> ## Documentation Index > Fetch the complete documentation index at: https://docs.voxworks.ai/llms.txt > Use this file to discover all available pages before exploring further. # FAQs > The FAQs fixed section: answers to what respondents ask during a survey call. Each FAQ's fields, the 41–99 priority band, the catch-all, asked-again lines, Always fires, the FAQ step levels, and a starter set including 'are you an AI?'. ## What are survey FAQs? Respondents ask things: *"Who is this?"*, *"How long will it take?"*, *"Where did you get my number?"*. The **FAQs** fixed section is where you write the answers. It is described as *"What the respondent may ask about the call, the study or the interviewer."* FAQs are live for the whole call — at the consent question and during every question. When a respondent asks one, the interviewer speaks your answer, then goes back to the question that was open and waits for the reply. The same FAQ can be answered as many times as it is asked. With none written, the pane warns: *"No FAQs authored. Without them the interviewer has no answer for “who is calling?” or “how long will this take?”."* Open **FAQs** in the **Outline** on the [Questions tab](/surveys/questions-tab), click **+ Add FAQ**, fill in the fields and click **Save**. The confirmation is *FAQs saved*. A survey can hold up to 50 FAQs. *** ## FAQ fields | Field | What it holds | | ------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Key** | A short name for the FAQ — lower case letters, numbers and underscores, up to 64 characters (*"FAQ keys are lower case letters, numbers or underscores."*). New FAQs start as `faq_1`, `faq_2`, and so on. | | **Priority** | Where the FAQ ranks against other routes — see [Priorities](#priorities). | | **Accept** | *"What counts as this — one condition per line, in the interviewer's words."* | | **Often heard as** | *"Recogniser mishears only. These are never seeded as keyterms."* | | **Reject** | *"What must not match here, and which route it belongs to instead."* | | **Answer** | Spoken exactly as written. Up to 10,000 characters. | | **Asked again** | *"Verbatim lines for a second ask, one picked at random. Leave empty to repeat the answer."* | | **Always fires** | *"Skips the check that decides whether the caller was answering the question instead. For a route where a miss costs more than a wrong fire."* | ### Writing Accept lines Accept lines describe what the respondent means, not an exact phrase to match. Write one condition per line, in plain words: * *Asks how long the survey will take* * *Asks how many questions are left* The interviewer matches the meaning, so *"is this going to take long?"* and *"how much longer?"* both fire the second line. ### Often heard as and Reject Leave these empty to begin with. Add to them only when you see a problem on a real call: * **Often heard as** — when the recogniser writes a question down wrongly and the FAQ misses it. Never add mishearings as [keyterms](/surveys/agent#keyterms). * **Reject** — when the FAQ fires on something it should not. Say what it is and where it belongs, for example *"Asking for the question to be read again — that belongs to the question"*. ### Answer and Asked again Write answers as one or two spoken sentences. The answer is said over a question that is still open, so keep it short and let the interviewer hand straight back. If a respondent asks the same thing twice, a line from **Asked again** is used instead of repeating the answer word for word. Two or three variations sound more natural than one. ### Always fires Every FAQ normally goes through a check first: when a reply could be either an answer to the current question or an FAQ, the interviewer decides which the respondent meant, and treats it as an answer by default. **Always fires** skips that check for this FAQ, so it answers whenever its Accept lines match. Turn it on only for an FAQ where missing the question is worse than answering it by mistake. See [Answers given early and interruptions](/surveys/riders-and-interruptions). *** ## Priorities Every FAQ has a **Priority** from 41 to 99. A lower number ranks higher: when two FAQs match the same words, the lower number wins. **+ Add FAQ** gives each new FAQ the next free number. * **No two FAQs may share a number.** The field warns *"`n` is used by another FAQ — the compiler refuses a duplicate."* * **Every specific FAQ must sit below the catch-all.** The field warns *"`n` is at or above the catch-all at `c` — the catch-all would answer instead. Give this one a lower number."* When every number below the catch-all is used, **+ Add FAQ** is disabled with *"Every FAQ priority is taken"*. FAQs rank below the survey's [Behaviours](/surveys/behaviours) (busy at 20 and stop at 30 by default), and above the questions' own handling of *"what does that mean?"* and *"can you read the options again?"*. An FAQ worded too broadly can therefore steal those requests from the question — use **Reject** to hand them back. *** ## The catch-all Below the specific FAQs is the **Catch-all** group: *"Answers any other question about the call, the study or the interviewer that no FAQ above covers. A specific FAQ that also matches always wins. Left empty, unexpected questions fall through to the question being asked."* The catch-all is the FAQ with the key `other`. It sits at priority 99 on new surveys; surveys created before 23 September 2026 keep it at 49, and new FAQs on those surveys must stay below 49. It has its own **Answer**, **Asked again** and **Always fires**. Write the catch-all answer as a polite deflection that steers back to the survey, for example *"That's a good question, but I'm not able to help with that on this call. Shall we carry on?"* Give it two or three **Asked again** lines, and add **Reject** lines so it does not swallow *"can you say that again?"* or *"what do you mean?"* — those belong to the question. *** ## FAQ step levels The FAQ pane ends with its own **Step levels** group: *"The gates every FAQ answer above is spoken under — its own, not the survey's. An answer is one line said over a question that is still open, so it waits briefly and hands straight back."* | Level | Unset here | | ------------- | --------------------------------------------------------------------------- | | **Eagerness** | High | | **Idle** | Low | | **Thinking** | The survey's own level from the [Agent](/surveys/agent#step-levels) section | Each can be set to **Survey default**, **Low**, **Medium** or **High**. These defaults mean a survey slowed down for thoughtful answers does not also slow down a one-line FAQ answer. See [Attempts and step levels](/surveys/attempts-and-step-levels). *** ## A starter set Most surveys need answers to the same handful of questions. A typical set: | Key | Accept | Example answer | | ------------------- | -------------------------------------------------------------------- | ----------------------------------------------------------------------------------- | | `how_long` | Asks how long the survey takes, or how much is left | *"It takes about ten minutes, and we're already part way through."* | | `who_is_calling` | Asks who is calling or which company this is | *"I'm Alex, calling from Northside Research, an independent research company."* | | `what_is_it_about` | Asks what the survey is about or who it is for | *"It's a survey about local community services, run on behalf of the council."* | | `where_number` | Asks where we got their number | *"Your number was chosen at random from Australian phone numbers."* | | `privacy` | Asks how their answers will be used, or about privacy | *"Your answers are combined with everyone else's and never reported on their own."* | | `are_you_ai` | Asks whether they are talking to a real person, a recording or an AI | Your organisation's own disclosure wording. | | `other` (catch-all) | — | A polite deflection back to the survey. | "Are you a real person?" and "Is this an AI?" have **no built-in answer**. If you do not write an FAQ for them, the question falls to the catch-all or back to the question being asked. Write the answer your organisation is comfortable with. Replace the example answers with your own facts — the interviewer says exactly what you write.