# Repetition exercise — wholly synthetic

These are invented bookkeeping states, not provider responses. No new run, cost, benchmark or customer observation is reported.

Question: can the review distinguish a completed run with no exposed query field from a timeout?

Conditions: fictional fixed topic “SEO tools”, English, all countries, native evidence mode. Provider/model identifiers remain placeholders; no actual API conditions are claimed. Cost cap: no requests authorised by this exercise.

Rules set before review:

- A completed run with no exposed query strings counts as a completed zero-query observation.
- A timeout has an unknown query count and stays outside the completed-run denominator.
- Compare only exact exposed strings; no invented queries or inferred source relationships.
- Raw provider output is absent. Public material is only this original editorial ledger.

| Attempt | Time | Completion state | Exposed queries | Cost | Deviation |
|---|---|---|---|---|---|
| A | fictional T0 | completed | 0 | not incurred | no query field in the invented case |
| B | fictional T0 + 1 | timeout | unknown | not incurred | no completed response |

Completed denominator: 1. Failed/timeout attempts: 1, separately reported. A conclusion such as “two runs returned no queries” is incorrect even in this exercise.

This demonstrates status accounting only. It says nothing about a real provider's probability of searching, output stability, rankings or consumer behaviour. Replace every condition and obtain the necessary cost/rights approval before doing actual research. Dispose of retained material according to the permitted deadline.