A · Open discovery
Signmos appeared in 0 of 9 candidate lists. This records the answers, not the cause of the absence.
Verification package
Signmos reference case
One product, documented in stages. This navigation keeps the current stage and its published artifacts together.
The measurement report expires. The summary, consumer-agent results, and public JSON digests remain as the durable record.
Recorded finding · Signmos S0 · 31 Aug 2026
What ChatGPT, Claude,and Gemini returned.
This is the readable record of 27 runs in browser-based consumer products before Signmos was changed. Auditmos operates both this program and Signmos; this is a disclosed reference case, not independent demand validation.
Compare the three productsConsumer runs27
Products3
Open discovery0/9
Qualified verdicts0/9
Consumer scope
Three tasks, repeated three times in each product.
Each prompt started in a fresh session and was submitted without a repair follow-up. The included scope covers consumer browser or desktop access only. A controlled-source API task was deferred and this record contains no API result.
Results by product
Signmos was not found or confidently qualified in any recorded run.
Open discovery, named-vendor evaluation, and requirement verification measure different things. The table keeps them separate.
| Product | Visible setting | Open discovery | Named vendor | Source verification |
|---|---|---|---|---|
| ChatGPT | Signed out · model not exposed | Signmos absent in 3/3 | 3/3 Insufficient Evidence | Relevant Signmos sources used in 3/3 |
| Claude | Opus 5 · High · Incognito | Signmos absent in 3/3 | 3/3 Insufficient Evidence | No Signmos source found in 3/3 |
| Gemini | Signed out · Flash-Lite | Signmos absent in 3/3 | 2 Insufficient Evidence · 1 wrong-vendor Disqualified | 1 relevant · 1 none · 1 wrong vendor |
Three recorded findings
The result changes with the task, but not into a qualification.
B · Named-vendor evaluation
Eight runs returned Insufficient Evidence. One Gemini run substituted SignsOS and returned Disqualified; it is an entity-resolution error, not evidence against Signmos.
C · Requirement verification
The nine answers produced 126 requirement grades. Retrieval varied sharply across products and one run attributed another vendor's facts to Signmos.
Requirement-level grades
Source retrieval produced materially different answer quality.
CS = correct and supported; CU = correctly left unknown; I = incorrect; U = unsupported; O = omitted; NR = an existing relevant source was not retrieved.
| Product | CS | CU | I | U | O | NR | Total |
|---|---|---|---|---|---|---|---|
| ChatGPT | 27 | 7 | 1 | 7 | 0 | 0 | 42 |
| Claude | 0 | 12 | 0 | 0 | 0 | 30 | 42 |
| Gemini | 2 | 7 | 15 | 2 | 0 | 16 | 42 |
| All | 29 | 26 | 16 | 9 | 0 | 46 | 126 |
Evidence boundary
This is a point-in-time consumer result, not a claim about provider internals.
The record establishes what these products returned for the fixed prompts in this run window. It does not reveal provider indexes, ingestion, caches, routing, personalization, geography, or the causal reason a source was missed. Later answers may differ.
Inspect the evidence
The readable result links to the machine record and every raw answer.The public JSON preserves the scope and aggregate counts. The repository record preserves grading decisions, complete outputs, citation captures, screenshots, and the SHA-256 manifest.