CrawlCheck, for writers and reviewers.
In one sentence
The CrawlCheck press page gathers what a writer needs to describe the product correctly: a one-sentence and one-paragraph description, dated facts and milestones, the live counts with the date they were read, the product's own vocabulary, the wordmark and mark as vector files, and the brand tokens.
The facts, the numbers with their dates, the words the product uses, and the marks — on one page so a description of CrawlCheck is right the first time, whether a person writes it or an agent assembles it from the media kit. Every line is something the site already publishes.
250,684
sites measured
143,426
real crawler visits logged
35
sections a report scores or reads
0
numbers a payment can change
Read live on 2026-10-12 UTC from the same counters the homepage prints. They regenerate; quote them with the date, and link the dataset for the current figure.
CrawlCheck is the verification layer for the agentic web. It measures what a site actually serves — fetched as GPTBot, ClaudeBot, PerplexityBot, Googlebot and eight other crawler identities, plus its machine files, crawl policy, declared rights and capabilities — decides findings by published rules, and signs the answer so an agent can verify it offline before it reads, cites, connects or acts. Site owners get the same measurement as a free A–F audit with every defect named.
The site’s own definition, served on the homepage and in llms.txt. Quoting it verbatim is preferred.
In one paragraph
CrawlCheck is the verification layer for the agentic web, built in Denver by VSNARY (Emmanuel Orta) and public since August 2026. Before an AI agent reads, cites, connects to or transacts with a website, CrawlCheck answers what that site permits, what it actually serves to agents (fetched as GPTBot, ClaudeBot, PerplexityBot, Googlebot and eight other crawler identities), whether its MCP tools behave as declared, and what to do about it, as one signed answer that verifies offline against a published key. Asking needs no key and no account. Site owners get the same measurement as a free A–F audit with every defect named; paid plans add a dated record, the evidence behind every finding, and the daily snapshot archive at volume. Its core is open source under MIT. No number it produces can be changed by a payment.
Facts on file
Each line links to the page that publishes it.
Name
CrawlCheck
Site
Built
7 August 2026, Denver, Colorado
Founder
Emmanuel Orta, creator name VSNARY — who runs this
Free
The signed answer about any domain (Resolve, preflight, receipts, the public snapshot), the full scan of any domain, and any single YouTube video check
Paid
Fix $749 once, with thirty days of monitoring · Watch $29/mo for 3 domains · Agency $249/mo for 40 with the API · API plans Developer $99/mo and Growth $499/mo · Evidence and Intelligence add-ons $49/mo each · Data add-on $199/mo (the 90-day snapshot archive, included in Growth) — pricing
For agents
Resolve protocol: one signed answer about a domain before an agent acts, verifiable offline against the keys at /keys; every rule at /rules, every dispute at /disputes
Open core
crawlcheck-core, MIT, dependency-free, parity-tested against production
Machine-readable
This page as data at /.well-known/media-kit.json (schema.org, the standard this site checks others for); the entity graph at /entitymap.json; every finding and guide at /feed.json; any page as Markdown with Accept: text/markdown
Crawler identity
Mozilla/5.0 (compatible; CrawlCheck/1.0; +https://crawlcheck.io/bot) — about 30 read-only requests per scan, from one address, identified honestly
Contact
hello@crawlcheck.io — read by the person who built it; there is no PR firm
Dated events
Each date is the one on the site’s own record.
- 2026-08-07Built. The date on the site’s own Organization record.founded
- 2026-08-15Cited by an answer engine. Perplexity reproduced the pricing table row by row and cited the homepage, the dataset and a findings post as sources — the product’s own thesis, on its own site.cited
- 2026-09-03Open core published. crawlcheck-core: the robots.txt reader, crawler identification and the log-line verifier, MIT, with parity tests against the live service.released
- 2026-09-03First editorial citation. A third-party AEO tool vendor wrote a piece around the /goto measurement and cited it as the primary source.cited
- 2026-09-05Mobile app design frozen. Version 1.0, 28 screens; the site began serving its design tokens and content manifest for it — the app.milestone
- 2026-09-17The app went live on the web. crawlcheck.io/app/run serves the design build against the live record; the phone's receipts route opened; the Android store package was built.milestone
- 2026-09-29Trust center, terms and release notes. /trust built from the running configuration; terms and a refund policy; the first release note, and every finding carrying a fingerprint and a lifecycle state.milestone
- 2026-10-01The verification layer went public. Signed fix receipts, the rulebook, the open observer protocol with a third network, the dispute ledger, the capability registry, a typed SDK and conformance fixtures; the census frame of 991 top-1,000 sites was frozen.milestone
- 2026-10-02Resolve protocol. One signed answer an AI agent reads before acting on a domain, with a change feed, outcome reports and a domain lifecycle record; signing keys published at /keys. A second outside security audit closed the same day.milestone
- 2026-10-03First census published. 991 of the Tranco top 1,000 read twice, every figure with its denominator and its change between runs — the dataset. Second release note.milestone
What it is not
Not a rank tracker
It measures what a crawler receives — the inputs — not where a page appears in anyone’s results.
Not an AI-content detector
It does not judge whether text was written by a model.
Not pay-to-grade
A licence unlocks lists, the record and cadence; it does not change a measurement, a finding or a grade. That sentence is in the pricing FAQ and has been quoted back to us by an answer engine.
The words it uses
Four tiers, named only on the pricing page. Not measured is never a zero.
- record
- not an account — the dated history of one domain
- reading
- one measurement
- finding
- a defect that holds a grade down
- receipt
- a dated, hashed change
- Reach
- can a named crawler get the page
- Read
- can it find the words
- Quote
- is there a fact it can state and attribute
- bands
- strong · adequate · weak · poor · not measured
Marks and colours
The wordmark is type: “Crawl” in ink, “Check” in the accent, set in Cabinet Grotesk. Vector paths, no font dependency. Use them to refer to CrawlCheck; do not recolour them, and do not present them as an endorsement.
Ink
#1A1916
Text, and “Crawl” in the wordmark
Accent
#006B6B dark #3DA8A8
“Check”, links, the mark’s short bar
Gold
#876810 dark #E6BE52
The site’s signal colour: labels, buttons, figures
Paper
#F5F5F3 dark #0F0F0E
Page background
Grade colours
#2D6A1F strong · #875F00 adequate · #AE430D weak · #A11B1B poor
Bands on every report
Type
Cabinet Grotesk · Satoshi · JetBrains Mono
Display · text · numbers and code
The same values are published for the mobile app at /tokens.css; the page and the app read one file.
Who runs it
Emmanuel Orta, whose creator name is VSNARY, builds CrawlCheck and writes its findings. He runs a Denver agency, and the sites he owns go through CrawlCheck — the public demo report is one of them.
The dataset · Proof and timestamps · How CrawlCheck behaves · Findings