Popular searches

Request consultation

Do you need support in running a pricing or product study? We can help you with agile consumer research and conjoint analysis.

Looking for an online survey platform?

Conjointly offers a great survey tool with multiple question types, randomisation blocks, and multilingual support. The Basic tier is always free.

Setting up a study on Conjointly
List of articles Back

Audit your survey for free with AI Helper

AI Helper relies on artificial intelligence and may make occasional errors. Please review the outputs before applying them.

Catching survey design flaws before launch prevents unusable datasets and eliminates unnecessary re-fielding costs. AI Helper’s survey review audits your survey script within minutes, flagging typos, broken scales, missing answer options, biased questions and more.

It is free on every licence tier with no setup required. The feature is unavailable if your organisation has disabled LLM features.

What a survey review typically catches

The examples below are taken from an AI Helper review of a draft “Consumer survey on functional foods”. Your review follows the same layout, though the specific categories and flagged issues will vary based on your script.

1. Typos and grammatical errors

Catches spelling and grammar slips in question text, answer options and labels.

AI Helper flagging a misspelt brand name, a missing letter in None of the above, and US spelling in an Australian English survey

2. MECE violations

Identifies answer options that are missing, overlapping, or leave respondents no valid choice.

AI Helper flagging a missing 55 to 64 age band, overlapping frequency options, and a missing escape option for respondents with no valid answer

3. Questionnaire flow issues

Detects misplaced questions, along with missing skip logic or display gates.

AI Helper flagging demographics placed mid-survey, a broken topic sequence, and missing display logic for non-users

4. Inconsistencies

Flags inconsistencies in content, formatting, or logic across questions.

AI Helper flagging brand images used in one question but not another and inconsistent product terminology across questions

5. Methodological issues

Identifies potential methodology issues, such as leading questions, biased phrasing, and mismatched scales.

AI Helper flagging brand images as recognition cues in an awareness question, a yes or no question forced onto a Likert scale, and a leading question

6. Response quality safeguards

Recommends screening questions, along with quality management options and settings that protect your data.

AI Helper suggesting a longer minimum answer length for an open-ended question, a decoy brand to catch overclaiming, and a consumption volume question

You can also work with AI Helper to fix what it finds.

Running a survey review

  1. Start the review from your survey setup page in one of two ways:

    • Click Yes on the prompt in the bottom-left corner (“Would you like AI Helper to review your experiment now?”). The analysis begins immediately.
    Bottom-left prompt asking to review the experiment
    • Or click the AI Helper button in the top-right corner, then Start review in the chat panel.
    Clicking Start review in the AI Helper chat panel

  2. AI Helper automatically submits a prompt asking for typos, inconsistencies, errors and opportunities for improvement, then returns its findings grouped by category.

  3. Review the findings, ask follow-up questions in the chat, and run another review once you have made your changes.

Triaging the suggestions

The survey script review functionality is built to look for problems, so it will always come back with something. Please read each finding as a question worth asking rather than an instruction to follow, and expect the wording to differ if you run the same script again.

It is worth sorting the findings before you change anything. Three groups usually cover them:

  1. Fix straight away. Genuine typos, missing answer options and duplicated questions are unambiguous, and there is rarely a reason not to act on them.
  2. Treat as a judgement call. Suggestions about question order, survey length, screening or methodology depend on your research objectives, your sample plan and what you have fielded before. AI Helper cannot see any of that, so weigh them against your own design.
  3. Set aside. Some findings will conflict with a choice you made deliberately, or repeat advice you have already applied. Leave those where they are.

Trackers deserve extra care because changing question wording can break comparability with your earlier waves.

Fixing issues with AI Helper

If you would like AI Helper to apply changes automatically, follow these steps:

  1. Select or tick the findings you wish to resolve.
  2. Ask AI Helper in the chat window to address the selected issues.
  3. Carefully review the proposed edits. A suggested fix can be technically correct while remaining unsuitable for your research objectives.
    User requests AI Helper to fix the flagged issues
  4. Click Confirm to apply the changes to your survey.
AI Helper proposing survey edits with a Confirm button to apply them

Getting the most out of AI Helper

  • Share your current screen. The camera icon in the chat toolbar sends AI Helper what you are looking at, and the marker icon lets you draw on it to point at the exact question or setting you mean. Both are quicker than describing the problem in words.

    Pressing the camera icon to attach the current screen

  • Ask what it can do for you. Type “What can you help me with?” at any time. AI Helper is continuously updated with new capabilities between releases, so supported features and assistance options expand regularly.

What the suggestions are based on

AI Helper’s review applies established questionnaire design principles. The guides below cover the same ground, and knowing them makes its findings easier to judge and easier to overrule when your study calls for something different.

Have questions about optimising your questionnaire design? Schedule a consultation with a research expert.