The method · in development
How we measure
How we intend to connect a buyer’s question to inspectable evidence, a reviewed website change and a later measurement.
From a buying question to a compatible later reading
This is the workflow JanusLabs is developing. The Pilot demonstrates an early part of it; it does not establish a fully operational commercial service.
- Buying questionRecord the exact question and why it matters to a buyer. Keep the comparison question fixed.
- Saved answer and sourcesRetain the original answer, date, service, UK settings and cited URLs. Record failed readings separately.
- Confirmed factCheck the claim against reliable business facts, including identity, audience and availability.
- DiagnosisAI can propose an assessment. Review the evidence and keep uncertainty where the cause is unknown.
- Justified priorityChoose a change supported by the evidence, buyer relevance and the owner’s ability to act.
- Prepared changeRecord the proposed edit and its reason. A draft is not a published improvement.
- Verified publicationCheck the actual live page and record the date. Approval and deployment are separate from preparation.
- Compatible later measurementRepeat unchanged questions under compatible conditions and compare each service separately.
Follow the JanusLabs identity question through this workflow →
Current Pilot scope
Our UK Pilot uses 20 fixed questions across ChatGPT Search, Google AI Mode and Google AI Overviews. This is the Pilot’s scope, not a statement that every planned reading completed.
Public app access and client onboarding are not available. Wider product coverage, commercial question allowances, monitoring frequency, prices and subscriptions remain to be confirmed. Current collection status still needs checking; this page does not promise that collection is running or paused.
Keep different outcomes separate
Classifications and proposed diagnoses can be wrong. Keep the original answer and supporting excerpts, review factual claims and record corrections without rewriting the source evidence.
Failures, denominators and uncertainty
A visibility count needs the number of usable answers it was counted over. A failed or unavailable reading is unknown and must be reported separately, rather than counted as absence. Investigate provider failures and unconfirmed charges before resuming collection.
A question set is a sample, not all buyers or all conversations. Repeated readings can be related; they are not independent buyers. Small samples and changing settings limit comparisons. Do not present a narrow range as proof of reliability or pool different services into a claim about every answer.
The Pilot has not established a complete accepted baseline. Compatible later readings and an accepted starting window are required before assessing change. Movement after publication does not prove the website edit caused it.
What you can inspect now
Our Pilot walkthrough contains an intentionally limited excerpt checked against the original saved answer, with its source list and remaining limits. The example report shows a fictional product report with original sample content. Neither is evidence of a customer outcome.
Use the buyer guide to compare evidence, accuracy reviews, workflows and equivalent UK coverage. Plans are provisional concepts, not subscriptions available to buy.