Mention coverage
Share of eligible prompts in which each brand is explicitly mentioned.
Python · AI Search · Competitive intelligence
AI Search Competitive Intelligence
An independent benchmark comparing Notion, Asana, ClickUp and monday.com across the same controlled buyer-intent prompt set. This is research, not client work for any benchmarked company.
At a glance / Evidence status
Problem
How can four SaaS platforms be compared in AI-generated answers without confusing brand mentions with citations or recommendations?
Method & tools
35 versioned buyer-intent prompts, seven intent families, public-source evidence, structured observation fields and Python-oriented scoring rules.
Deliverables
Benchmark protocol, measurement definitions and an auditable observation design.
Verified status
Protocol prepared; AI response observations and comparative performance results have not yet been collected or published.
Dataset v1
35 frozen prompts across seven intent families: category discovery, alternatives, comparison, capability, pricing/value, implementation and recommendation. Prompt definitions are versioned before observations are scored.
Measurement model
Citation coverage
Share of eligible prompts with evidence cited for the brand.
Recommendation coverage
Share of recommendation prompts in which a brand is included.
Intent consistency
Family-level visibility is preserved so a strong result in one intent cannot hide weakness in another.
Audit trail
Every observation records engine/run, prompt, brand, mention, citation, recommendation inclusion, ordered position where meaningful, source type, source URL and an evidence note. Repeated runs append evidence instead of overwriting history.
Scoring guardrail
Mean position is calculated only for explicitly ordered shortlists and is never presented as a universal AI-search rank. Missing observations are never imputed. The Python pipeline refuses to invent results.
Status
Protocol and dataset are ready. Results remain unpublished until real observations are collected and auditable.