METHODOLOGY

How we know what we say we know.

AI visibility is only as trustworthy as the method behind it. This page summarizes how Searchestra measures, what each number means and what we do not claim, so you can judge our numbers rather than take them on faith.

Read the full methodology

WHY WE PUBLISH OUR METHOD

Two tools can measure the same brand and disagree.

Different questions, platform coverage and definitions produce different numbers. That is why the method matters more than the headline score, and why we would rather you trust a number you can interrogate than a confident figure you have to take on faith.

THE UNIT OF MEASUREMENT

A buyer question, an AI platform, a real answer.

Our unit is a buyer question paired with an AI platform, run as a real request rather than a simulation. Results are reported per platform, so a strong showing on one surface never hides invisibility on another.

Coverage: this is our catalog of AI platforms. A platform is only queried when access to it is configured, so coverage depends on the platforms enabled for an account and we always report per platform.

WHERE THE QUESTIONS COME FROM

Demand first, then questions.

The question set for a category is generated from a taxonomy of topics and realistic buyer scenarios, so coverage is systematic rather than picked toward questions a brand happens to win.

MarketTürkiye · Kitchen
Demand universeSmall kitchen storage
Question universeVersioned buyer questions
AI answersMeasured per platform

Demand universe: one buyer need and decision context within a market. Question universe: the versioned set of buyer questions that measures it. Results are always read in that chain: market, demand universe, questions, answers.

Real search-demand signals inform the set; synthetic expansion only fills coverage gaps
Classified by intent: informational, commercial, transactional, navigational and comparison
Brand-neutral by construction: the measured brand never shapes what is asked
Versioned and immutable: every run records the question-set version it used
Each question carries its intent, scenario, source type and version

THE FOUR LAYERS

Four separate measures, not one blended score.

We report along the four layers of the IAB framework so you can see where representation breaks down. The layer names are the IAB framework’s; the metrics are how Searchestra reports each layer.

Layer	The question it answers	How Searchestra reports it
Presence	Are you in the answer at all?	Mention rate, share of voice
Prominence	How central are you when present?	Average rank, citation share
Portrayal	How are you described?	Sentiment
Persuasion	Are you actually recommended?	Recommendation strength

Naming: our operational metrics are not always the IAB metric of a similar name. Citation share is a brand’s share of all citations, not the IAB Citation Rate; recommendation strength follows the IAB definition, scored with the rubric below.

Recommendation strength rubric: for each answer that mentions you, we classify the three dimensions the IAB names: endorsement (specific reasoning, generic, or none), prominence in the response (primary, one of several, or peripheral) and qualification (unconditional or qualified). The score runs from 0 to 4. No endorsement scores 0; otherwise specific reasoning scores 2 and generic endorsement 1, plus 1 if the recommendation is primary and plus 1 if it is unconditional. We report the share of answers scoring 3 or more as active recommendations, together with the full breakdown. A negative mention is not a recommendation; tone is reported under Portrayal. When the rubric changes, earlier scores are kept separate and never mixed into the new measure. Post-citation click-through is not measured, because it depends on click data that AI platforms do not share.

THE FORMULAS

Where a standard definition exists, we follow it.

These metrics carry a shared, standardized definition, and we compute them to that definition, with no proprietary reweighting.

Share of voice: the competitive set is the brands tracked in each market, defined at setup and visible to you. The denominator is the sum of mentions across exactly that set, nothing hidden.

REPRODUCIBILITY AND HONESTY

Unavailable means unavailable.

Our principle is simple: no fabricated certainty. AI answers vary between runs, so a single answer proves nothing. We report what we collect and flag what we cannot.

Answers are aggregated across runs, so one volatile response never masquerades as a trend
A platform change is treated as a new baseline, not as a brand win
If a signal is not measured for a run, the field stays empty rather than filled with a guess
Some signals, such as sentiment, are treated as signals rather than verdicts
An API result can differ from what a signed-in person sees, and we say which layer a number reflects

DIRECTIONAL VS DECISION-GRADE

Not every number is safe for every decision.

Directional
Early signals, trend reading and internal briefings.
Decision-grade
Budget allocation, agency evaluation and executive strategy. These need enough questions, intent coverage, a steady cadence and a documented method.

We build toward the IAB decision-grade criteria: versioned questions, aggregation across runs, per-platform reporting and brand-neutral construction. Where a metric only supports a directional read, we label it that way.

Read the full methodology.

The complete version covers every definition, formula and limitation in detail, with the reasoning behind each choice.

Read the full methodology