Study and run context
| Field | Synthetic value | Meaning |
|---|---|---|
| Study | Example study | Demonstration context only |
| Wave | Example wave | No real wave exists |
| Protocol | Draft example | Not an approved protocol |
| Prompt | Synthetic prompt fixture | Links to the exact synthetic prompt below |
| Service | Synthetic service A | Not a real provider |
| Run status | Fixture only | No call or timestamp |
Exact synthetic prompt
Which types of provider should a small example retailer compare when planning a fictional returns service?
Fabricated output fragment
“A fictional retailer might compare Example brand A and Example brand B. Requirements should be checked directly.”
This sentence was written solely as a synthetic example. It was not generated by an answer service.
Synthetic annotation
| Field | Synthetic value | Review state |
|---|---|---|
| Brand | Example brand A | Fixture |
| Mention position | First position | Illustrative only |
| Citation present | No | No URL was invented |
| Claim assessment | Unavailable | No factual claim or source exists |
| Coder | Synthetic fixture generator | Not a human research coder |
Why the unavailable state matters
Because the fixture has no factual source, claim support is unavailable. Treating that state as unsupported would create a false negative; treating it as supported would invent evidence.
What a real record would still need
A permitted raw-evidence reference, UTC timestamp, displayed service context, immutable panel version, coder identity, independent review state, disagreement log and release decision. None is present here.
How a genuine record would move through review
The execution log would first confirm whether the run was valid, failed, blocked, empty or excluded with a reason. A valid response would then be coded against the entire frozen panel, not only brands noticed at first reading. A reviewer would inspect a declared subset, record disagreement and preserve any correction. Only approved aggregate fields would move to a public table.
Interpretation boundary
The fixture demonstrates data shape, not model behaviour. Its mention position is not an observed rank, its unavailable claim state is not a negative score, and its lack of a citation is not evidence about a real service. Readers should not compare these labels with any company, product or market.