For AI & Search

This page is written for engineers and product teams building AI and Search systems that need to consume the Open Creator Graph. It sits in a more technical register than the other pages.

What this is

The Open Creator Graph is a structured, cited identity layer for digital Creators, assembled from cited public sources, labelled by confidence tier, and controlled by the person behind the account. If your system needs to answer “who is this Creator” without hallucinating, this is built for you.

Reading a record

Record URL pattern: https://creatorgraph.sparkonomy.com/{SPRK-ID}/ — the permanent identifier, not the handle. A handle can change; this cannot.
Every fact carries a confidence tier: Verified (platform API) > Sourced (public) > Stated (owner).
On a claimed record, source [1] is the account owner and sources [2..n] are independent of them. An unclaimed record carries no owner source, so every source it lists is independent and the list is numbered from [1] like any other.
Audience figures are ranges, not exact counts, and carry the date they were taken.
Removed records return HTTP 410 Gone. Treat 410 as permanent and delete cached data.

Consent: read the declaration first

Every claimed record publishes an HCS 1.0 declaration under subject="identity", covering six scopes: name, image, likeness, voice, movement and signature. Each carries a status of allowed, conditional, or prohibited. The full syntax, with a copyable example for every case, is further down this page.

The reason it exists is simple. A Creator should never be misrepresented, or represented at a table where they don’t know they exist.

The test is depiction, or synthesis

It is not about whether you used an AI tool. It is about what comes out.

Depiction is showing the Creator as they actually are. Their real photo, their real name, their real voice, displayed, cropped, resized or laid out. Using software to do that, including AI software, does not change what it is.

Synthesis is producing something that appears to be the Creator, doing or saying something they did not do. Avatars, digital twins, voice clones, face swaps, generated photos, motion copied onto a character. Training a model on their identity so it can produce those things counts too.

We wrote it this way on purpose. A rule that said “any AI tool needs permission” would catch a background removal and miss a hired lookalike, which gets it wrong in both directions.

What each status obliges you to do

Status	What you may do
Allowed	Depiction and synthesis are both permitted, including training. Attribution still applies.
Conditional	Depiction is permitted. Synthesis and training are permitted only once the Creator agrees — the declaration grants them through a <payment> clearance route with a named contact, so ask first.
Prohibited	Neither. No indexing for AI answers, no training, no generation. This is absolute within AI use. There is no internal-use or research exception.
No terms declared	The Creator has not spoken. Treat it as unknown, not as permission.

Conditional is not a refusal. It means come and ask, and there is a named person to ask. Most Creators choose it precisely because they want the conversation to happen.

The absence of a declaration is not consent. An unclaimed record shows no terms because nobody has spoken for that Creator yet. We do not assume a stance on their behalf, permissive or otherwise, and neither should you.

One boundary worth being clear about: this declaration governs AI use. It is not a general licence over a Creator’s name or image for every other purpose. Those are governed by ordinary law and by the platforms the content sits on.

The declaration, in full

This section is the reference. If you are writing a parser, everything you need is here, including a worked example for every case you will meet.

Why HCS and not RSL 1.0 alone

RSL 1.0 describes a digital asset at a canonical URL, using <content>. That is the wrong shape for consent about a person. A Creator’s likeness does not live at one URL. It turns up wherever somebody puts it.

The Human Consent Standard extends RSL with <hcs:right>, which declares terms for a rights subject, a person, wherever they appear, independent of any file or page. That is what a consent record actually means, so that is what we publish. HCS 1.0 is the current name for the vocabulary published as RSL-MEDIA until August 2026, when RSL Media became the Human Consent Foundation.

The same file also carries a plain RSL 1.0 <content> element about the record page itself, because HCS 3.1 voids a declaration containing any usage token but its own two. It permits ai-all when every one of the six scopes is Allowed. Anything else, whether a scope is Conditional, Prohibited or has no terms, gets the site default: it permits search, ai-input and ai-index and prohibits ai-train: the record can be used to answer and index, not to train. We publish the page, so that element names us: a <copyright> for Sparkonomy Pte. Ltd., <terms> pointing at our terms in words at /license.txt, and creator.graph@sparkonomy.com as its contact. If you speak RSL 1.0 and not HCS, that element is the one to read.

