denverpost.com

JSON-LD is an RDF serialisation, so every schema block above is already a graph of triples. This section reads it as one: 68 statements about 0 identified subjects, 1 reference between them, none dangling. The triple count is approximate: each property value is one statement and each @type one more; nested objects without an @id are counted as values, not expanded. Weight 0.8 in the overall, below the schema section it overlaps.

What this can and cannot say. E-E-A-T is a judgement an evaluator makes about experience, expertise, authoritativeness and trust; it is not on the wire and this scanner does not pretend to score it. Each row above is one machine-readable statement a site makes that a reader could use as evidence toward that judgement, measured from the delivered homepage and its JSON-LD. A full row of passes does not create expertise, and a row of fails does not disprove it; what they decide is whether a machine reading this site can find the evidence at all. Weight 1.0 in the overall, the same as payload: what these rows measure is whether the evidence is findable by a machine, and that is the part this scanner can stand behind.

What this is not. Not a WCAG audit: contrast, focus order, keyboard traps, motion and the experience of using the page with assistive technology need a rendered page and a person, and none of that is measured here. These rows read the document the server returned and ask whether the structure a screen reader depends on is present at all. Heading order, zoom lock and link text are scored in their own sections and are not counted again here. Weight 0.8 in the overall. Counts in the Read stage from scoring version 26, because parsers find content through the same structure a screen reader uses; reports issued earlier kept it outside the stages.

3 corroborated · 0 unreciprocated · 0 unverifiable · 0 dead. A sameAs link is an identity CLAIM, and an identity claim can be reciprocated — the target either points back at this domain or it does not. Platforms that refuse datacentre readers are counted as unverifiable, never as failures: the claim can be neither confirmed nor accused from here, and pretending otherwise would manufacture findings. Corroborated claims pass, unreciprocated and dead claims fail, and unverifiable or declared-only claims are left out of the score — so a profile page that blocks readers can never cost a point, and a link that is dead or never mentions this site does.

Sentences an engine could lift as they stand:

• “Here’s what to expect in the final 3 years.”
• “Much of the I-70 Floyd Hill Project is taking place above drivers’ heads.”
• “Here’s what to expect in the final 3 years.”

The lead, as delivered: “Sign up for Newsletters and Alerts Sign Up”

What this scores, and what it is not. These rows read the delivered HTML for the shape of quotable text: nothing here judges whether the copy is good, true or wanted. Counts come from body paragraphs and list items with the navigation, header, footer and forms removed; a page with fewer than five is not scored here. Thresholds were set from a 176-homepage corpus pass and each row states the corpus figure it was set against. Weight 0.8 in the overall and part of the Quote stage, from score version 15.

A crawler arrives with a budget. Every row here is a fetch that produced no new page: a redirect, a parameter variant, a case variant, a declared page that points elsewhere. Counted from the delivered homepage, the sitemap, the sampled declared URLs and the daily whole-site read where one exists. None of this moves the grade.

A document does not compete for a keyword alone; the index links it to entities by their Knowledge Graph identity and ranks documents against each other for that entity (the Webref layer in the recovered Geostore schema). This section reads the structured data on the delivered homepage and asks the question that layer asks: which business is this page about, and how is every other business on it related? Chain membership, subsidiary, department and containment are declared relations; a second business with no relation is a claim the engine must arbitrate. Measured, not scored, until the corpus shows how often each verdict occurs.

Measured on this exact URL. These are the 75th percentile of what real Chrome users experienced over the last 28 days, not a test run from here — the thresholds are Google’s published ones, the only ones that are.

Where the bytes go

1 of 1 map identity links point at places, not this business — 5990 Washington St, Denver, CO 80216. sameAs asserts that two URLs describe the same entity.

robots.txt ✓ → sitemap ✓ → llms.txt ✗ → entitymap.json ✗

A ✗ breaks the chain at that point: everything downstream is only reachable by a crawler guessing the conventional path.

Checked against the page’s own visible text, with no extra request. Node type: NewsMediaOrganization.

This reports EXPOSURE, not intent. Most hidden text is an old SEO habit or a collapsed menu, and a match here is a prompt to go and read it — not a finding that someone attacked the site.

One URL, fifteen fetches in the same second — an unnamed client, twelve named crawlers, a mobile browser, and the browser again as a control. The control fetches agreed (self-similarity 1), so text differences below are differences, not noise. A user-agent is a claim, including when we are the one making it.

Per-engine retrieval. Each answer engine reads through named crawlers with different jobs — one builds the index, one fetches live when a user asks, one collects training data. They are separate permissions and a site commonly grants one and refuses another. Nothing here is scored: these same facts are already scored once above, and this measures what an engine is permitted and given — never what a model has retained or would cite, which no scanner can see from outside.

