# Questions, answers and engines

> How Prefera tracks your brand: the questions it asks, the answers it saves, the engines it uses, when it checks, which countries and which competitors.

Source: https://prefera.si/docs/concepts/tracking
Docs index for agents: https://prefera.si/llms.txt

Prefera asks buyer questions to AI engines, saves every answer and shows you where your brand is named. This page explains each part. Limits and prices depend on your plan: see [Plans, add-ons and billing](/docs/account/plans). How your answers are spent (daily, weekly or a mix) is on [How tracking works](/docs/concepts/how-tracking-works).

## Questions

A question is something a buyer would ask, for example "best invoicing tool for freelancers". Each question has a buyer stage (Discovery, Comparison, Decision or How-to), a buying intent and a country.

### How questions get added

- **During onboarding.** You pick from candidate questions. See the [Quickstart](/docs/quickstart).
- **Add question.** On the **Question map** screen (`/app/questions`), press **Add question**. Type one question or paste several, one per line. **Buyer stage** can be left on "Pick for me". Prefera checks each line and says why it refuses one.
- **Import CSV.** On the same screen, import a file with one question per line.
- **Discover questions.** This button searches Google and Search Console for real searches and adds them as suggestions. It can take a minute or two.
- **Suggested questions.** The panel beside the list holds suggestions. Press **Track this** to start tracking one, or dismiss it.

Each question shows where it came from, for example "Added by you" or "From Search Console".

### Status

Each question shows a status taken from recent runs.

| Status | Meaning |
| --- | --- |
| Won | Your brand is named in more than half of the valid answers. |
| Contested | You are named in some answers but not most. |
| Winnable | Most answers do not name you yet, but AI cites pages for it. |
| Dead | Two runs in a row with no answer naming you and no page cited. |
| New | No finished run has asked it yet. |
| Paused | Archived: kept, but not asked on runs. |

### Archive, restore and pin

Open the actions menu (the three dots, **Question actions**) on a row.

- **Archive** stops the question being asked. It stays saved and shows when you press **Show paused**.
- **Restore** brings an archived question back. It needs a free question slot. If there is none, archive another question first.
- **Check this instead** appears on a question marked **Not checked** because you are over your plan's question limit. It pins the question so runs ask it first. A pinned question shows a **Pinned** tag.
- **Let intent decide** removes the pin.
- **Watch daily** moves a question into your daily checks. See [Schedule](#schedule).

You can also change a question's stage or country from this menu.

> [!NOTE]
> Only editors and owners can change questions. Viewers can look but not edit.

## Answers

A check asks each question on each engine. A weekly check asks 1 to 3 times per engine, because AI answers vary from ask to ask; three answers give a steadier number, and 1 makes proof weaker. A daily check asks once per engine each day, and your numbers count its last 7 days. Each ask is a **sample**, and every one counts as one of your AI answers.

For every sample Prefera saves:

- the answer text;
- whether your brand is named;
- who else is named, including your competitors;
- the pages the engine cited.

Open **AI answers** to read them. Pick a question, then an engine tab and a sample. The panel on the right (**Sources the engines used**) lists the cited pages. Older runs are in the run menu at the top.

## Engines

Open **Tracking** in the sidebar. The **Engines** card has one switch per engine.

| Engine | How Prefera reads it |
| --- | --- |
| ChatGPT | Read from chatgpt.com, logged out |
| AI Overviews | Read from a live Google search page, logged out |
| AI Mode | Read from Google AI Mode, logged out |
| Gemini | Gemini API with Google Search grounding on |
| Perplexity | Perplexity API (Sonar) with web search on |
| Claude (add-on) | Claude API with web search on |

Every plan includes the first five. Claude is the **Claude add-on** on every plan: its row shows **Add it** with the price until you turn it on in **Team and billing**. With **Claude on the questions you choose**, the workspace has a monthly Claude answer budget (1,300 Claude answers on Growth: about 100 questions weekly × 3, 300 weekly × 1, or 42 daily). In **Tracking settings**, under **Which questions get Claude**, pick **Auto** (Prefera picks your most important questions weekly × 3) or **Choose**: tick questions from every brand, set each to **Daily** or **Weekly × 1 / × 2 / × 3**, and the meter shows whether it fits before you save. Daily gets 1 Claude answer every day; weekly × 3 gets 3 answers on each weekly run. On **Questions**, editors switch the **Claude** tag on a question (weekly × 3) or off. A member limited to some brands chooses for those brands only, inside the same budget; viewers see the picks. Removing the add-on, or moving to the every-question tier, clears the picks at that point; with **Claude on every question**, it answers every question you track on every run. Either way it is on top of your AI answers.

To turn an engine off, switch it off and press **Save changes**. The screen shows the effect before you save. Every answer counts as one of your AI answers whichever engine gave it, so a brand on fewer engines can check more questions daily or twice a week. Your question limit stays the same.

## Schedule

On **Tracking** you choose how often each brand's questions are checked: the recommended mix, daily, weekly with 1 to 3 answers, or twice weekly with 3 answers. Every paid plan can use every mode; your AI answers decide how many questions each allows, and the screen shows the count before you save. See [How tracking works](/docs/concepts/how-tracking-works). Each brand has its own start hour between 06:00 and 17:00 UTC. Scheduled results arrive within an hour of the start.

To choose the weekday and time of day, use **When checks run**: pick the day (or two days when the brand runs twice a week), the **Time of day** and the **Time zone**, then press **Save schedule**. **Use the default** goes back to the default schedule.

To run everything again now, press **Run now** on **AI answers** and confirm.

### Daily checks

A daily question gets one answer per engine every day, counted over the last 7 days. In the mix you say how many questions are daily; in the daily mode every question is. Prefera picks the daily questions once a day, for the next day, in this order: questions with a fix that just went live, questions you watch, questions that just moved, then your highest-priority questions. If you would rather pick them yourself, mark them **Watch daily** and choose to pick them yourself.

## Countries

Each question is asked as if from one country, so engines answer with that country's results, prices and brands. Questions without a country of their own follow your **Main market**. Change it in **Tracking** under **What we ask**.

To track another country:

1. On the **Question map**, press **Add a country**.
2. Pick the country and press **Suggest questions**. Prefera writes questions for it, including People Also Ask from there.
3. Tick the ones to keep and press **Add N to country**.

Adding a country does not raise your question total. Each country gets a share, which starts equal and which you can edit in the same panel. You can also move one question with **Change country** in its menu, or tick several and use **Move to country**. When a brand has more than one country, a country switch appears on the answers screens.

## Competitors

Open **Competitors** in the sidebar. It shows who answers your questions: questions led, change since the last run, how many answers name each one and each competitor's top winning page.

- **Add competitor**: enter a **Name** and **Domain**, and optionally **Aliases** (comma-separated). Changes apply from the next run.
- **Aliases** are other names for the same competitor, for example "Acme Inc". Prefera counts a mention of an alias as a mention of the competitor. Press the aliases text on a row to edit it.
- **Track** next to a suggestion adds a competitor that AI engines recommend but you do not track yet. If no website is known, Prefera looks for one and asks you to check it.
- **Remove** appears on a competitor not seen in AI answers.

The number of competitors you can track is shown on the screen as "N of M competitors tracked".