I'm Emmanuel Orta, and VSNARY is my creator name. I am a digital marketing consultant: I get businesses found by Google, by Google Maps and by the answer engines — ChatGPT, Claude, Perplexity — and I built CrawlCheck, the verification layer for the agentic web: one signed answer about any site that an AI agent can check before it reads, cites or acts, and a free A–F audit any site owner can run. I learned the work on my own money, on a network of Denver home-service brands I own and rank myself, not on a client's. The second thing I do is conservation: I keep and breed threatened reptile species, and I document them where the record is public — Wikimedia Commons, Wikidata, a lineage registry with accession numbers.
This page is a record rather than a pitch. Every claim below is followed by the place you can go and check it, and where a credential would have helped me and I do not have one, it is simply not here. That constraint removed the testimonials, the award badges and the invented statistics — and it is the only reason the rest is worth reading.
The stack this is actually built and measured with
I am a digital marketing consultant for businesses that want to be found — in Google, on Google Maps and in the answer engines people now ask instead of searching. Four levers, no more: each one is something I do inside my own companies first, every week, and then for yours.
What ChatGPT, Claude, Perplexity and Google actually receive from your site — not whether they can reach it, but what it costs them to read it, and whether the answer differs by who is asking. Measured from outside, and from your access logs when you can give me them.
Profiles rebuilt and kept aligned with the site — categories, services, hours, the fields an assistant is actually asked for. At multi-location scale, where the usual failure is fifteen declared entities and one real listing.
The errors validators pass: a sameAs array that claims your business is fourteen cities, two nodes sharing one identifier, a keyword list where a topic graph should be. Every value legal, the meaning wrong.
When the measurement does not exist, I build it. CrawlCheck, SpeciesVault and CredentialRecord all started as something I needed for my own work and could not buy. Cloudflare Workers, WordPress, and whatever the job actually requires — so a consulting engagement comes with its own instrument, not a slide about one.
Emmanuel Orta lives between Arvada and Denver — both, really. VSNARY is his creator name: one name for the businesses, the sites and the software, and one person behind it.
A working day is three things in whatever order the day demands: creating, self-improvement, and the care and maintenance of the animals. He started six or seven years ago with a single site, treeservicedenverllc.com, and built outward from it.
What he is proud of is not a number. It is the person he has become: someone who never needs to work a job, because he creates them.
This is the portfolio, and it is also the consulting portfolio: every ranking, profile and machine layer here I built and hold myself. No logos, no “trusted by”, no client names. Each row is a live site you can open, score with any tool you like, and judge for yourself.
These aren't logos on a wall. They're companies I am inside of every week — building, ranking and tightening the systems until they compound.
ChatGPT, Claude, Perplexity and Google all send crawlers. Almost nobody checks what those crawlers were actually given. I built CrawlCheck because I needed the answer for my own sites and no tool would tell me.
It sends fifteen client identities to a homepage from one address inside one second and keeps every reply — status, bytes, content type, cache header — then reads the machine files, the structured data and the delivered HTML across 38 sections, 24 of them scored. The most common serious finding so far is a file returning 200 to every crawler while serving a verification page instead of the content.
The scan is free and stays free. What costs money is the record: a dated series that says what changed, which a single fetch can never produce after the fact. Plans run from $29 to $249 a month, with a Network tier from $690, priced by how many domains you keep a record of, and every report is hashed and anchored to Bitcoin daily so the record can be proved rather than remembered.
Websites say “licensed, insured, certified” by default — increasingly because a model wrote the page and nobody read it back. There is no public place to check the claim against the register that issues it.
CredentialRecord publishes the lookup, never the conclusion: the claim as published, the registry consulted, what it returned, and the date. No business can ever pay for an outcome.
Tree, fencing, concrete, excavation, landscaping. I own the domains, the rankings and the phone numbers; crews do the work and get paid for it. This is not a client roster — it is a portfolio.
Lead generation is the half I am good at and the half that is cheap to replicate. Capacity is what decides how much of it turns into money. A ranked site attached to a crew that answers the phone is worth many times the same site attached to nobody, and I have run both.
A specialty breeder of reptiles and invertebrates raised in bioactive setups. Low volume, high value, mostly referred — the sort of trade where a buyer is paying as much for the record as for the animal.
My photographs illustrate the Wikipedia articles for Gonatodes daudini (on six language Wikipedias, Wikispecies and Wikidata) and Cnemaspis chanthaburiensis. Terrarium Station is named as the collection in the daudini infobox on English Wikipedia, and on the Commons file record. The images themselves were adopted by editors I have no connection to, on wikis I do not edit — which is why that use is worth more than anything I could write about myself.
