Methodology

Don't guess. Check systematically.

Every value carries its AI search engine, model version, measurement path, measurement time and denominator. This page shows the three measurement paths, why we ask every question more than once — and where the measurement honestly reaches its limits. For ongoing work this method is bound to the plans; the Deep Audit additionally reviews visible interfaces by hand.

Prepare project start See the Deep Audit

Measurement paths

Three paths from search prompt to value.

A measurement point always carries its origin. The three paths get their own denominators and timelines — they never disappear into a shared average. Click a layer to highlight its path.

Search prompt → API → observation

The planned, comparable measurement path.

API measurement paths for OpenAI, Google Gemini and Perplexity are technically prepared for the planned launch. They will be used in production only after documented provider, contract and quality approval. Model version, search mode and time stay attached to the value.

Measurement plan & stability

Only repetition makes a value.

A single hit may be chance. That is why search prompts are repeated separately. Recurring patterns become visible; one-off observations stay labelled and never become a recommendation without review.

Two runs, one verdict

First we measure, then we compare: the run views only show what was observed. Only the comparison reveals what recurs — and what stays a single observation that gets reviewed rather than sold.

Provider A mentionedobserved
Provider B mentionedobserved
Source X citedobserved
Source Y citedobserved
Provider A mentionedobserved
Provider B mentionedobserved
Source X citedobserved
Provider C mentionedobserved
Provider A mentionedin both runs
Provider B mentionedin both runs
Source X citedin both runs
Source Y citedrun 1 only
Provider C mentionedrun 2 only
recurringsingle observation — gets reviewed
Value passport · observation 0184complete
47%Visibilityfictional sample value
AI search engine
Google Gemini
Model version
documented
Search mode
API · standard
Measurement time
day 02 · run 2
Denominator
50 prompts
Sources
as supplied

Without these fields, no value becomes a finding.

Privacy

A lean data flow.

Prompt sets are designed so that no client personal data or credentials are needed. Prohibited content is removed before measurement. The flow from prompt set to report stays deliberately short; the details will be governed by the privacy policy of the published site.

Path of the data

Four stations, a short path

From the approved prompt set to the report it takes four steps — nothing more is needed.

  1. Your approved prompt setWritten for market questions and checked for inadmissible content before measuring.with you
  2. Measurement via the interfacesThe prompts run across the agreed AI search engines; answers and sources are recorded.documented
  3. Review and editingWarning cases are opened; internal raw data and keys stay separate from the client result.at peaksly
  4. Report to youChecked values and sources — no third-party account data, no raw-data sprawl.delivered
Excluded

What is never collected

The simplest data protection is not collecting data at all.

We measure market questions — not people.

Honest limits

What this measurement cannot do.

No guarantees

Mentions, positions, rankings, traffic or revenue cannot be promised — by anyone.

principle
No full coverage

The market of AI search engines changes fast. We measure the agreed, documented scope.

principle
No interface mirror

API measurements and visible interfaces are documented and reviewed separately.

principle
Sentiment & share of voice

Appear as values only after passing their acceptance review — until then shown as “in review”.

status
Prepare project start See the Deep Audit