promptrack
Getting started

Quick start

Four things to do on your first day. Each one takes a few minutes, and none of them needs you to understand the product first. The guides are there when you want the why.

Read time 4 min · Doing time 10 min

1. Add your brand

Your task: give one domain and check what comes back.

The domain is the only thing you have to look up. From it we read your site to work out what you sell and to whom, then pre-fill your brand name and the market your buyers are in, and propose the brands you compete with. Correct anything that looks off before moving on: those three fields travel into every query afterwards.

i
The competitors we suggest are not the ones you would name from memory. They are the domains that show up alongside you on your own topics. Expect a surprise or two, and keep the ones that are really competing for the same answer.

2. Check your questions

Your task: read the proposed list, delete what does not fit, add what is missing.

This is the step worth five real minutes, because this list decides every number you will read afterwards. A weak set produces numbers that look fine and mean nothing.

Two things to look for as you read:

KeepDelete
Questions a buyer asks before choosing: best X for Y, X against Z, where to find XAnything a definition would answer. Those replies explain a concept and never name a brand.
Questions with a constraint in them, which is what makes an answer name specific brandsAnything that already contains your own name. The point is to see whether the engine brings you up on its own.

You can track 200 active questions across your whole workspace, so there is room. Spend it on covering your buying themes before piling up variations of the same one.

Guide 2 goes into what makes a question worth tracking.

3. Read your first run

Your task: open the Overview and read four numbers, without concluding anything about a trend.

Your questions go out to three engines by default, and the answers land one by one. You do not have to wait on the screen.

NumberWhat it answers
Visibility scoreOn what share of the analysed answers are you named at all?
Average positionWhen you are named, how high do you sit?
Share of voiceOf every brand mention on your questions, how many are yours?
Positive sentimentWhen engines say something about you, is it good?
!
The arrows mean nothing yet. Your first run is a baseline. Every arrow compares a question on one engine with the previous time that same pair was measured, so they only start saying something next week.

Guide 4 explains what each number is divided by, and when one of them misleads.

4. Ship one action

Your task: open the action plan, take the card at the top, do it, and stop there.

Every run ends with a ranked list built from what was actually measured: a domain the engines really cited, a page that really failed, a search an engine really ran. Cards are ordered by the visibility points we estimate you could gain.

Do one. Not three. Model answers move a little on their own, so a single week is never proof, but changing one thing at a time is the only way to build a reading over a month. Ship five at once and you will never know which one mattered.

i
A short plan is not a bug. When the data does not support a recommendation, we would rather say nothing than fill the board.

What happens next, without you

From here the product runs on its own and you come back to read it.

WhenWhat happens
Every MondayYour questions are re-asked and a new run lands. This is the run that produces trends, and the one that sends your client report if you set one up.
Every TuesdayThe site audit checks whether AI crawlers can reach and read your pages, so a change you ship Monday evening is already in it.
When something dropsYou get an email naming the questions that fell the most, so you do not have to open the dashboard to find out.

Come back next Monday and read the same questions. That comparison, repeated, is the whole product.