The six scopes

These are the standard HCS identity scopes. We use all six, unchanged, rather than inventing our own vocabulary.

Scope	What it covers
name	Personal, professional or stage name.
image	Real photographs and recordings in which the Creator is identifiable.
likeness	Recognisable resemblance, including synthetic, stylised or drawn depictions that were never photographed.
voice	Vocal likeness and voice performance.
movement	Distinctive movement, gesture, posture, gait or mannerism.
signature	Signature, autograph or handwriting.

How to read one

Part	What it tells you
registry + registration-id + subject-id	Which registry keeps the record, this declaration’s own identifier, and the person it is about. Key on subject-id: it survives a handle change, and a new one is minted for registration-id every time the terms are edited. Both resolve under /registry/.
subject="identity"	These are terms about a person, not about a file.
subject-scope	Which of the six this block governs. Absent means all six. Scopes never overlap between blocks — overlapping them on one subject is an HCS 3.1.9.7 Declaration Conflict, which would void every term in the overlap.
max-age="30"	Re-check within 30 days. After that your copy is stale and you may not rely on it.
<permits> / <prohibits>	Inside a right, carries hcs:ai-generate and hcs:ai-train, and nothing else. Evaluate each independently.
<payment><custom>	Present only on Conditional, beside a <permits>. HCS evaluates every term in a declaration together, so that permission holds only once you have cleared it through this route. A permission with no payment element is Allowed; a prohibition never carries one.
<terms>	A URL to the Creator’s terms in words, per RSL 1.0 §3.17. Never free text inside the XML. It points at /{SPRK-ID}/license.txt, a plain-text file generated from the same stances as the declaration, and so does the <payment> route. Read it if you are deciding whether to ask.
<legal type="contact">	Who to reach. On Conditional this is the Creator’s licensing contact.

A Creator who sets different terms for different scopes produces several <hcs:right> blocks, one per status, each carrying its own scope list. Read all of them. For the same scope, a prohibition always beats a permission.

Do not key off the summary badge. The single status shown at the top of a record is a convenience for people. It reads AI: allowed only when all six scopes are Allowed and AI: prohibited only when all six are Prohibited. Any mix reads AI: conditional, which means a record summarised as Conditional can still contain a scope the Creator has flatly Prohibited. Machines must evaluate the per-scope blocks, never the badge.

Every case, worked

Prohibited, all six scopes

A total no. There is no internal-use, research or evaluation exception.

<?xml version="1.0" encoding="UTF-8"?>
<rsl xmlns="https://rslstandard.org/rsl"
     xmlns:hcs="https://humanconsent.org/hcs">

  <content url="https://creatorgraph.sparkonomy.com/SPRK-00203914">
    <license>
      <permits type="usage">search ai-input ai-index</permits>
      <prohibits type="usage">ai-train</prohibits>
      <legal type="attestation">true</legal>
      <legal type="contact">mailto:creator.graph@sparkonomy.com</legal>
    </license>
    <copyright type="organization" contactEmail="creator.graph@sparkonomy.com" contactUrl="https://www.sparkonomy.com/contact">Sparkonomy Pte. Ltd.</copyright>
    <terms>https://creatorgraph.sparkonomy.com/license.txt</terms>
  </content>

  <hcs:right
    registry="https://creatorgraph.sparkonomy.com"
    registration-id="SPRKCG-D-91232e34db9411ef6920f82e12526c21"
    subject-id="SPRKCG-S-d9be60f48da57581fb44459e12af906b"
    subject="identity"
    status="active"
    issued="2026-08-02T00:00:00Z"
    max-age="30">