An F-number means nothing unless the chain behind it exists. Most of this trade runs on a claim in a forum post: someone says an animal is F3 and there is no way on earth to check it.
SpeciesVault gives every animal an accession number, links sire and dam as records rather than as text, and derives the generation by walking that chain. A number that cannot be derived renders differently and says so. It exports Darwin Core, so an institution can read it without asking me for anything.
The measurement is the centre: a site fetched as fifteen client identities, GPTBot, ClaudeBot, PerplexityBot and Googlebot among them, its machine files, crawl policy and declared capabilities read, findings decided by published rules, and the answer signed so an agent can verify it offline before it reads, cites, connects or acts. Around it sit the things built because the measurement kept needing them — a second vantage from your own browser, the open core, an API and SDKs, a glossary, a library of dated findings, and the monitoring that turns one measurement into a record. Each one below is a link, and the counts are read live.
Two green anoles from a PetSmart, at eight years old. The collection came first and the business came after; Terrarium Station is a lifestyle that was turned into a company, not a company that acquired a hobby. What it became is a conservation practice: threatened species kept and bred in captivity, each animal on a lineage register with an accession number, and the photographs placed where the public record lives.
The register behind it is SpeciesVault: accession numbers, derived generation, Darwin Core export — the same fields a museum collection carries, applied to a living one. The photographs go to Wikimedia Commons under one account, Vsnary, where Eight of them are in use on 18 pages across 9 projects, re-counted hourly from the Commons API.
Some of the animals in it are the only photographs of their species on Wikimedia Commons. On 9 September 2026 he published the first photograph of a Gonatodes daudini hatchling recorded on Wikimedia Commons, Wikidata, iNaturalist or GBIF — a Critically Endangered gecko, hatched here, photographed minutes old on a fingertip. It is now the species’ juvenile image on Wikidata and in its Wikipedia article. Those are the claims in this section that are not his own word, and each is checkable at the link.
| Species | IUCN status | Provenance | Record |
|---|---|---|---|
| Gonatodes daudiniUnion Island gecko | Critically Endangered | bred here · first hatchling image on record | Q3110413 |
| Phelsuma klemmeriNeon day gecko | Endangered | bred here | Q882598 |
| Correlophus ciliatusCrested gecko | Vulnerable | bred here | Q24276246 |
| Varanus kordensisBiak tree monitor | Data Deficient | captive-bred | Q2013825 |
| Gekko pradapdao | Not assessed | — | Q107283404 |
| Cnemaspis chanthaburiensisChanthaburi rock gecko | Least Concern | breeding pair, hatched here | Q2980830 |
| Gonatodes annularisAnnulated gecko | Least Concern | — | Q3110407 |
| Gehyra marginataHalmahera giant gecko | Least Concern | — | Q3100138 |
| Chlamydosaurus kingiiFrilled dragon | Least Concern | captive-bred | Q501566 |
This is the work I sell, shown on myself. An answer engine does not read a brand; it resolves a node. One Person identifier, one Brand, one Organization per business, one identifier per product — declared once, reused by every site that mentions them, and published in the machine files as well as the page. The table is the graph in the <script type=application/ld+json> above, row for row.
@id, reused by crawlcheck.io and terrariumstation.com rather than re-declared; twelve profile edges; every organisation with its own canonical identifier; and the machine files that publish the same graph. Nothing here is a second copy that can drift — the files are built from the constants the page is built from.https://emmanuelorta.com/#personthis page; reused verbatim by crawlcheck.io/author and terrariumstation.comhttps://emmanuelorta.com/#vsnarythis page; every organisation below points brand → herehttps://emmanuelorta.com/#consultingthis page; provider → #personhttps://crawlcheck.io/#orgcrawlcheck.io, every page; founder → #personhttps://crawlcheck.io/#appcrawlcheck.io; offered by #orghttps://crawlcheck.io/extension#appcrawlcheck.io/extension; install URL on the Chrome Web Storehttps://github.com/emmanuelorta/crawlcheck-core#codeGitHub, MIT; version and CI read hourlyhttps://crawlcheck.io/docs/api#apicrawlcheck.io/docs/apihttps://crawlcheck.io/glossary#setcrawlcheck.io/glossary; 768 DefinedTerm nodeshttps://crawlcheck.io/data#datasetcrawlcheck.io/data; 4,111 scanshttps://crawlcheck.io/directory#datasetcrawlcheck.io/directoryhttps://crawlcheck.io/pricing#catalogcrawlcheck.io/pricinghttps://credentialrecord.com/#orgcredentialrecord.comhttps://terrariumstation.com/#organizationterrariumstation.com; foundingDate 2022-08-04https://terrariumstation.com/nature-database#speciesvaultterrariumstation.com/nature-databasehttps://github.com/emmanuelorta/PhotoFind-iOS#codeGitHub, Swift, MITone @id per domain, trading names onlyeach site; legalName only where the registered name is the trading namethe Commons file URL is the @idWikimedia Commons; creator and copyrightHolder → #personNo vague retainer and no discovery theatre. The first step produces something you can read whether or not we ever work together, and the last one is the only part a single audit can never give you.
