NovolAI Search Visibility Benchmark 2026 — prompt set
A frozen bilingual prompt set for measuring mention, citation and referral visibility across five AI engines.
- Author:
- Vladislav Novoloake
- Reviewed by:
- Novol editorial review
- Updated:
Direct answer
The benchmark defines 30 EN/RU prompts and separate mention, citation and referral fields for five AI engines. The baseline is intentionally marked pending until the runs are captured with dates, engine labels and cited URLs.
What matters
- 30 frozen prompts: 15 English and 15 Russian.
- Five engines at baseline and days 30, 60 and 90.
- Mention, citation and referral remain separate metrics.
What is measured
A mention records whether Novol appears in an answer. A citation records whether a Novol URL is used as a source. A referral is an actual classified visit. These events must not be collapsed into one visibility score.
Prompt and run controls
The set covers five brand/entity, six product and four service prompts per language. Every run records the engine, date, model label when shown, response presence and cited URLs.
- No personalized account context where avoidable
- Same prompt wording at each wave
- Screenshots or exports retained as evidence
- No causal claim from a single change
Publication status
The prompt set and blank run template are public. Baseline scores remain pending rather than inferred from search impressions or manually selected examples.
Open research files
Sources and evidence
Download the frozen prompt set
Novol software studio, operated by FLOWPIX LIMITED.