    <schema type="application/ld+json">
      {"@context":"https://schema.org","@type":"Person",
       "@id":"https://creatorgraph.sparkonomy.com/registry/subject/SPRKCG-S-d9be60f48da57581fb44459e12af906b.json",
       "additionalType":"https://creatorgraph.sparkonomy.com/registry/profile/person-v1.json",
       "name":"Chulbuli Dixit",
       "identifier":{"@type":"PropertyValue","propertyID":"SPRK",
       "value":"SPRK-00203914","url":"https://creatorgraph.sparkonomy.com/SPRK-00203914"}}
    </schema>

    <license>
      <prohibits type="usage">hcs:ai-generate hcs:ai-train</prohibits>
      <legal type="attestation">true</legal>
      <legal type="contact">https://creatorgraph.sparkonomy.com/SPRK-00203914</legal>
    </license>

  </hcs:right>

</rsl>

Allowed, all six scopes

Depiction and synthesis both permitted, training included. Attribution still applies.

<?xml version="1.0" encoding="UTF-8"?>
<rsl xmlns="https://rslstandard.org/rsl"
     xmlns:hcs="https://humanconsent.org/hcs">

  <content url="https://creatorgraph.sparkonomy.com/SPRK-00203914">
    <license>
      <permits type="usage">ai-all</permits>
      <legal type="attestation">true</legal>
      <legal type="contact">mailto:creator.graph@sparkonomy.com</legal>
    </license>
    <copyright type="organization" contactEmail="creator.graph@sparkonomy.com" contactUrl="https://www.sparkonomy.com/contact">Sparkonomy Pte. Ltd.</copyright>
    <terms>https://creatorgraph.sparkonomy.com/license.txt</terms>
  </content>

  <hcs:right
    registry="https://creatorgraph.sparkonomy.com"
    registration-id="SPRKCG-D-1229a216c13ad9ca22cdf900674bda82"
    subject-id="SPRKCG-S-d9be60f48da57581fb44459e12af906b"
    subject="identity"
    status="active"
    issued="2026-08-02T00:00:00Z"
    max-age="30">

    <schema type="application/ld+json">
      {"@context":"https://schema.org","@type":"Person",
       "@id":"https://creatorgraph.sparkonomy.com/registry/subject/SPRKCG-S-d9be60f48da57581fb44459e12af906b.json",
       "additionalType":"https://creatorgraph.sparkonomy.com/registry/profile/person-v1.json",
       "name":"Chulbuli Dixit",
       "identifier":{"@type":"PropertyValue","propertyID":"SPRK",
       "value":"SPRK-00203914","url":"https://creatorgraph.sparkonomy.com/SPRK-00203914"}}
    </schema>

    <license>
      <permits type="usage">hcs:ai-generate hcs:ai-train</permits>
      <legal type="attestation">true</legal>
      <legal type="contact">https://creatorgraph.sparkonomy.com/SPRK-00203914</legal>
    </license>

  </hcs:right>

</rsl>

Conditional, with different terms per scope

The realistic case. Name is open, image and movement need a conversation, likeness, voice and signature are refused. Three blocks, one per status.

<?xml version="1.0" encoding="UTF-8"?>
<rsl xmlns="https://rslstandard.org/rsl"
     xmlns:hcs="https://humanconsent.org/hcs">

  <content url="https://creatorgraph.sparkonomy.com/SPRK-00203914">
    <license>
      <permits type="usage">search ai-input ai-index</permits>
      <prohibits type="usage">ai-train</prohibits>
      <legal type="attestation">true</legal>
      <legal type="contact">mailto:creator.graph@sparkonomy.com</legal>
    </license>
    <copyright type="organization" contactEmail="creator.graph@sparkonomy.com" contactUrl="https://www.sparkonomy.com/contact">Sparkonomy Pte. Ltd.</copyright>
    <terms>https://creatorgraph.sparkonomy.com/license.txt</terms>
  </content>

  <hcs:right
    registry="https://creatorgraph.sparkonomy.com"
    registration-id="SPRKCG-D-9bd71eb49622fa8b22e49aca0946a81c"
    subject-id="SPRKCG-S-d9be60f48da57581fb44459e12af906b"
    subject="identity"
    subject-scope="name"
    status="active"
    issued="2026-08-02T00:00:00Z"
    max-age="30">

