What Landing Page Lie Detector Flags — And What It Cannot Prove

1. Hypothesis

The hypothesis was that most bad landing page copy isn't born from malice, but from a habit of relying on established marketing tropes (e.g., "revolutionary", "best-in-class"). If the tool could flag these common tropes locally, writers would replace them with concrete facts before publishing.

2. What I built

Landing Page Lie Detector — a purely client-side, regex-based tool that flags text into four categories:

• Vague promises (e.g., "game-changing")
• Absolute claims (e.g., "never", "everyone")
• Fake urgency (e.g., "act now", "limited time")
• Proof gaps (claims that lack nearby numbers, links, or evidence)

I chose not to use a generative AI model. A regex engine is faster, fully deterministic, and requires no API calls — consistent with the privacy principle behind this tool.

3. Example Input/Output & UI Decisions

Input: "Our revolutionary platform is guaranteed to boost your conversions overnight. Everyone is switching already. Limited time offer ends tonight."

The UI Decision: The detector calculates a "Trust Score" and highlights problematic words. I implemented a secondary check for Proof gaps: if a claim word (like "increases") is found, the tool scans the adjacent 90 characters for numbers or links. If evidence is nearby, the penalty is significantly reduced.

Output: A "Courtroom chaos" rating. "Revolutionary" is flagged as too shiny, "guaranteed" and "everyone" as absolute claims, and "limited time" as urgency pressure.

4. What I observed

No user tracking was in place when this note was written. These are design observations and intended use cases, not measured outcomes:

• Observation: The regex flags common SaaS marketing tropes — absolute claims, urgency language, and vague proof. Coverage of edge cases and industry-specific language is untested.
• Observation: The tool cannot prove a claim is actually true. It only detects whether a claim is structurally supported by nearby evidence — a number, a link, or a named source within 90 characters.
• Design intent: The intended use is catching lazy copywriting patterns before publication, not verifying factual accuracy. The proof gap check is meant to force claims to appear near supporting data rather than standing alone.

5. What changed

No expansions to the evidenceRegex were made after the initial build. The four flag categories — vague promises, absolute claims, fake urgency, proof gaps — remain as implemented.

6. Status

kept

FAQ

Can the tool verify whether a claim is actually true?

No. The tool checks whether a claim is structurally supported — whether a number, link, or named source appears within 90 characters of a claim word. It cannot verify whether those numbers are accurate or whether the underlying claim is factually correct.

Does it send pasted copy to a server?

No. The regex rules run entirely in the browser. Nothing you paste reaches a server or an external API.

What counts as a Proof gap?

A proof gap is flagged when a claim word like "increases" or "improves" appears without a number, link, or named source nearby. The tool scans 90 characters before and after the claim word. If supporting evidence is present, the penalty is reduced.