AI Logic.

AI Logic automatically generates survey logic from your brief or prompts, reducing the need for manual setup.

When to use this

Logic helps you skip irrelevant questions, personalise the survey flow, and screen out respondents. AI Logic lets you generate and apply this logic automatically based on your research goals, without manually setting up rules.

How AI Logic works

AI Logic applies survey logic in three ways:

  1. Inferred from your brief
  2. Specified in your prompt
  3. Imported from Word or PDF

Inferring logic from your brief

AI Logic can infer where logic should be added, even if you do not explicitly describe it.

For example, if your brief includes a screener section and your topic is ice cream, AI Logic may recognise common research patterns and apply logic where it would typically be expected—such as exiting respondents who do not select “ice cream” at Q1.

Click image and press Ctrl, Ctrl to magnify.

In another example, AI Logic may add display logic to follow-up questions when a respondent meets a condition (e.g. showing Q13–Q14 only if Q12 = Yes), even if that instruction is not explicitly written in the brief.

Click image and press Ctrl, Ctrl to magnify.

What to expect

Because AI Logic is inferring intent:

  • It will usually apply the correct logic
  • It may occasionally miss or misinterpret something

When that happens, you can correct it by providing a clear, explicit follow-up prompt describing the logic you want.

Viewing applied logic

Questions with logic applied show a branching-arrow icon.

Expand the question to view the logic underneath, highlighted with an orange ribbon.

Specifying logic in your prompt

AI Logic works best when you give it explicit instructions.

Here are some example prompts you can use:

  • Exit if ‘ice cream’ is not selected at Q1.
  • Display Q13–Q14 if the respondent bought AFL merchandise in P12M at Q12.
  • Show Q24 if ‘Brand N’ was selected as ‘ever bought’ at Q22, but not as ‘bought in P3M’ at Q23

  • Add least-fill logic at S2 so only 1 of 6 prices is shown.
  • Only display questions Q4–Q9 when the respondent has the correct tag for each question.
  • If the respondent is unlikely to buy at Q4, jump to S3.

  • Remove the ‘jump to’ logic from Q4.
  • Quality exit if the respondent selects a fake brand at Q10.

Importing logic from Word or PDF

If your survey document includes logic instructions, Glow imports both the questions and any supported logic. This allows you to bring existing surveys into Glow without needing to rebuild logic manually.

Supported logic types

AI Logic supports the full range of logic types available in Glow, with one exception.

Display logic

Show questions, statements, choices, or variables based on conditions

Action logic

Submit, exit, quality exit, jump to a question or section, add or remove tags, retain least-filled tags, set variables, or add to a list

Conditions:

Based on whether a respondent:

  • selected an option
  • was shown an option
  • answered a question
  • has a tag
  • and more

For more detail, see the Logic Editor help section.

Limitations and when to use manual logic

AI Logic is powerful, but there are a few limitations to be aware of:

  • AI Logic does not currently support Hide Groups (hiding an entire group of questions)
  • Because AI Logic may infer intent, it can occasionally miss edge cases.

Use the Logic Editor directly when you need precise control over complex or edge-case conditions.

Reverting to a previous version

If you make a mistake while adding logic, you can revert to an earlier version of your survey.

To do this, click Load on any blue message bubble in the conversation history on the left.

Your next prompt will use the version of the survey you most recently loaded.

AI Logic Automation

AI Logic can generate tens, hundreds, or even thousands of lines of repetitive but predictable logic. Tasks that once took hours or even days to set up manually can be completed in minutes with the right prompts. This is especially useful for complex survey types where large volumes of structured logic are required.

See: AI Logic Automation [WIP]

Did this answer your question? Thanks for the feedback There was a problem submitting your feedback. Please try again later.