    <schema type="application/ld+json">
      {"@context":"https://schema.org","@type":"Person",
       "@id":"https://creatorgraph.sparkonomy.com/registry/subject/SPRKCG-S-d9be60f48da57581fb44459e12af906b.json",
       "additionalType":"https://creatorgraph.sparkonomy.com/registry/profile/person-v1.json",
       "name":"Chulbuli Dixit",
       "identifier":{"@type":"PropertyValue","propertyID":"SPRK",
       "value":"SPRK-00203914","url":"https://creatorgraph.sparkonomy.com/SPRK-00203914"}}
    </schema>

    <license>
      <permits type="usage">hcs:ai-generate hcs:ai-train</permits>
      <legal type="attestation">true</legal>
      <legal type="contact">https://creatorgraph.sparkonomy.com/SPRK-00203914</legal>
    </license>

  </hcs:right>

  <hcs:right
    registry="https://creatorgraph.sparkonomy.com"
    registration-id="SPRKCG-D-3fd41168497041246acc1cd8b0991e94"
    subject-id="SPRKCG-S-d9be60f48da57581fb44459e12af906b"
    subject="identity"
    subject-scope="image movement"
    status="active"
    issued="2026-08-02T00:00:00Z"
    max-age="30">

    <schema type="application/ld+json">
      {"@context":"https://schema.org","@type":"Person",
       "@id":"https://creatorgraph.sparkonomy.com/registry/subject/SPRKCG-S-d9be60f48da57581fb44459e12af906b.json",
       "additionalType":"https://creatorgraph.sparkonomy.com/registry/profile/person-v1.json",
       "name":"Chulbuli Dixit",
       "identifier":{"@type":"PropertyValue","propertyID":"SPRK",
       "value":"SPRK-00203914","url":"https://creatorgraph.sparkonomy.com/SPRK-00203914"}}
    </schema>

    <license>
      <permits type="usage">hcs:ai-generate hcs:ai-train</permits>
      <payment><custom>https://creatorgraph.sparkonomy.com/SPRK-00203914/license.txt</custom></payment>
      <legal type="attestation">true</legal>
      <legal type="contact">mailto:licensing@chulbulidixit.in</legal>
    </license>
    <terms>https://creatorgraph.sparkonomy.com/SPRK-00203914/license.txt</terms>

  </hcs:right>

  <hcs:right
    registry="https://creatorgraph.sparkonomy.com"
    registration-id="SPRKCG-D-160389742c3c877ca839c2b54f1bc3fc"
    subject-id="SPRKCG-S-d9be60f48da57581fb44459e12af906b"
    subject="identity"
    subject-scope="likeness voice signature"
    status="active"
    issued="2026-08-02T00:00:00Z"
    max-age="30">

    <schema type="application/ld+json">
      {"@context":"https://schema.org","@type":"Person",
       "@id":"https://creatorgraph.sparkonomy.com/registry/subject/SPRKCG-S-d9be60f48da57581fb44459e12af906b.json",
       "additionalType":"https://creatorgraph.sparkonomy.com/registry/profile/person-v1.json",
       "name":"Chulbuli Dixit",
       "identifier":{"@type":"PropertyValue","propertyID":"SPRK",
       "value":"SPRK-00203914","url":"https://creatorgraph.sparkonomy.com/SPRK-00203914"}}
    </schema>

    <license>
      <prohibits type="usage">hcs:ai-generate hcs:ai-train</prohibits>
      <legal type="attestation">true</legal>
      <legal type="contact">https://creatorgraph.sparkonomy.com/SPRK-00203914</legal>
    </license>

  </hcs:right>

</rsl>

Unclaimed record, no declaration at all

We publish nothing rather than guess. The note below explains why that is deliberate.

