Skip to main content
hitak

Methodology

Depth beats frequency.

One isolated answer is not enough to measure a brand. hitak repeats observations, freezes their conditions and documents their limitations.

One answer · AI 1 Present True on that day, for that wording.
12 observations · same question Top 3 in 9 of 12 answers Corpus v3 · frozen

IGPI

Global AI presence index.

A summary measurement based on a frozen corpus and repeated observations.

Corpus v3 · 8 questions × 12 observations Top 3 Mentioned lower down Absent

back mattress9 / 12

firm or soft7 / 12

best 20266 / 12

side sleeper5 / 12

customer reviews4 / 12

warranty3 / 12

delivery3 / 12

€500 budget2 / 12

IGPI · Your brand 41 % 39 of 96 observations

One answer would say “present” or “absent”. Repeated measurement shows your brand in the top 3 of 9 out of 12 answers for “back mattress”, but only 2 for “€500 budget”.

Three principles

What makes a measurement defensible.

Corpus v3 · frozen same question · 12 observations

Repeat without moving the target

The corpus and its variations are frozen so repetitions address the same measurement question.

Pure AI · 44%

Web search · 29%

Average: 37%

Separate search regimes

General-purpose assistants and web-native engines are not added together as if they produced the same observation.

●●● High ●●○ Medium ●○○ Marginal Predicted uplift: +12%

State the limitations

An incompatible corpus or scope stops the comparison. A recommendation remains high, medium or marginal: never a predicted uplift.

Honesty

What hitak does not do.

  • No quantified uplift promised

    A recommendation remains high, medium or marginal. It never comes with a predicted uplift.

  • No trend across a break

    When the corpus or scope changes, comparison stops rather than being extrapolated.

  • No average across regimes

    Pure AI and web search produce different observations; they remain read separately.

  • No invented causality

    A cited source is not a demonstrated cause. Analysis describes what is observable.