Use case · Market research
Before you enter a market, ask the engines who already holds it
A new segment, a new country, a product idea: the engines are already answering your future buyers. Epovest asks them that market's buying questions and reads who gets named, how often, and on whose pages the answers rest.
Views from the app. Your figures replace the sample at the first survey.
A market, read through the answers your future buyers get
Nothing of yours is measured here. The tracker asks a market's buying questions and you read the shape of the answers: who is named, how many different names come back, how much of the mentions the leaders hold, and which pages the engines lean on to answer. One tracker per candidate market, the same questions on each, and the fields line up side by side.
What you measure
Who the engines name
The names that come back on the market's buying questions, engine by engine. An answer that names nobody is a reading of its own.
How crowded the field is
How much of the mentions the leading names hold: four names owning a field and fifteen sharing it are two different markets.
The pages behind the answers
The sources cited on those questions, ranked by AI Authority: the players' own pages, or independent third parties.
Fields side by side
The same grid on every candidate, run the same day: what you compare is markets, and the reading carries its date.
The recipe
How to run a market study with Epovest
The full working guide for this use: one tracker per candidate field, the same grid of buying questions on each, a first run that names the players, four readings that tell a held field from an open one, and a decision that leaves a dated baseline behind it.
One field, one tracker, and no name of yours
The unit of this study is the field, not the player: one project per candidate market, one tracker inside it, and nothing of yours in the settings.
That is what sets this use apart. You hold no ground here yet, so there is no canon to settle, no page to align, no mention to keep: the tracker is the whole instrument, and what it measures is the state of a market in the engines' answers. Keep the candidates in separate projects, one each: the sources cited on ITS questions stay its own, and the day you enter one of them, everything already collected is waiting there.
The same grid of buying questions on every field
A comparison holds only if the questions have the same shape everywhere: three buying questions per field, one per intent, and only the trade changes from one field to the next.
"Which [trade] would you recommend for [situation]?"
"What is the best [product category] for [type of buyer]?"
"[trade or category]: how do I choose, and what does it cost?"
Write them the way a buyer asks, in the language and the country you would sell in: the engines do not name the same players in French and in English, and a field can be wide open in one market and held in another. Three questions covering the neighbourhood teach you more than one question asked three times: what you are after is the field's answer, not one wording's.
The first run names the players, the second counts them
You do not supply the names: the answers do. That inversion is the point of this use, and it is what a market panel used to be bought for.
1 First run, no keyword
Start the field with no keyword at all: your questions are the study. The first answers name who is installed, in the order the engines reach for them, and they name some you had on no list.
2 Second run, the names as keywords
Seed the recurring names as keywords with the presence analyst, and leave discovery on: from there the field is counted instead of read, and discovery keeps the roster honest by surfacing the domains that keep coming back.
The fields where the answers stay generic, a category explained and nobody named, deserve a second look: their citation slot is open, and the first name that becomes readable there is the one the engines will reach for.
Reading a field: four questions, in this order
The same four readings on every candidate. Together they say whether there is room, and what entering would take.
- 1 Do the answers name anyone at all? An engine that explains a category without naming a provider is describing a field where nobody is installed yet. That is the strongest signal this study produces, and the cheapest to obtain.
- 2 How many names come back Count the distinct names over the whole run, never over one answer: a field that returns the same four names is settled, one that returns fifteen different ones is still sorting itself out, and the two call for different entries.
- 3 How concentrated the mentions are Read the share the leading names hold engine by engine, never as an average: a field can be held on one engine and open on the next, and that gap is where an entry starts.
- 4 Whose pages the answers rest on Open the field's Atlas: the sources cited on its questions, ranked by AI Authority. Answers built on the providers' own pages describe a field you enter by publishing; answers built on independent third parties describe one you enter through those third parties, and the Atlas names them one by one.
Sound every candidate in one wave
Fields are compared, so they are measured together: the same grid, the same settings, started the same day.
Give them all the same cadence and the same resolution, then launch the first run on all of them at once: what separates two fields is then the field itself, and not the week it was measured. Monthly suits a decision of this size, and a second wave a month later adds movement to the picture, which fields are filling up and which are standing still. Your assistant can open a whole wave from one instruction and read it back to you field by field.
From the wave to a decision, and what it leaves behind
This study ends on a choice, and the instrument survives the choice: whichever way it goes, the record stays.
The field you enter
It already carries its dated zero, the answers as they stood before you existed, which is the baseline every launch wants and few have. Keep the tracker running: the launch playbook takes over from there, and the curve starts at a measured zero rather than a remembered one.
The fields you set aside
Pause the tracker and archive the project: both keep their record intact and stop asking for work. The day the question comes back, the study is where you left it, with its dates.
Date the decision itself in the logbook, with what made it. Six months later the series reads with its reason in front of it, and the field you left aside is still arguable on evidence rather than on memory.
Worked example
Five fields, one wave, one decision
Five candidate markets, one project each, three buying questions apiece, monthly, all started the same afternoon. The first run comes back: two fields return the same four names on every engine, one returns nothing but generic advice, two sit in between. The Atlas separates them further, in the crowded fields the cited pages are the providers' own sites, in the open one they are trade bodies and forums that sell nothing. The names seen in the first run go in as keywords, the wave repeats a month later, and one field has already moved. The decision goes into the logbook with its date; the four other trackers pause, their record intact.
With your AI assistant
Your AI assistant runs this recipe with you
Connect Epovest to the assistant you already use, one-click connector or API key. It creates trackers, reads the results and keeps the registers with you; a payment only ever happens with your approval.
The instruction to give it
Connect to Epovest and sound three markets for me: one project each, one tracker each with the same three buying questions in [language], no keyword, monthly. When the first run lands, read me for each field who gets named, how many distinct names come back, and which sources the answers rest on.
$0.10
per prompt, per engine, per check
Pricing
Whatever you configure, the unit never changes. Your total shows live in the configurator before anything runs; you start whenever you want, from a $10 top-up. No subscription: only the checks that run are debited.
Your shortlist moves as you learn
Add a candidate at any time: its tracker starts its own series, next to the others. A field you set aside pauses and keeps its record; the one you pick keeps its series, and everything already collected becomes the dated zero of your launch.
The other use cases
Make AIs recommend you
Create a tracker in minutes: your prompts, your engines, your cadence. First checks within 24 to 48 hours.
No subscription. Credits from $10, valid 12 months. Pause anytime: the history stays yours.