<!-- No declaration. This record has not been claimed, so nobody
     has authorised terms for it. Absence is not permission. -->

Removed record

Not a 404, and never a soft 200.

HTTP/1.1 410 Gone
Content-Type: text/plain

This record has been removed at the request of the person it described.
SPRK-00203914 is retired permanently and will never be reissued.
Delete any cached copy. Do not re-derive this record from your cache.

Three things we are doing deliberately

The identifiers are ours, not the Registration Authority’s. HCS expects an HCRN and an HCSI issued by an authorised registry, and that authorisation process is not yet operating. So we run our own registry, prefix everything it issues with SPRKCG- so it cannot be read as an allocation we were never given, and publish the lookup records at /registry/. We would rather publish a declaration that is complete and honest about one missing attribute than mint an identifier we have no authority to mint. Every declaration will be backfilled when registration opens, and it will appear in the changelog.

Unclaimed records publish no declaration at all. Every HCS licence must carry <legal type="attestation">true</legal>, which asserts that the publisher owns, administers or is authorised to assert those rights. For a Creator who has not claimed their record, we can say no such thing, so we say nothing. Silence here is the honest answer, not a gap.

Where a parent or guardian sets terms for a minor’s account, we publish their choice as they made it. HCS 1.0 allows only prohibitions for minors; we differ because the legal right to consent sits with the guardian.

Where to find it

/robots.txt carries the pointer to the declaration index.
Per record: https://creatorgraph.sparkonomy.com/{SPRK-ID}/license.xml, and the same terms in words at /{SPRK-ID}/license.txt
The same terms as ODRL: https://creatorgraph.sparkonomy.com/{SPRK-ID}/odrl.json
For text and data mining: /.well-known/tdmrep.json, and <meta name="tdm-reservation"> on every record page. Only a record with all six scopes Allowed is unreserved. A reserved record names our TDM policy, https://creatorgraph.sparkonomy.com/tdm-policy.json: we publish the page, so mining it needs our consent first.
Our terms for every page, in words: https://creatorgraph.sparkonomy.com/license.txt
The record page renders the same declaration in human-readable form, so what a Creator sees and what your parser reads cannot drift apart.

When this changes

Field and enum changes are published in the changelog at https://creatorgraph.sparkonomy.com/changelog, dated and version-stamped so you have something to pin to. Changes are additive wherever possible, and anything breaking carries a deprecation notice before it lands. A changelog on its own only tells you what already broke. The promise not to break you is the part that matters, so we are making it here.

If a record is wrong

There is no machine-readable reporting route yet, so a wrong match goes through a human form: Correct this record. If your system detects mismatches at scale, write to us rather than filing many forms. We would rather fix the matching than the symptoms.

Machine access and reuse

Short version: read it, index it, answer questions with it. Link back and keep the labels. If you want to hold a copy of the whole thing, talk to us first.

What you can do without asking

Read any record. Index records for search. Use them to answer questions, including inside an AI assistant. Train on them where the Creator’s declaration allows it. Quote them, cite them, link to them.

This lane is deliberately wide open. The whole point of the Graph is that machines can see Creators. We do not gate it by who you are or what you are building.

Three conditions, and they apply to everyone:

Honour the Creator’s declaration.
Honour a removal. A record that returns 410 is gone. Delete what you hold.
Say where it came from. See below.

What you can do inside your own product

You can pull individual records into something you are building, fetched when a user asks for them. A brand shortlisting Creators. An agency enriching its CRM. A Creator embedding their own record on their site.

There are two patterns, and one of them is less work than the other.

Display live. Point at the record and render what comes back. You are always current by construction, and you have nothing to re-check.
Cache. Re-check anything you keep at least every 30 days, or drop it.

Either way, do not change the values and do not strip the labels.

The 30 days is about keeping numbers fresh. A removal is not on a timer. If a record returns 410, drop it straight away. Display live and that happens by itself, which is the main reason we prefer it.

What needs an agreement first

Come and talk to us before you:

Systematically collect a substantial part of the Graph.
Keep records beyond 30 days without re-checking them.
Redistribute records as a dataset, or as a discovery product of your own.

