> ## Documentation Index > Fetch the complete documentation index at: https://docs.recepai.ai/llms.txt > Use this file to discover all available pages before exploring further. # How written answers are read > What happens to the words a guest types into a survey — how they are grouped, counted, quoted, and when they alert your team. A **Written answer** question, and the *Ask for details* question under a low score, give you the guest's own words. Star reads every one of them a few minutes after it arrives, so a guest never waits on the reading, and turns them into the **What guests wrote** block on [Results](/front-desk/reviews/surveys-results). This page explains what that reading does — and what it does not. *** ## One answer, read on its own Every written answer is read separately, in the language the guest wrote it. The reading names each thing the guest talked about — a *subject* — and how they felt about it. One answer can carry several subjects: a sentence praising breakfast and complaining about the noise is two subjects, with two different readings. ## A fixed list of subjects Subjects are matched to **one fixed list of hotel topics** — 69 subjects in 11 areas — so a towel complaint written in Russian and one written in German count as the same subject: | Area | Examples of its subjects | | ------------------------- | ------------------------------------------------------------------------------------------------- | | Housekeeping | Room not cleaned · Towels or linen · Bathroom hygiene · Bad smell | | Maintenance and technical | Air conditioning or heating · Hot water or water pressure · Repair not done or slow | | The room | Bed, mattress or pillows · Noise · View · Room not as booked | | Food and drink | Breakfast variety · Breakfast quality · Restaurant or bar service · Room service | | Reception and arrival | Check-in waiting time · Reception attitude · Bill, deposit or charges · Complaint not followed up | | Staff, in general | Staff friendliness · Staff competence · Language barrier | | Facilities | Pool · Beach · Spa and wellness · Wi-Fi · Parking · Accessibility | | Price | Value for money · Unexpected extra charges | | Location | Location · Getting there | | Safety and health | Safety and security · Insects or pests · Illness during the stay | | The stay as a whole | The stay in general · Would come back or recommend · Not as advertised | When the words say something the list has no name for — *a bike rental would be great* — the subject is shown as **Other**, with the words the reading gave it, rather than forced into the nearest topic. ## Four readings Each subject carries one of four readings: **positive**, **negative**, **neutral**, or **mixed**. Only a negative reading counts as a complaint. A wish, a remark, or praise is kept and shown, but it is never treated as a problem. ## The guest's own words Every subject carries a **quote** — the guest's own text, checked against the answer before it is stored. A subject whose quoted words cannot be found in the answer is dropped, and the screen says how many were dropped that way. What you read under a subject is always what the guest wrote, never a summary. ## Which department A subject is placed in a department by the guest's own words alone. The department of the question they were answering does not decide it — a guest can complain about the restaurant inside a housekeeping question, and the complaint goes to the restaurant. When the words do not say which department a subject belongs to, it is shown under **Not placed in a department** — counted there rather than guessed into one. ## How the block counts * **Responses** counts answered surveys — one per stay. A guest who stayed twice and answered twice is two. It never counts sentences. * **Mentions** counts how many times a subject came up across those responses. * **Was N** is how many responses named the subject in the period of the same length before this one. It is shown only when both periods have been read in full; otherwise the screen says why it is not compared. * The line beside the heading says how many written answers were **read**, how many are **still waiting to be read**, how many **said nothing about the stay**, and how many **could not be read**. Every count in the block is over the answers that were read. * Under each department, subjects several guests named come first; subjects one guest named are folded under *more subjects, each from one response*. Two sentences at the top of the block are chosen for you: **Most complained about** and **Most praised** — the subject the most responses complained about, and the one the most responses praised, each with last period's number when the periods can be compared. ## When an alert fires A low **score** alerts your team on its own — the overall question, or any department question at or below your follow-up threshold. Written answers add a second road: a **negative subject with the guest's own words** alerts the same people, in the same way, even when the guest gave five stars — *the air conditioning has been broken for three days* under a 5-star answer reaches your team tonight. * Only complaints alert anyone. Praise, neutral remarks and wishes never do. * The alert shows the guest's own words, with the subject and the department. * One alert per answer. The rating is never changed by the reading. * Feedback your team has already acted on, or that a low score already alerted, is left alone — the words were on that alert. Who is alerted, and on which channels, is set under [Alerts](/front-desk/settings/reviews) — the same people who hear about a low rating. ## What it is not This is a count of what was said, not a measure of what matters most. A subject named by six responses is a subject to look at; it is not proof that it decided anyone's rating. The same reading of a guest's words also feeds [Guest Pulse](/front-desk/reviews/analytics) once, alongside your published reviews, so a survey answer is never counted there twice.