Core Concepts

The whole system is one loop: you describe a brand, you write the questions its customers ask, the platform sends each question to each AI assistant in each market, and every answer is graded. Seven words carry that loop.

The subject Target One brand: name, website, markets, optional brand context. Everything else hangs off it.

The question Prompt A question a real customer would type into an assistant. Classified into one of four intents.

The measurement Probe One prompt, one engine, one region, one execution — and the graded answer it produced.

The batch Run Every probe launched together. One row in score history, one set of stats.

Target

A target is one brand you monitor. It holds:

Property	Notes
Name	The brand as customers say it — what the grader looks for in an answer
URL	Your homepage. Disambiguates same-named companies and is what the Site Audit fetches
Regions	Ordered list of markets. The first is the primary market
Brand context	Optional free text passed to the analysis and recommendation models

How many targets an account may hold is a plan limit; see Plans & usage.

Prompt

A prompt is a question, stored in the target’s library, not a template. It is written the way a customer would ask it — “quietest massage gun for a small office”, not “massage gun keywords”.

Prompts are deliberately geography-neutral: no “in the UK”, no “in Germany”. The region is a parameter of the run, applied at retrieval time, so the same prompt text can be compared across markets. A prompt that names a country would measure the country twice.

Every prompt carries a source (AI, Reddit, GEO, Manual) and an intent.

Intent

Intent is what a prompt is trying to measure. There are four, in two buckets:

Intent	Names your brand?	Example	Bucket
Category	No	”best massage gun 2026”	Discovery
Scenario	No	”quietest massage gun for a 10-person startup”	Discovery
Comparison	Yes	”Dyson vs Dreame hair dryer”	Branded
Trust	Yes	”is Kospet legit in 2026”	Branded

The bucket decides the rubric. Discovery prompts never name you, so a mention is earned visibility and is scored by where and how you appear. Branded prompts already name you, so a mention proves nothing and is scored by the verdict the answer reaches instead.

This distinction is why Mention Rate counts discovery prompts only — mixing in prompts that guarantee a mention would inflate it toward 100%. Full detail in Prompt intents.

Engine

An engine is one AI assistant: ChatGPT, Gemini, Claude, Perplexity, Google AI Overview, Google AI Mode, Grok, DeepSeek, Doubao. Engines differ in how they retrieve — some browse the live web, some answer from their own retrieval backend — which is why an answer to the same question differs between them, and why per-engine breakdowns are worth reading separately.

Engines are billed identically: one answer is one probe, with no cost weighting for expensive models. See Engines & regions.

Region

A region is a market, written as a two-letter country code. Region is not a property of the model — a model’s weights have no location. It is a property of retrieval: what the engine finds when it searches from that country.

That yields three kinds of engine, and the app reports them separately:

Region signal	Meaning	Reported as
Adjustable	Retrieval genuinely runs from the target country	Per-region, and the only set used for cross-region comparison
Fixed	Retrieval is locked to one market by design	Grouped under that market
None	Retrieves without a location signal	”Global engines”

Probe

A probe is the atom: one prompt × one engine × one region × one execution. It stores the answer text, the citations, the graded scores, the competitors named in the answer, and the model’s recommendations.

If grading fails after retries, the probe is marked not analyzed: the answer and citations are still kept, but the probe is excluded from every metric — including the mention-rate denominator. An unknown result is never rounded down to a zero.

Run (probe group)

A run is every probe launched together, plus the statistics computed over them and an AI-generated interpretation of the batch. Each run is one point on the score history chart and one row in the run log, and you can switch the panel back to any past run.

How the pieces produce a score

probe → AI Score (0–100) run → average AI Score → Final Score = 0.8 × avg AI Score + 0.2 × SEO Score The SEO Score is site-level, from a Lighthouse audit of your homepage — identical in every region.

AI Score → How a single answer becomes a number. Final Score → How that number is blended with your site's technical health.