warning: Only one sensitive comments import can run at a time. If another admin is already importing, you will see Import already in progress and need to wait for theirs to finish.
Step 4 (optional): Restore a comment you redacted
A redaction is not permanent, and putting a comment back to it's original state is the same process as the initial redaction. Export the comments, find the row, copy the text from Original comment text into Updated comment text so the two match, and import again.
Matching text records the comment as Not redacted rather than Redacted, and it will not return to Needs review status.
Fix a file that did not pass validation
Validation is all or nothing. One bad row means the whole file is rejected and nothing in it is applied, so a failed import never leaves you with half your decisions in place. Your file is untouched, so you can fix it and try again.
When rows fail, you can download a spreadsheet of the errors. Its Validation errors sheet keeps your rows in their uploaded order and shows Original comment text, Updated comment text, Comment ID, and a Warning column naming the problem and the row number. Fix the warnings in the original file, then upload it again.
What the warning says | What to do |
A missing Comment ID | Copy the value back from your original export. |
A Comment ID that is not valid | Spreadsheets sometimes reformat the value on paste. Copy it again from the export. |
Missing original comment text | Restore the comment exactly as exported. |
Missing updated comment text | Add your wording, or copy the original across to leave the comment as it is. |
The same Comment ID on more than one row | Remove the duplicate row. |
No comment found for a Comment ID | The response was deleted or reset. Remove the row. |
The comment on record does not match the text in your file | Original comment text was edited by accident, or the comment changed after you exported. Start again from a fresh export. If the cell ends with |
The comment is no longer flagged as sensitive | It cannot be redacted here. Remove the row. |
What redaction changes, and when to do it
Redaction replaces the text of the comment on the survey response itself. From then on, everywhere the comment appears shows your new wording: the All comments tab, the Comments Report, shared reports, and exports. The wording the employee typed survives only in the Sensitive comments record, available to account and survey administrators in the Original comment text column of the export.
⚠️ Warning: Nothing holds a flagged comment back while you review it. Unredacted sensitive comments stay visible in the Comments Report and will appear in shared reports, so redact before you share.
If you have already shared or published a report, people with access may have read the original wording, and redacting afterwards changes only what they see from then on. Guide to Report Sharing explains who gets access and when, which helps you plan the gap between a survey closing and results going out.
Comments flagged as Self-harm
The Sensitive comments tab never shows you who wrote a comment, including a comment flagged as Self-harm, and nobody is alerted automatically when a comment is flagged. This tab is something you need to open manually and read rather than something that will push notifications to you (as is the case with all comments flagged as sensitive).
If you believe a comment needs a response beyond redaction and you may need to identify a participant for this purpose, open a Support Conversation and 'ask a person' to be connected with support. They can refer you to our relevant legal and privacy channels to see if we can lend a hand.
What Sensitive Comment Management does not do
A few limits to plan around:
Redaction happens in a spreadsheet. There is no way to edit a comment directly on the tab.
No charts or insights on sensitive comments. You get comments to read and decide on, not summaries.
The eight sensitive topics are fixed. You can add keywords under a topic, but not create a topic of your own.
Keywords are added one at a time, with no bulk upload available.
Flagged comments are not held back. A flagged comment stays visible everywhere it normally appears until you redact it.
Troubleshooting
Your redaction has not appeared yet
Allow 25 to 30 minutes from the moment you imported, then refresh the page.
You are able to leave the page at any stage and the import will continue in the background.
If you miss the 'import completed notification,' you should still be able to tell once it is completed by seeing the status changes on the redacted comments, and the import/export buttons no longer being inactive.
A comment you tried to restore still shows as Redacted
Compare your Updated comment text with Original comment text in a fresh export. The two have to match exactly (in the language the comment was written in); otherwise, the status will show redacted.
FAQ
Why do we see this tab when AI features are switched off for our account?
Why do we see this tab when AI features are switched off for our account?
The Sensitive comments tab is part of survey reporting, not your AI feature settings, and the review check runs on every comment either way. There is currently no way to turn off the flagging of sensitive comments for eligible surveys.
Can we still change the sentiment on a sensitive comment?
Can we still change the sentiment on a sensitive comment?
A sentiment change on the 'All comments' tab carries across while a comment is in Needs review or Not redacted status. Once it is Redacted, the sentiment shown against it stays as it is.
Can we completely remove any comments we aren't redacting from the import file?
Can we completely remove any comments we aren't redacting from the import file?
Technically, yes! Deleting any comment rows that don't need redactions to be made will allow the import file to be accepted, but it will also leave those comments in 'Needs review" status on your Sensitive comments tab.
Due to this, it is recommended to leave all flagged comments on the import file, and simply copy the exact original comment text into the updated comment text tab if no redaction is needed. This ensures the status will move to not redacted rather than needs review, making it easier to track and manage.
How do we request re-analysis of sensitive comments, and what are our options?
How do we request re-analysis of sensitive comments, and what are our options?
Only our Product Support team can re-run the analysis, so open a Support Conversation and request a "re-sync" or "re-analysis" of your sensitive comments, being as clear as possible about which option you need based on the below:
Option | When to ask for it |
Update Keywords and analyse comments (useful when new keywords have been added mid-survey) | You added or changed keywords after the survey launched, and want the comments checked against your current list. Comments left after the new keywords are added will automatically search for them, but comments left before the keywords were added will only pick them up if this resync is actioned. |
Analyze sensitive content (useful for surveys that closed before the feature was released) | The tab says Sensitive comments not available for this survey, usually because it closed before the feature was released. This takes much longer (up to 2 hours) and should only be requested if no sensitive comment analysis has been completed on the survey at all. |
Save your keywords before you ask, as a keyword re-sync has nothing to match against if the list is empty. Re-syncs run on one survey at a time, and cannot start while a redaction import is in progress. Only comments at Needs review are looked at again, so anything you have settled as Redacted or Not redacted stays as it is, and because the model does not re-read a comment it has already read, a resync cannot change what it flagged, so it should not be requested in cases where you simply want the analysis to run again when it already has, with no new keywords added.
➡️ Next steps
The Science Behind Senstitive Comment Management for further information on our people science-backed recommendations
Getting Started with Report Sharing for planning when results go out and to whom.
Text Analytics in Comments Reporting for the comment topics and sentiment you already use in reporting.
Translate Comments to English for how comment translation works across your account.
Assigning Report Viewers and Owners for checking who can currently open a report.
💬 Need help? Open a Support Conversation to speak with our Product Support team.