Your site is fetched from outside your network as fifteen different crawler identities inside one second. You get a graded report showing exactly what each one received — status, bytes, how much of the payload was readable text, and which machine files answered. Free, no account, and yours to keep even if the conversation ends there.
The findings get corrected, not forwarded. Schema that validates but asserts the wrong thing, a robots rule silently voiding itself, a listing pointing at a page that costs a crawler 300KB to read a paragraph. Then the site is re-measured, so the fix is a before and an after rather than an assurance.
Sites break quietly: a plugin update, a cache rule, a host toggle. Weekly re-measurement turns “it is broken today” into “it broke on the 14th, and here is the receipt”. That record cannot be bought after the fact at any price — it exists only if something was already watching.
Not a story arc. A list of dates, each with the place that records it — a registry, a page that recomputes itself, a register that says when it opened.
treeservicedenvercolorado.com, the domain the tree business traded on before it moved. It still redirects, path for path.
Verisign registry recordThe date the site itself declares as its founding, and the date the shop has been keeping records since.
terrariumstation.comThree brands, one crew, and the first time the model was run at more than one listing at once.
Rows 07–09 in the registerTree Service Denver stops relying on third-party review widgets and records its first first-party review through its own moderated system.
The reviews pageBuilt because I needed to know what my own sites were handing a crawler, and no tool would tell me.
crawlcheck.ioThe author page counts every piece since, with dates, and recomputes the count when it is opened.
The author pageA register that states the lookup and never the conclusion. Its own manifest records the opening date.
credentialrecord.com/llms.txtOne quote request on treeservicedenverllc.com carried utm_source=chatgpt.com. One lead is not a rate, and the note says so — but it is the first outcome on a record that had only measured inputs.
The readers at the centre of the measurement — robots.txt policy, user-agent group shadowing, crawler-IP verification — extracted verbatim from production and put on GitHub with their tests. The repository records its own first commit.
The repositoryHatched here from adults in the collection and photographed minutes old. As of that date no other juvenile image of the species existed on Wikimedia Commons, Wikidata, iNaturalist or GBIF; the file page carries that dated, platform-scoped sentence and the upload timestamp.
The Commons fileA second vantage: what a crawler is served from your own address, after your JavaScript runs. At launch it sent nothing back; since 1.0.3 the grade badge sends the site’s hostname and signed readings are shared, each behind its own switch. The store listing carries its own publication date.
Chrome Web StoreA versioned public API with a schema, a trust center that states retention from the code, and every finding carrying a state — new, persisted, worsened, resolved — across scans. The release note is dated and lists each one.
The release noteOne signed answer about any domain, read by an AI agent before it acts, specified as the Resolve protocol with a typed SDK and conformance fixtures; the first agentic-web census of the Tranco top 1,000; every rule published, every fix ending in a signed receipt, disputes resolved in public. The release note names where each one can be checked.
The release noteMost personal sites ask you to believe a number. This one gives you the third-party page instead — and names the one claim that has nothing behind it, because leaving that out would defeat the exercise.
I own and operate eleven live domains.
Each one is listed above with its URL. Open any of them; there is no logo wall standing in for a link.
The registerI built CrawlCheck, and it is a working product.
Put any domain into it. The free scan runs without an account and returns a graded report you can read in full. The findings from 4,111 of those scans are published as an open dataset that recomputes itself on every request.
crawlcheck.ioI publish research, including my own mistakes.
Ninety-seven notes, each one a measurement with the method attached. Several of them document defects in my own instrument — an audit that labelled one page and measured another, a forgery rate that was counting my own extension.
The writingMy photographs illustrate Wikipedia articles in six languages.
Eight own-work images on Wikimedia Commons (user Vsnary) are in use on 18 pages across 9 projects: the Gonatodes daudini article on six language Wikipedias plus Wikispecies, the crested gecko on Cebuano Wikipedia, Arborist, Synthetic fence and Chanthaburi rock gecko on English Wikipedia, and five Wikidata items. Terrarium Station is named as the collection on the daudini file record; the full usage list is on Commons.
