Open-source case study - completed February 9, 2026

21-Day AI Entity Visibility Experiment

The Miracle Berry Protocol is a public NoorRank experiment measuring how a newly launched entity was recognized and cited by AI answer engines.

The direct answer

From January 20 to February 9, 2026, NoorRank published a new research entity and tracked its visibility in ChatGPT, Perplexity, and Claude. The experiment recorded first indexing, entity recognition, citation behavior, and cross-platform differences over 21 days.

The results were mixed: ChatGPT recognized the project on Day 9 and remained stable through the observation window; Perplexity first recognized it on Day 17 but was volatile; Claude did not recognize it by Day 21. These are observations from one public case study, not a universal timeline or performance guarantee.

21 days

January 20 to February 9, 2026

3 platforms

ChatGPT, Perplexity, and Claude

Open evidence

Code, metrics, and reports published

Results by platform

The final report distinguishes recognition from sustained establishment.

ChatGPT

Recognized on Day 9

The protocol was recognized and noorranklabs.com was cited in documented tests. Recognition remained stable for more than 11 days of observation.

Perplexity

Recognized on Day 17; volatile

Recognition was observed on Day 17, disappeared in a later check, then reappeared in a documented Day 20 follow-up. This was treated as a volatility finding, not proof of permanence.

Claude

Not recognized by Day 21

The protocol was not retrieved in the documented end-of-experiment tests, showing that timelines and retrieval behavior vary across answer engines.

What was measured

Entity definition

Persistent Schema.org identifiers and consistent descriptions were used to distinguish the protocol from the miracle berry fruit.

External corroboration

The project was documented through the public site, GitHub repository, and supporting technical and community content.

Search visibility

Google Search Console milestones were tracked alongside publication and indexing events.

Answer-engine checks

Fixed direct and entity-focused queries were tested across the three platforms, with recognition and citations documented as milestones.

Limits of the experiment

  • This is one public case study, so its timing and outcomes should not be generalized to every website, brand, or answer engine.
  • Answer-engine responses can change with query wording, retrieval indexes, model updates, and source competition.
  • Perplexity's observed volatility shows that a first citation is not the same as stable, long-term recognition.

Inspect the evidence

The repository includes the milestone dataset, structured-data implementation, and day-by-day research reports used for this case study.