Methodology 1.0
Airankscan is designed to answer a practical question: does this brand's public evidence give AI systems enough clear, credible material to understand and recommend it? The scan separates what was observed from what was modeled so the report is useful without overstating certainty.
Updated July 17, 2026
01
Airankscan reviews public pages that can help an answer system understand a brand: the submitted site, relevant category and product pages, third-party profiles, reviews, directories, articles, comparisons, and competitor evidence discovered during the scan.
02
The scan creates realistic discovery, comparison, validation, local, and decision-stage questions from the submitted category, market, competitors, and goal. These are research scenarios—not a claim that real buyers used those exact words or that the prompts have a known search volume.
03
Observed evidence is converted into consistent signals such as category clarity, owned-source coverage, third-party corroboration, competitor differentiation, and support for the modeled buyer questions. Conflicts and missing evidence remain visible instead of being treated as proof.
04
Fixed scoring rules assign preset points and caps to the normalized signals. The AI model helps research and structure evidence, but it does not freely choose the final score. The same normalized evidence under the same methodology version produces the same readiness score.
What the score means
The 0–100 score summarizes how well the available public evidence supports a brand across the report's buyer scenarios. It is not a measured position in ChatGPT, Google, Perplexity, or another engine.
Each normalized signal uses defined points, limits, and status rules.
A thin or single-source footprint cannot earn full readiness through confident wording alone.
Reports should identify the methodology used so results can be compared responsibly.
Subscores point to the evidence gaps that can be improved, not just a headline number.
Published scoring rules
Each modeled prompt earns 100 points for Strong, 65 for Partial, 30 for Weak, or 0 for Missing. Brand presence is the average of those prompt points.
65% comes from verified-source volume—12.5 points per verified source, capped at 100—and 35% comes from the share of prompt findings supported by at least one citation.
70% comes from brand presence and 30% from inverse competitor pressure, so repeated competitor dominance lowers this component.
Each evaluated claim earns 100 points when Supported, 55 when Partial, or 10 when Unverified. Claim readiness is the average of those claim points.
75–100
Strong
55–74
Promising
30–54
Thin
0–29
Hidden
Confidence
Confidence describes support in the observed evidence. It is not a statistical probability and does not predict a future ranking.
Source provenance
A report distinguishes the submitted brand site, observed third-party evidence, competitor evidence, and model-assisted interpretation. Source links are preserved only when they are valid public web addresses associated with the scan evidence.
A source link shows where supporting material was observed. It does not prove that every AI engine cited or relied on that page, and it does not mean Airankscan endorses the source.
If a source cannot be validated or the evidence is too weak, the report should lower confidence, mark the gap, or omit the claim.
Direct engine checks
The standard scan uses OpenAI for model-assisted analysis and live public web research. It does not directly query Google AI Overviews or AI Mode, Perplexity, Gemini, Claude, Copilot, or each consumer version of ChatGPT unless the finished report explicitly identifies a separate direct observation from that engine.
Modeled prompt findings estimate how the observed evidence could support an AI answer. They must not be presented as screenshots, quotations, or rankings captured from an outside engine.
Public pages, crawl access, reviews, competitor claims, model behavior, web indexes, and source availability all change over time. A scan is a dated baseline, not a permanent verdict. The most useful comparison is a repeat scan using the same scope and methodology after meaningful site or source improvements have been published.