Transparent methodology
How the Credalume promise audit works
Credalume is an automated first-pass review of visible ecommerce wording. This page explains what the live product reads, how it prioritizes findings, and where qualified human review remains essential.
Short answer
Credalume extracts readable content from one public product or landing page—or uses text supplied by the user—then detects concrete promises, visible evidence signals, risk language, and transparency cues. It ranks findings using deterministic rules and shows the reason for each signal. It does not silently invent a source, fact, customer result, or legal conclusion.
What the live audit accepts
Public URL
One standard HTTP or HTTPS page that can be accessed without authentication. Private, local, internal-network, oversized, and non-HTML targets are rejected.
Pasted text
At least 80 characters of visible marketing copy, including product descriptions, landing-page copy, ad copy, or a pasted video transcript.
Supported interface languages
English, French, Spanish, and Arabic. Detection depends on the exact wording and the content visible in the submitted input.
Four-stage analysis
- Extract. Read visible headings, paragraphs, list items, quotations, and relevant links while excluding scripts, styles, menus, and common interface noise.
- Detect. Identify candidate promises and evidence language. Candidate strength increases when wording contains a result, a number, a timeframe, a guarantee, a comparison, or a health-related verb.
- Prioritize. Rank absolute, guaranteed, medical, comparative, quantified, and time-bound claims above broad descriptive language.
- Explain. State the risk, distinguish a visible signal from independent proof, and propose a safer direction without adding unsupported facts.
Risk taxonomy
| Signal | Why it receives attention | Safer review question |
|---|---|---|
| Absolute | Words such as “always,” “never,” “perfect,” or “100%” leave no room for conditions or variation. | Can the benefit be stated without certainty that exceeds the evidence? |
| Guarantee | A guaranteed outcome may create a stronger buyer expectation than the product or policy can support. | Are the conditions, exclusions, and remedy visible? |
| Time-bound | A precise timeframe is testable and should match the population, method, and outcome in the source. | Does the cited evidence support this exact timeframe? |
| Quantified | A percentage or multiplier looks precise and therefore needs equally precise context. | Are sample, method, date, baseline, and source disclosed? |
| Health-related | Treatment, prevention, diagnosis, or cure wording can be regulated and high stakes. | Has a qualified reviewer approved the claim for the target market? |
| Comparative | “Best,” “leading,” or “better than” depends on a defined comparison set and method. | Who or what was compared, when, and using which measure? |
Evidence labels are intentionally conservative
- Evidence signal
- The submitted content contains evidence wording and a reference-like link near a specific claim. This is not independent verification.
- Needs evidence
- The promise is understandable, but the page does not clearly connect it to a directly relevant source.
- Higher risk
- The wording combines a stronger risk marker with insufficient visible support and should be reviewed before traffic is scaled.
How the score should be read
The preliminary score combines visible evidence coverage, specificity, transparency, and claim-safety signals. Higher-risk and unsupported promises reduce the score; relevant transparency and evidence cues can improve it. The score is capped below 100 because an automated page scan cannot establish full truth or compliance.
Important: the score compares the submitted page with the audit rules. It is not a scientific confidence percentage, legal-compliance grade, conversion forecast, or substitute for customer research.
What the free snapshot reveals
- The highest-priority promise as a risk preview, without exposing the paid evidence review or final rewrite.
- One medium-risk promise with a complete explanation, evidence check, and safer direction.
- One additional medium-risk preview when the page contains it.
- A preliminary score, visible reference signals, method summary, and explicit limitations.
What Credalume cannot conclude
The launch product cannot validate a scientific study, judge whether evidence is legally sufficient in every country, access private documents, observe real customer behavior, evaluate uploaded video files, or prove that a rewrite will increase conversion. Those tasks require source inspection, market-specific expertise, accessibility testing, experimentation, or research with real users.
Need to audit a page?
Use the free snapshot on one public URL or paste the text customers see. Do not submit private, personal, regulated, or confidential information.
Run the free snapshot ↗