ModularDot
PUBLIC SPECIFICATION · MODULARDOT-DETERMINISTIC-V2 · 2.0.0

Sixteen checks.
No crystal ball.

The ModularDot score is a reproducible description of explicit public inputs under a declared deterministic method. It is not a prediction of rankings, citations, model preference, traffic, conversion, or revenue.

HOW A SCAN IS PRODUCED

One retrieval, two root files, sixteen records.

  1. Normalize a public HTTP or HTTPS URL.Credentials, localhost, private IP ranges, and internal hostnames are rejected.
  2. Retrieve the page safely.The scanner follows at most three public redirects, requires an HTML response, and reads at most 1.5 MB. It analyzes returned markup and does not execute client-side JavaScript or perform a browser session.
  3. Check page HTML and root-file reachability.Most checks inspect the returned page. Access checks separately test root /robots.txt and optional /llms.txt.
  4. Produce a record for every check.Each stable ID returns pass/fail, possible points, awarded contribution, recognized excerpt plus element or URL when practical—or a failure reason and suggested fix.
  5. Sum four weighted groups.Identity contributes 30 points, access 20, proof 25, and decision support 25. No model-generated opinion changes the score.
IDENTITY · 30 POINTS

Can the page identify itself?

8 POINTS · identity.title

Specific page title

Observes
The text inside the HTML <title>; passing length is 15–70 characters.
Evidence
The recognized title and the title element. Failure reports the observed length.
7 POINTS · identity.meta_description

Useful meta description

Observes
A standard meta description; passing length is 70–180 characters.
Evidence
The description excerpt and meta[name="description"]. Failure reports the observed length.
7 POINTS · identity.primary_heading

One clear primary heading

Observes
The count of HTML <h1> elements; exactly one passes.
Evidence
The recognized heading and h1 element. Failure reports the number found.
8 POINTS · identity.structured_entity

Machine-readable entity type

Observes
JSON-LD containing one supported responsible entity, offer, application, event, or creative-work type.
Evidence
The matched @type and JSON-LD element. It does not validate every property or prove the claims.
ACCESS · 20 POINTS

Can public systems retrieve unambiguous routes?

7 POINTS · access.robots

Robots file reachable

Observes
Whether a GET request to root /robots.txt returns an OK response.
Evidence
The resolved robots URL. It does not interpret every crawler directive.
2 POINTS · access.llms_summary

Experimental LLMs summary

Observes
Whether root /llms.txt returns an OK response.
Evidence
The resolved file URL. This is deliberately low weight and is not presented as a ranking factor.
5 POINTS · access.canonical

Canonical URL declared

Observes
An HTML canonical link and its declared href.
Evidence
The declared URL and link[rel="canonical"]. It does not prove search engines selected it.
6 POINTS · access.open_graph

Share metadata present

Observes
At least one Open Graph title or description meta tag.
Evidence
The recognized content and Open Graph element. It does not test social-preview rendering.
PROOF · 25 POINTS

Can a visitor inspect responsibility and evidence?

10 POINTS · proof.attributable_evidence

Attributable evidence visible

Observes
A linked evidence artifact or language naming sources, methodology, case studies, support, funding, or comparable attribution.
Evidence
The qualifying link or text excerpt. Generic “reviewed” language does not pass.
6 POINTS · proof.responsible_identity

Responsible identity visible

Observes
Natural-language operator wording such as “operated by” or a clear About, team, company, organization, story, or mission statement.
Evidence
The recognized visible excerpt. It does not independently verify the named identity.
5 POINTS · proof.contact_path

Actionable contact route

Observes
A visible email address, mailto: link, or clearly linked contact, request, inquiry, or enquiry destination.
Evidence
The link text, address, and destination where available. Vague “email us” wording alone does not pass.
4 POINTS · proof.public_context

Enough public context

Observes
At least 250 scanner-visible words after script, style, and HTML tags are removed.
Evidence
The counted words and body element. Quantity does not establish usefulness or accuracy.
DECISION SUPPORT · 25 POINTS

Can a visitor decide what to do next?

8 POINTS · decision.pricing_or_action

Pricing, participation, or next step

Observes
An explicit EUR, USD, or GBP price, or bounded fit-check, quote, booking, application, or case-by-case language paired with an actionable route.
Evidence
The price or bounded-action excerpt and destination when available.
6 POINTS · decision.faq

Questions answered directly

Observes
Visible FAQ, “frequently asked,” or “questions answered” heading, or FAQPage JSON-LD.
Evidence
The qualifying heading or JSON-LD match. Presence does not establish answer quality.
6 POINTS · decision.fit_context

Comparison or fit context

Observes
Visible suitability, non-fit, comparison, alternative, “when another option is enough,” or similar decision language.
Evidence
The recognized visible excerpt. It does not judge whether the comparison is fair.
5 POINTS · decision.policies

Basic policies linked

Observes
A visible link whose URL or text refers to privacy or terms.
Evidence
The link text and destination. It does not provide legal review of the policy.
DELIBERATE NON-CLAIMS

What the scanner does not establish.

No third-party outcome is promised.

The scanner does not query or rank model answers; predict citations; measure search placement, traffic, leads, conversion, revenue, buyer behavior, or model preference; audit factual accuracy; provide legal compliance review; execute client-side JavaScript; or prove that detected copy is persuasive. Human review can qualify the pattern results, but the paid Fix Pack still does not guarantee any external outcome.

Run the scanner →Inspect an annotated scan →Review the Fix Pack scope →