An Organization object in JSON-LD can make a brand's identity easier to inspect. It cannot prove that an AI service will cite the page or rank the brand above a competitor. A useful technical audit therefore needs a narrower question: which machine-readable facts were visible in the permitted static response, and which contradictions or gaps should an editor review?
Schema.org for LLMs Audit build 0.1.4 reads supported public page facts and produces a deterministic readiness review. It does not call an LLM, render JavaScript, log into a site, or measure answer-engine behavior. Its score belongs to the audit rules, not to a search or citation experiment.
Submit the expected identity explicitly
An illustrative input checks a public page against the brand name you intend to find:
{
"urls": ["https://apify.com/"],
"expectedBrandName": "Apify",
"brandAliases": [],
"maxConcurrency": 1
}
The example defines the request shape; no fresh grade is asserted here. The schema accepts one to fifty public URLs, optional expectedBrandName, up to twenty accepted brand aliases, and concurrency from one to ten. Fragments and exact duplicates are removed before collection.
The expected name is used for deterministic identity-consistency checks. It is not sent to an AI service to infer what the company probably meant. Use aliases to express spellings that your organization actually accepts, rather than widening the list until an inconvenient mismatch disappears.
Interpret a readiness finding as a page observation
The collector compares supported JSON-LD, Schema.org Microdata and RDFa, OpenGraph, canonical metadata, and expected brand identity from initial HTML. The result can describe an entity inventory, gaps, contradictions, and prioritized fixes for supported organization, product, and article observations.
readinessScore ranges from zero to one hundred, with grades A, B, C, D, or F. Those values summarize this product's deterministic checks. They are not a complete Schema.org validator, a universal GEO score, or a forecast of AI referrals. Keep the underlying findings beside the score when discussing a fix with an editor.
Observation identity matters too. auditId includes a timestamp and should not be used as the permanent join key for a page across repeated observations. Keep a normalized page identity and the particular audit observation separately, so a changed score does not create a fictional new page.
Keep collection diagnostics outside successful audit counts
The runtime respects robots.txt and checks origins reached through redirects. It blocks private-network targets. The installed fetch limits include 750 KB of HTML, 100 KB of robots content, a twenty-second timeout, and up to five redirect hops. A denied or unsafe URL is not a low-readiness page; it is a collection diagnostic.
Successful fetched audits go to the Dataset and can be billable even when the page receives Grade F. The paid service is the completed audit, not a good score. Robots denials, unsafe targets, failures, and other non-delivery diagnostics are recorded as bounded free KVS evidence rather than paid pseudo-results. Retain both the Dataset and the diagnostic scope when reporting batch coverage.
For acceptance, use a controlled page with an explicit Organization identity, a page with a deliberate name mismatch, and a denied target. Confirm that the mismatch is inspectable and the denied target stays outside completed audits. These are proposed checks; they are not newly observed runtime results.
After an editor fixes a finding, re-observe the page under comparable conditions and compare the specific evidence. Separately measure any citation, traffic, or conversion change over a defined period. That separation lets a team report an actual technical correction immediately while waiting for evidence about the commercial outcome.
Cost and source boundary
The live FREE-tier event prices checked on 20 October 2026 are $0.03 per result-found event and $0.02 per automatic Actor-start event. The start event is memory-scaled: one event per GB, with a minimum of one event. A free result outcome therefore does not make an entire run free. Other account tiers have their own rates; check the live pricing panel before execution. These prices describe the installed contract, not measured customer revenue or a new runtime test.
Inspect the Actor, input, and current pricing.
AI-assisted drafting; technical statements reviewed against the installed Actor source and current API metadata. Examples are illustrative unless explicitly identified as saved observations.
Top comments (0)