Crawl and classify
Find canonical pages, extractable sections, links, structured data, robots rules, sitemap, and `llms.txt` signals.
Methodology
Measure AI visibility without hiding the evidence.
CiteAura treats crawlability, extractable facts, model answers, citations, and product-surface observations as different evidence streams. The result is a defined measurement, not a universal ranking claim.
Evidence model
Question × provider × sampling mode
Every visibility result keeps its provenance so teams can compare like with like and see what is still unmeasured.
01Public site
02AI answer
03Verification
Five inputs define a useful measurement.
01
Public source
The canonical domain, page set, rendered HTML, machine files, structured data, and official facts form the inspectable source layer.
02
Question set
Questions are grouped by intent, such as definition, comparison, recommendation, pricing, and problem diagnosis. Brand probes are kept separate from target questions.
03
Provider cohort
Only configured and successfully sampled providers enter a measured cohort. Missing providers remain visible as unmeasured instead of becoming an implied zero.
04
Sampling mode
API · Model knowledge, API · Web-grounded retrieval, and Manual · Product surface answer different questions and are never silently merged.
05
Time window
Before-and-after comparisons use a defined measurement period. Model updates, retrieval changes, and question changes are recorded as cohort changes.
From evidence to an acceptance check.
Replay raw answers
Keep answer text, citations, provider, question ID, and mode so an aggregate can be traced back to its evidence.
Re-check the change
Use deterministic page checks for technical work and a comparable re-sample for visibility work. A draft remains a draft until reviewed.
Interpretation boundary: a measured mention or citation is evidence for the defined cohort and period. It is not a promise about every user, model, region, or future answer.
Methodology questions
Does an API answer equal a live AI search result?
No. API model knowledge, web-grounded retrieval, and manual product-surface observations stay separate.
What makes two visibility measurements comparable?
The question set, provider cohort, model, sampling mode, market, and measurement period must be defined and comparable.
Why does CiteAura show “unmeasured”?
Because insufficient evidence is a result. The product exposes provider and sample coverage instead of estimating missing observations.
Read the operational steps in CiteAura measurement documentation, inspect the sample GEO diagnostic report, or review the AI visibility audit workflow.