Transparency

How we measure and create tasks

We ask AI assistants the questions your customers ask, and from their answers and your site we pick what to improve. Here are the principles we follow, plus one example from measurement to a finished task.

Walk through the example
What people are askingIllustrative example, made-up data

In the Tents category, assistants didn’t mention you in a single one of 20 answers

What you do

Write answers to the 5 most common customer questions into the Tents category description, with concrete numbers, not generalities.

How we measure

We ask what your customers ask

We put the questions to ChatGPT, Gemini, Google AI Mode, Perplexity and, depending on your plan, Claude.

You can read every answer

For every answer we work out whether the assistant mentioned you. You can open it in full in the app and check it yourself.

The score is a plain share

Your AI score is the share of answers that mention you. Where the assistant ranked you is shown separately.

Principles

How we create tasks

  1. 1

    Fixed rules, same tasks

    Fixed rules turn what measurements and site checks found into tasks. So the same results always give the same tasks.

  2. 2

    Too few answers, no task

    A rule based on assistant answers only decides once it has enough of them. Until then, it stays silent. But silence doesn’t mean everything is fine.

  3. 3

    You see why a task is there

    Every task says what we found and why it’s worth doing. You don’t have to guess where it came from.

  4. 4

    One finding, one task

    A finding you already have as a task, even a snoozed or dismissed one, won’t create another one.

  5. 5

    Most worthwhile on top

    The order is our estimate of what’s most worthwhile for you. It’s an estimate, not a promise.

  6. 6

    Only a few at a time

    Your week holds only a few tasks at a time. The rest wait in the queue, and finishing, snoozing or dismissing a task frees up a spot for the next one.

An example of how we recommend tasks

  1. 1

    A measurement finds something

    We asked the assistants several questions about the Tents category and got 20 answers. Not one of them mentioned the made-up store.

  2. 2

    A rule turns it into a task

    There are enough answers and the store is missing from all of them, so the rule creates a task. Its title says what we found and how many answers it’s based on:

    Task title

    In the Tents category, assistants didn’t mention you in a single one of 20 answers

  3. 3

    You get one clear action

    Under the title is one concrete action, followed by step-by-step instructions and how you’ll know it’s done. The action reads:

    What you do

    Write answers to the 5 most common customer questions into the Tents category description, with concrete numbers, not generalities.

  4. 4

    A measurement then shows whether the finding is still there

    Once you mark the task done, we look at it in one of the later measurements. The task itself tells you whether you need to run that measurement yourself. A measurement can say only one of two things: “The finding is gone” or “The finding is still there”. Neither claims your change caused it. If a measurement can’t tell, we don’t claim anything.

What we don’t promise

  • We don’t promise you a position in AI answers or on Google. Nobody on the outside can guarantee that.
  • We don’t promise how many visits or orders a change will bring you, and we don’t take credit for your work.
  • We don’t change anything on your site. You make the change, or whoever you send the brief to.

You can see where Custela itself uses AI on our AI transparency page.

FAQ

Frequently asked questions

Why didn’t I get new tasks after a measurement?

The measurement found nothing, or a rule didn’t have enough answers to go on. A finding you already have as a task, even a snoozed or dismissed one, won’t create a new task. And when your week is full, a new task waits in the queue.

Why don’t you publish every rule and number?

Because this page shows how the app works, not the detailed recipe. The rules and their thresholds are our know-how, and the app is built on them. What you need for your own decisions is in the app: the title of a task that comes from assistant answers says how large a sample it’s based on, and you can open every answer of a measurement in full under AI visibility.

See how it works out for your own site

Start 7-day free trial

Once per account, and you enter a card to start. If you cancel the subscription during the trial, you pay nothing. Otherwise you start paying for Profi, 1 490 Kč a month excl. VAT.