AI in Candor is optional

You write the survey. If you’d like help, you can switch AI on, and turn off any part you’d rather not use.

Off until you switch it on

AI help is off in a new Candor account. You switch it on in your own account, and you can switch off any AI feature in Settings at any time.

One feature reads what people wrote

Only Response themes sends your respondents’ written answers to the AI. The other AI features work from your own questions or from the numbers.

Included in every plan

AI help costs nothing extra. Each plan includes a monthly allowance.

You stay the author. Nothing the AI suggests changes your survey unless you add it, and no AI feature changes the answers people gave. Everything the AI writes is labeled as AI-generated, in Candor and when you copy it.

What does each AI feature do, and what does it send?

The first five can each be switched off on their own. The second opinion comes with the main AI switch.

Feature What it does What it sends Keep in mind

Question suggestions

 What it doesSuggests questions that fit the survey you’re writing. What it sendsYour survey’s title, description and questions. No answers. Keep in mindNothing is added unless you add it.

Response themes (“Find themes”)

 What it doesSuggests themes in the written answers to one question, and shows how many answers fall under each. What it sendsYour respondents’ written answers to that question: all of them when there are 384 or fewer; above that, a random sample of 384, and Candor says so on screen. No names or email addresses are attached, but anything a respondent typed is sent as written. Keep in mindSuggested themes can be wrong. Check them against the answers.

Results summary (“Summarize all”)

 What it doesWrites a plain-language summary of your results. What it sendsYour questions, response counts and per-question summaries. Not the answers people wrote. Keep in mindIt works from counts, not from what people wrote.

Question interpretation (“AI perspective”)

 What it doesExplains what one question’s results suggest. What it sendsThat question’s wording, counts and statistics. Keep in mindIt explains the numbers on screen; check it against the chart.

Cross-tab explanation

 What it doesExplains how the answers to two questions relate. What it sendsBoth questions’ wording and a table of counts. Keep in mindCheck it against the table.

A second opinion, before you publish

 What it doesReviews your survey before you publish, if you ask. It never stops you from publishing. What it sendsYour survey’s questions. Keep in mindIt comes with the main AI switch.

What the AI writes comes back in your language.

What stays the same with AI off?

  • The checks that can stop a survey from publishing are fixed rules, not AI.
  • “Tell me what you need” matches what you tell it to our templates. It isn’t AI.
  • With AI off, the guide beside you as you write still introduces your template and helps you try your survey on a few people first. Its questions about who you’re surveying, and its question ideas, appear only with AI on.

Where does it go, and what’s kept?

  • Candor’s AI features send what each one needs to Anthropic, the makers of Claude.
  • Candor saves each AI result for up to 24 hours so it can show it again, then it expires and is deleted. Candor doesn’t use your surveys or your respondents’ answers to train or tune any AI model, and each AI request starts from the same fixed instructions with only that survey’s data.
  • How long Anthropic keeps what it receives: see Anthropic’s data retention policy.
  • Candor also uses AI to translate its own templates and app text into other languages. None of your surveys or answers are involved.

Facts last checked: September 27, 2026.

For researchers and reviewers: answers for your ethics or procurement form

If a form asks about Candor’s AI, here are the answers. Where the answer is yours to decide, we say so. Facts last checked: September 27, 2026.

What’s sent, and to whom?
The table above lists what each feature sends. It goes to the provider named above.
How long is it kept?
Candor keeps each AI result for up to 24 hours; after that it expires and is deleted. Your survey and its responses aren’t affected. For the provider, see its retention policy.
Is it trained on, or does it adapt to, what participants write?
No. Candor doesn’t use your surveys or your respondents’ answers to train or tune any AI model, and each request starts from the same fixed instructions.
Are personal details removed before sending?
No. Names and email addresses aren’t attached, but anything a respondent typed into an answer is sent as written.
Who can turn AI off?
Each person, in their own account. There’s no organization-wide switch, and one person’s setting doesn’t apply to anyone else’s account.
Does turning AI off stop requests on Candor’s servers?
No. Turning AI off hides the AI features in Candor, but Candor’s servers don’t refuse AI requests from an account that has AI off.
Are respondents told?
Candor doesn’t tell your respondents that AI may read their written answers. If you use Response themes, whether and how to tell them is yours to decide. Some ethics boards require it, and AAPOR’s 2026 task-force report recommends it.
Is Candor approved for my data?
That’s your institution’s decision; Candor can’t make it for you.

More for ethics submissions: Research Mode and the supporting documents.

The longer view: “Should you trust AI with your survey?” Questions about Candor’s AI? Ask us — a person writes back.

Start your first survey free No credit card required

Candor Surveys℠ by Golden Hills Software, Inc. ©