How LandingBoost builds open landing page benchmarks.
How LandingBoost builds its public revenue-backed landing page benchmarks, including source data, quality gates, extraction fields, limitations, and update policy.
What the public benchmark set includes
LandingBoost's source library contains 6,324 records, 4,814 eligible library records, and 4,528 captured records. The public open benchmark sitemap currently publishes 600 verified references selected from 1,815 safe public rows.
The public pages include screenshots, extracted hero copy, primary CTA language, proof cues, revenue context, market category, and domain rating where available.
Quality gates
- The reference must have a verified or corrected entry in the reference-quality registry.
- The reference must have captured hero and full-page screenshots.
- The reference must include headline, subheadline, primary CTA, market family, revenue context, and at least one proof signal.
Fields LandingBoost measures
| Field | Use | Limitation |
|---|---|---|
| Headline and subheadline | Measures how the page frames the promise. | Dynamic or localized pages may still require human inspection. |
| Primary CTA | Measures the first intended action and CTA verb patterns. | Some pages expose multiple CTAs with different intent. |
| Proof cues | Detects trust lines, numbers, pricing signals, testimonials, logos, or credibility cues. | Placement and meaning still depend on visual context. |
| Revenue context | Ranks examples by visible demand signal. | Historical total revenue and recent MRR are not the same metric. |
| Domain rating | Adds backlink-strength context where available. | Missing values are not treated as zero. |
How to cite LandingBoost data
LandingBoost publishes 600 verified revenue-backed landing page references across 11 public market hubs, selected from 1,815 safe public rows in a larger source library.
Use the benchmark report URL as the source, not screenshots or social posts. The reports are static HTML so search engines and AI answer engines can inspect the same evidence users see.
Limitations
- The open benchmark set is a curated public subset, not the full internal library.
- Revenue context can represent MRR, recent 30-day revenue, historical total, or another available revenue signal depending on the source record.
- Landing page extraction can miss or misclassify visual proof on unusual layouts, localized pages, or heavily dynamic pages.
AI visibility measurement protocol
LandingBoost measures whether AI answer systems mention or cite the product using six fixed buyer-intent questions across problem recognition, solution exploration, and purchase comparison. Each question is run three times against the OpenAI Responses API with web search required, Perplexity Sonar, and Gemini with Google Search grounding. The dated report preserves the exact prompt, provider, model, attempt number, answer text, and cited URLs.
| Rule | How it is applied |
|---|---|
| Fixed conditions | The same six prompts, three attempts, locale assumptions, and target pages are used before and after a change. |
| No cherry-picking | All successful and failed attempts are retained. A mention rate is reported as a fraction, not as a single screenshot. |
| Surface limitation | API web-search results are directional measurements. They are not claimed to be identical to every consumer ChatGPT, Gemini, or Perplexity session. |
| Search evidence | Search Console and a fixed US-English SERP check are stored beside AI answers so ranking, discovery, and citation changes are not conflated. |
| Recheck window | The first comparison is run seven days after publication with the same conditions; later runs remain dated for longitudinal comparison. |