My Commons uploadsThe tree business is real work done by a real crew.
Tree Service Denver is operated by Eric Petruska. The site carries first-party reviews submitted by customers, not testimonials I wrote.
Customer reviewsEvery CrawlCheck report can be proved to have existed, unaltered, on its date.
Each report is reduced to its measurements, hashed, and each day’s digests are anchored to the Bitcoin blockchain. That proves the report existed in that form at that time — not that the measurement was right, and the page says which claim is which.
Timestamp proofEverything I have written is dated, counted and listed on one page.
97 findings, sixty-seven guides and the glossary, with first and latest dates, recomputed from the post registry each time the page is opened. The pieces about my own mistakes are listed on purpose.
The author pageThe extension is published, not planned.
The CrawlCheck browser extension is listed on the Chrome Web Store under the CrawlCheck name, with its own listing page, permissions and publication date. Install it and read the network tab: it sends nothing back.
Chrome Web StoreMy developer identity is public, under one name.
The GitHub profile carries this same register, and the public repositories are listed there. Code that is private is described as private, not left out.
GitHub profileThe verification layer is open at its core.
crawlcheck-core publishes the robots.txt policy reader, the user-agent group shadowing detector and the crawler-IP verifier as MIT-licensed, dependency-free JavaScript with no build step. The test suite runs in CI on every push — passing at the last hourly check —, including byte-parity checks against the production service’s own live endpoints. Clone it and check my results against my code.
crawlcheck-coreI hold no certification I have not earned.
There is no ISA arborist claim, no agency award, no press badge on this page. Where a credential would help me and I do not have it, it is absent.
Nothing to link — that is the pointEvery note is a measurement with the method attached, so you can run it yourself and disagree with me. Several of them are corrections to my own work, which is the only reason to trust the rest.
Seven findings from the open dataset: the most common defect is a stale cache, a quarter of sites carry a grade-capping finding, opt-outs travel by copy-paste, and a third of crawler traffic is forged.
2 Sep 2026Read →FindingA quote request on a Denver tree-service site carried the tag ChatGPT appends to links it hands out. One lead is not a rate, but it is the first outcome point on a record that had only measured inputs until today.
2 Sep 2026Read →FindingThe same page that answered a nonexistent path answered /llms.txt. One layer of our scanner noticed; the other called it present.
1 Sep 2026Read →FindingA customer's schema pointed at a page on our own domain. From outside it answered 200. From inside the scanner it answered nothing, and we reported it dead.
1 Sep 2026Read →FindingA self-audit rule that scans every stored record for the token NaN found three of them. All three were inside the letters of one URL slug.
1 Sep 2026Read →FindingExperience, expertise, authoritativeness and trust are judgements an evaluator makes. A fetch cannot make them. It can check whether the evidence for them exists in a form a machine can find.
1 Sep 2026Read →Ninety-seven notes so far, plus an open dataset of what 4,111 scanned sites actually returned. No gated PDFs and no email wall.
Read all 97I don't build content. I build assets that compound.
A retainer ends the month you stop. A ranked domain with a tracked number attached is worth something whether or not I am working on it today. I would rather own one asset than bill for ten.
If a number cannot be linked to the thing that produced it, it does not go on a page. That rule cost this site its testimonials, its award badges and its invented statistics, and the page is better for it.
Captive-bred means captive-bred. ISA member is not ISA certified. Representative photo means representative photo. The businesses that last are the ones that never had to walk a claim back.
A diagnosis handed over with nothing done about it is not a deliverable. If the fix is within reach, it ships before the message does — and the message reports both.
Emmanuel Orta, VSNARYWritten and maintained by me, between Arvada and Denver. If anything here stops being true, the page is wrong and I want to know.
A small number of engagements at a time, because the work is done by me rather than passed to a team — and every one starts with a measurement you keep whether or not we go further.
If you are in a truck more than at a desk: when someone nearby searches or asks an assistant for what you do, you come up, you look legitimate, and the phone rings. I run this playbook on my own brands first.
Start with the free scan — it shows what a crawler sees before we talk. Scope and pricing on the first call.
You already sell GEO, AI SEO or local search. I supply the instrument and the hands, white-labelled under your brand.
Scope and pricing on the first call. Monitoring across a roster is priced publicly at crawlcheck.io/pricing.
Correspondence
Agency looking for the layer under your pitch, a local business that wants the phone to ring, or a keeper with a problem SpeciesVault might solve. One form, one reply, from me.