How Laup measures what automated systems can retrieve from your site.
The AI readiness scan is a technical diagnostic. It tests crawler access, discoverability, structured product information and how much page content is available before JavaScript runs. It does not measure whether AI recommends you.
Whether Laup can reach the site, discover candidate pages, retrieve their raw content and identify the structured signals the scanner checks.
Whether an AI engine knows you, recommends you or chose you because of a particular technical signal. AI Visibility measures that separate question.
The same page, two views.
Many crawlers begin with the server response, while a browser can reveal content assembled later by JavaScript. Laup records both views and compares five kinds of page content.
Illustrative page comparison
/solutions
| Signal | Weight | Raw HTML | Browser rendered | Visible |
|---|---|---|---|---|
| Visible text(characters) | 40% | 950 | 1,000 | 95% |
| Headings(elements) | 20% | 5 | 5 | 100% |
| Images(elements) | 20% | 4 | 4 | 100% |
| Structured-data blocks(blocks) | 10% | 2 | 2 | 100% |
| Tables(elements) | 10% | 1 | 2 | 50% |
Content visibility
Ratios are capped at 100%. Signals absent from the rendered page are excluded and the remaining subweights are renormalized.
95 × 40% + 100 × 20% + 100 × 20% + 100 × 10% + 50 × 10% = 93
Illustrative data for explaining the method. The page and values do not represent a customer result.
Four stages, with capture before interpretation.
The order follows real dependencies. A site must respond before Laup can discover pages, and the raw and rendered captures must exist before the score can be calculated.
01
Reach
Resolve the responding domain, inspect robots.txt and test whether deeper pages accept automated requests. The llms.txt check runs beside the access check.
02
Discover
Follow sitemap declarations, expand sitemap indexes and identify candidate pages. Product brands can fall back to product-line links found in site navigation.
03
Compare
Fetch up to 10 sampled pages as raw HTML and through a browser. Compare visible content, then inspect Product and homepage structured data in the raw response.
04
Score and explain
Calculate the score mechanically from captured signals. A separate written assessment is generated afterwards, and only when the crawl is reliable enough to support it.
Five signals, with one explicit N/A rule.
Every normal scan starts with the same nominal weights. Missing measurable data scores zero, except when a healthy crawl completes the schema check without finding product pages to sample. Product schema is then excluded and the remaining weights are renormalized.
Worked example
Illustrative software site
The crawl is healthy but finds no measurable product pages. Product schema is therefore not applicable, leaving 70% active weight.
94/100
Content visibility
Nominal weight 40%
93 × 40%
37.2 points
Product schema completeness
Nominal weight 30%
Not applicable
Excluded
Crawler access
Nominal weight 15%
100 × 15%
15.0 points
Discoverability
Nominal weight 10%
85 × 10%
8.5 points
Homepage schema
Nominal weight 5%
100 × 5%
5.0 points
Renormalized total
Product schema is excluded because the healthy crawl found no product pages to sample.
65.7 ÷ 70% = 93.9 → 94
40%
Content visibility
Compares raw HTML with the browser-rendered page across visible text, headings, images, structured-data blocks and tables. Each raw-to-rendered ratio is capped at 100%.
30%
Product schema completeness
Measures the fields exposed in Product JSON-LD on sampled pages. The direct-seller mode checks 10 fields; product brands use seven and are not penalized for missing offer data.
15%
Crawler access
Inspects 14 crawler identities. Only the three AI-search crawlers and four user-triggered agents affect the score; blocking seven training crawlers is treated as a policy choice.
10%
Discoverability
Awards 50 points for a sitemap, up to 30 for lastmod coverage, 15 for image-sitemap coverage and five for an llms.txt file. The signal is capped at 100.
5%
Homepage schema
Checks the raw homepage for Organization and WebSite structured data. Each contributes half of this signal.
The same standard, adjusted for what the site sells.
A manufacturer should not lose points for omitting a checkout price. A healthy software site without a product catalogue should not be treated as a broken webshop. The selected site type changes only the relevant Product-schema checks and page discovery behaviour.
Direct seller
Retailers and webshops
Product completeness is scored across 10 fields, including price, price currency and availability inside an offer.
Direct seller
SaaS and software
The same direct-seller scan runs. When a healthy crawl samples no measurable product pages, Product schema becomes not applicable instead of an automatic zero.
Product brand
Product brands and manufacturers
Offer fields are excluded, so Product completeness is scored across seven fields. Navigation can supply product-line pages when no product sitemap is available.
Product-brand scans also record where-to-buy and documentation signals. They remain informational observations and do not add points to the readiness score.
What a readiness score can support.
The score is calculated without manual grading or an LLM deciding what looks good. Its usefulness still depends on what the scanner could reach and the pages it sampled.
A technical proxy, not an assistant session
The browser comparison shows what disappears between raw HTML and a fully rendered page. It does not record ChatGPT, Gemini, Copilot or Claude browsing the site.
A sample, not a complete crawl
Laup inspects up to 10 candidate pages. The report exposes the sampled URLs and page-level findings so the result can be checked rather than treated as exhaustive.
Limited access remains visible
A WAF, challenge page or repeated fetch failure can make deeper signals unreliable. In that case the report is marked limited instead of presenting the number as a normal score.
A dated observation
Robots rules, sitemaps, rendered content and structured data can change after the scan. Re-running creates a new measurement of the site as it exists then.
The written assessment and recommendations are generated after the score from the captured facts. They do not alter the arithmetic. On limited scans, Laup skips that assessment rather than filling missing evidence with generic advice.
Before you rely on the score.
Does the readiness scan run ChatGPT or another AI assistant through my site?
No. It is a controlled technical scan. Laup fetches the site directly, compares raw HTML with a browser-rendered page and inspects the access, discovery and structured-data signals described on this page. AI Visibility is the separate measurement of what AI engines say when answering buying questions.
Why can Product schema be marked not applicable?
When the crawl is healthy and the schema check completes without finding product pages to sample, Laup cannot evaluate Product schema. It excludes that signal and renormalizes the remaining weights. A blocked, errored or unreachable crawl keeps the missing signal at zero because the scanner could not complete the check reliably.
Does blocking AI training crawlers lower the score?
No. Laup displays access for seven training crawlers, but blocking them is treated as a legitimate policy choice. The Crawler access score uses only three AI-search crawlers and four user-triggered agents.
How much does an llms.txt file affect the score?
It contributes five points inside the 10% Discoverability signal. When all five top-level signals are active, that is at most half a point in the final score. Laup reports llms.txt because it is useful context, not because it proves readiness by itself.
Does a high readiness score mean AI will recommend me?
No. A high score means the scanner found strong technical access, discoverability and machine-readable content under this method. Recommendation depends on a broader set of signals and is measured separately through AI Visibility.
Why can the score change when I run it again?
Every scan is a dated observation. Robots rules, sitemaps, sampled pages, structured data, JavaScript rendering and access controls can all change. A later run measures the site as it exists at that time.
How we measure AI Visibility
Readiness checks the site. AI Visibility measures whether engines name you when answering the buying questions you care about.
Read the AI Visibility methodologyHow evidence becomes an Action
The Action plan investigates likely root causes, ranks the work and makes clear whether the evidence is measured, inferred or still open.
Read the Action methodologySee what automated systems can retrieve.
Run the same technical checks on your own site. No account required.
Usually completes in about 30 seconds.