Exploratory observatory

Measure before turning intuition into a trend.

The observatory collects reproducible protocols on answer-engine visibility. It only publishes a trend after comparable waves and preserves collection failures.

Independent No affiliation Primary sources Updated 13 August 2026
Wave 01 · collection status

Four executions, no model ranking.

The chart reports collection usability only. It does not compare answer quality, citation accuracy or engine performance.

CSVInitial exploratory collection.
Wave 02 · 13 August 2026

A reproducible protocol, four truncated outputs.

The exact questions, model identifiers, limits, collection states and answer hashes are preserved. No model ranking or trend is defensible from these incomplete outputs.

JSON wave 02Research indexCost: $0.715186. Comparable waves: 0. Trend: unavailable.
Baseline 01

French-language benchmark, 12 August 2026

An exploratory first wave, insufficient to rank engines.

01.01

10 planned queries

Identical questions on SEO, GEO, crawlers, measurement, ecommerce, B2B and monitoring tools.

01.02

4 executions

One complete output requiring review, one truncated output, one interrupted output and one HTTP 504.

01.03

Permitted conclusion

Collection quality varies strongly. This wave cannot support a comparative model ranking.

Wave 02

Frozen protocol, 13 August 2026

The first repository-preserved wave establishes a reproducible baseline, but not a quality comparison.

02.01

4 current models

Claude Opus 5, GPT-5.6 Sol Pro, Gemini 3.1 Pro Preview and Grok 4.5 were called through OpenRouter.

02.02

4 truncated outputs

Every execution reached its output limit before completing the ten-question set. The failure is retained as data.

02.03

$0.715186 observed cost

The run stayed below its $10 ceiling. Exact prompts, limits, statuses and hashes are public.

Open wave 02 JSON →
Access

Data and protocol

Files are readable without an account, cookie or JavaScript.

03.01

Method article

Protocol, outcomes, disagreements and primary sources.

Read in French →
03.02

CSV data

Observed status and usability of the four executions.

Download →
03.03

Wave 02 JSON

Frozen questions, exact model identifiers, limits, costs, states and answer hashes.

Open →
03.04

JSON index

Machine-readable inventory of the research programme.

Open →
Next waves

Requirements before calling a trend

Stability and repetition matter more than novelty.

04.01

Same query set

Keep intent, language, locale and qualification criteria stable.

04.02

Repeated runs

Measure response variance before aggregating mentions or citations.

04.03

Source verification

Retain outputs within applicable limits and verify every attribution.