> ## Documentation Index
> Fetch the complete documentation index at: https://docs.citino.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# How answers are collected

> How Citino asks each question, what it stores from every answer, what it reads out of it, and how the questions themselves are built.

*Version 1 · 7 November 2026*

Every number in Citino starts as an AI engine's answer to a question a buyer would ask. This page sets out how we ask, what we keep and what we read out of each answer.

## What we ask

Each project has its own questions: the ones you wrote, or the ones we suggested from your website at sign-up and you edited on Review. Every day, we send each question, word for word, to every engine the project tracks, from the project's market. On an engine we reach through its API, the question is preceded by a one-line note naming the reader's market. Today that market is the United States, in English.

A question's wording never changes once it has been asked. To ask it differently you replace it, through the [API](https://docs.citino.ai/api/writes.md), and the new wording starts a new history. As the app puts it, "A question is never edited, only replaced, so its history stays comparable."

## How an answer reaches us

* **Signed out.** Every answer is collected signed out. There is no account, no chat history and no memory behind it, so nothing about a past conversation shapes it.
* **From the surface your buyers use, where we can.** Each answer comes either from the engine's public web page, as a signed-out visitor sees it, or from the engine's documented API with web search switched on. Every stored answer records which of the two it came from.
* **From your market.** Questions are asked from the project's country and language. On an engine reached through its API, that one-line note is how the market is given.

## What we keep from every answer

* The full answer, exactly as the engine gave it.
* The pages the engine says it read while answering, and the ones it cited, where the engine reports them.
* The searches it ran, where the engine reports them.
* Which engine answered, its version, the language and market, and whether web search was on.

Stored answers are never edited or deleted. You can read every one on the **Answers** screen.

## What we read out of an answer

A separate reading step goes through each stored answer and records:

* **Which brands it names**, and in what order.
* **How favourably** it describes each one.
* **What it says about the claims you track**, with the sentence it said it in.

Two rules keep this honest:

* **A brand counts only if its name is in the answer.** Nothing the reading step infers becomes a mention.
* **Sources come from the engine, not from the reading step.** Which pages an answer read and cited is what the engine itself reported.

Because the answers are kept, they can be read again. When we improve the reading step, or you add another name for your brand in Project settings, the stored answers are re-read. As Project settings says, "Re-scanning stored answers for this alias costs no engine calls and uses no prompts."

## What counts as an answer

| What happened | Counted as an answer? |
| - | - |
| The engine answered | Yes |
| The engine declined to answer | Yes. It is an answer that named nobody. |
| The call failed and never succeeded that day | No. The day is recorded as missed. |
| Google showed no AI Overview for the search | No. The surface showed nothing to read. |

## Which questions your headline numbers count

Visibility, share of voice, position and sentiment count your **unbranded** questions, the ones that name no brand, in the segments you have claimed. A question that names your brand would name it nearly every time and swamp the average, so questions that name a brand are kept apart from your headline numbers. So are the questions Citino asks about your brand for [Objections](https://docs.citino.ai/guides/screens/objections.md).

## How questions are built

We publish the rules our suggested questions follow, not anyone's question list. A published list would invite people to optimise for the exact strings instead of the questions buyers ask.

When we suggest questions for a new project, each one:

* is something a real buyer would plausibly type, in their own everyday words;
* describes the buyer's situation, often with the one constraint that matters to them, such as their size, budget or what they use today;
* is answered naturally by naming products or companies, rather than by an explanation;
* names no company, product, brand or website;
* asks one thing;
* does not repeat another question in different words;
* avoids keyword-list phrasing such as "top 10" or a year.

Together, the questions vary in length and cover the moments a buyer goes through: just starting, a specific problem, choosing between approaches, cost and trust. Each sits in a segment, one job buyers hire these products for. You edit all of it before the first collection, and you can add questions at any time, or replace them through the [API](https://docs.citino.ai/api/writes.md).

### The questions Citino asks about your brand

For projects in an English-language market, Citino also asks five questions of its own about your brand each week, to find the objections engines raise. With *Acme* as the brand, they read:

* Why might someone decide not to choose Acme?
* What are the main drawbacks of Acme?
* What do customers complain about with Acme?
* Why do people switch away from Acme?
* What should I know before choosing Acme?

They don't count toward your plan, they are never in your headline numbers, and you can pause them in Project settings.

## Related

> [card] [Cadence](https://docs.citino.ai/guides/methodology/cadence.md)
>
> When each question is asked.

> [card] [How bands are computed](https://docs.citino.ai/guides/methodology/bands.md)
>
> How sure each number is.

> [card] [What we do not measure](https://docs.citino.ai/guides/methodology/what-we-do-not-measure.md)
>
> The limits of this method.

> [card] [Engines and opt-ins](https://docs.citino.ai/guides/account/engines.md)
>
> Which engines a project is asked on.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.