A blocked training column beside an allowed index column is coherent policy, not a defect: it says "answer with me, do not train on me." The last column is whether your llms.txt or agents.md addresses that engine by name — robots.txt is a permission, those two files are where an operator actually talks to an agent. Naming nobody is the norm and is not scored.

no vantage provider is configured and no extension reading exists for this site in the last 14 days, so this was not measured

Checked because this site declares a local business entity (LocalBusiness). This measures the site side of Google Business Profile alignment — whether the business points at its own profile and carries the fields a profile is matched on. It does not read the profile itself: that needs an API key, and inventing facts about a listing we cannot see would be worse than reporting nothing. Profile link found: https://maps.app.goo.gl/aK261HiEe923wgBV7.

Every profile below is one the site itself declares in sameAs. Nothing here was discovered by guessing at directories — undeclared listings need an index this scan does not have.

Directory coverage

Ten sources an answer engine is likely to reach for when asked about a local business, checked against what this site declares. Declared means the site names the profile in its own structured data — the only thing this scan can verify. A listing that exists but is not declared will read as missing here, and that is itself worth fixing: an engine reading your site has no way to find it either.

9 of 10 are not declared. Each one is a place a retrieval system could have found a second, independent statement of your name, address and phone — and the agreement between those statements is what makes any of them trustworthy.

Machine layer over time. Our own history of a site starts the first time we scanned it; the Internet Archive holds what came before. Nothing here is scored — a site’s past is not a defect, and the Archive’s coverage is uneven, so a missing snapshot says nothing about the site.

This compares the sitemap against the links on the homepage only, so it finds pages the sitemap omits — it cannot prove a declared page is unreachable, because that needs a full crawl. The reverse gap (declared but linked from nowhere) is real and is not measured here. A page missing from the sitemap is still findable by following links; a page missing from both is findable by nothing.

What an engine can learn from the pictures. Five rows score: missing alt, declared dimensions, nothing lazy-loaded above the fold, an entity image in JSON-LD and an og:image. Empty alt, file format and hero priority are reported and never scored — empty alt is correct on a decorative image, and format is a preference, not something an engine fails to read.

CCBot is blocked by robots.txt, so future crawls will skip this site. Checked against September 2026 Index, from a stored result.

0 of 6 pages checked were captured.

A capture means the page is in an open dataset. It does not establish that any model trained on it, and nothing here claims otherwise.

the deep entity read runs for domains under continuous record; this scan used the keyless resolver

The identity this site publishes is unchanged since 2026-09-24. Same name, same phone, same address, same declared profiles.

Fingerprint 8153bbf8989417ea971c5250c7185c50… · recorded 2026-09-24

Schema type newsmediaorganization. We hold no population figure for this type, so no rarity is claimed.

A fingerprint of the identity this site publishes, recomputed on every scan. A change here is a change the site made, not a judgement about it. Anyone can recompute this from the published page: the canonical form and the hash are both public at /api/entity-drift/latest.

Percentiles come from sites this scanner has measured itself, not from a published study — so they describe this corpus, not the web. The corpus is not a random sample and is weighted toward sites that were submitted or seeded, which is why the rank is shown next to the raw number rather than instead of it. A metric is left blank rather than ranked when fewer than 8 peers carry it. This rank is against the wordpress/publisher cohort (60 sites), not the whole corpus of 866, because a WordPress local-service site and a static personal site are not the same measurement problem.

Comparing against the last reading that differed, taken 2026-09-24T14:10:19.087Z.

History keeps the last 24 changes on this domain — repeated identical readings collapse into one entry that records when the state began and how many scans confirmed it, so re-running a scan can never push real history out of the window. A finding that comes and goes is the pattern worth acting on: a permanent failure gets noticed, an intermittent one only shows up if something is looking at the right moment.

Measured on 7 separate days between 2026-08-20 and 2026-09-24. 1 defect has stopped being measurable and 8 are still open.

Headline measurements, first reading against latest

Defects that stopped being measurable

Median time from first measurement to no longer measurable: 21 days.

Still open

This section is reported and never scored: a site that has improved a lot and a site that was always fine should get the same grade for the same measurements today.

Not measured here, and not measurable here. Everything else in this report was produced by fetching your pages from outside. Clicks, enquiries and orders happen where no scanner can see them — in your Search Console, your inbox and your checkout. Connect one of those and this section fills in against the dates above. Until then it stays empty rather than showing an estimate, because an invented traffic figure on a report that measures honesty would be the whole product undone.

Every row in this report is either a measurement of your site or nothing at all. These lookups did not return a measurement, so they were left unscored rather than counted against you. They are listed because a silent absence is indistinguishable from a clean result, and that is the difference this tool exists to keep.

This covers the third-party and enrichment lookups only — CrUX, the Google profile link, the Internet Archive, the geocoder, and the two well-known files. Your own server’s refusals are never listed here: a 403 to GPTBot is the finding, not a gap in our instrument, and it is reported as a finding above.

Not scored, and therefore excluded rather than counted as zero: Agent surfaces (emerging, not scored), Content entities (measured, not yet scored), Service area, as declared (measured, not scored), The whole site, not just the front door (measured, not scored), Freshness, as the site declares it (measured, not scored), Local intents, the Maps vocabulary (measured, not scored), Conversion path on the page (measured, not scored).