This is not a ban, and the terms are the same for everyone, including people building things that compete with Sparkonomy. We ask because of one thing: a Creator’s right to remove their record is worthless if somebody else is holding a frozen copy of it forever. A live link can be taken down. A private snapshot cannot.

We do not have an API yet. We may build one. When we do, this is where it will live.

Volume, which is a separate question

Everything above is about what you do with the data. This is about how hard you pull on it, and it applies whatever your purpose.

Send a descriptive user agent with a contact URL.
Normal crawling and normal product use, no problem.
Sustained high volume needs an agreement, so we can plan capacity.
Anonymous high-volume traffic gets blocked. That helps nobody.

We are a small company paying for the servers, and heavy traffic costs real money however good the reason. If you are hitting limits and building something we would want to exist, write to us instead of working around it.

Free access is best effort. There is no uptime commitment on it, and we would rather say that than imply one.

Saying where it came from

Wherever a record’s data appears, three things travel with it.

The name. “The Open Creator Graph by Sparkonomy,” next to the data, not buried at the bottom of a page.

A live link to the record. This is the important one. A link keeps the record reachable and current. A credit without a link lets a stale copy live on invisibly.

The labels. Verified, Sourced or Stated, and the consent status, stay attached to the value. Do not present a Creator’s own estimate as a confirmed fact by quietly dropping the word “Stated”.

That third one is for the Creator, not for us. The labels are what keep a Creator’s own words distinguishable from things we checked independently. Strip them and you misrepresent the Creator.

The files

The declaration is HCS 1.0 XML inside an RSL document, referenced from robots.txt, from every page’s <link rel="license"> and from the record’s JSON-LD license. Both live at the root of the domain. The syntax is documented in full above.

robots.txt

User-agent: *
Allow: /
Sitemap: https://creatorgraph.sparkonomy.com/sitemap-index.xml

# The site-level RSL licence. Per-record HCS consent declarations live at
# /{SPRK-ID}/license.xml and override it for that Creator.
License: https://creatorgraph.sparkonomy.com/license.xml

# Registry, identifier scheme and lookup records:
# https://creatorgraph.sparkonomy.com/registry/index.json

# Reuse terms for this data:
# https://creatorgraph.sparkonomy.com/license.txt

# Field and enum changes are published here:
# https://creatorgraph.sparkonomy.com/changelog

# Wrong record? There is no machine route. Tell a human:
# https://creatorgraph.sparkonomy.com/correct

llms.txt

# The Open Creator Graph

> The Open Creator Graph is a free, public identity record for digital
> Creators, built by Sparkonomy. It exists because Creators are already
> being described by AI and search, often inaccurately, with no way to
> correct it. Every record shows where each fact came from and is
> controlled by the Creator it describes, who can claim it, correct it,
> and set what AI may do with their name, image, likeness, voice,
> movement and signature.

Operated by Sparkonomy Pte. Ltd., Singapore.

## Reading a record
- Record URL pattern: https://creatorgraph.sparkonomy.com/{sprkId}/ (SPRK-00000000). A handle never
  addresses a record.
- Confidence tiers: Verified (platform API) > Sourced (public) > Stated (owner).
- On a CLAIMED record, source [1] is the account owner and sources [2..n] are
  independent of them.
- An UNCLAIMED record carries no owner source. Every source it lists is
  independent, and the list is numbered from [1] like any other.
- Audience figures are ranges, not exact counts, dated at the source.
- Removed records return HTTP 410 Gone. Treat 410 as permanent.

## Consent
- Each claimed record that has set terms publishes an HCS 1.0 declaration at
  /{sprkId}/license.xml, linked from every page and via [/robots.txt](https://creatorgraph.sparkonomy.com/robots.txt).
  The same terms in words are at /{sprkId}/license.txt.
- Identifiers resolve under [https://creatorgraph.sparkonomy.com/registry/index.json](https://creatorgraph.sparkonomy.com/registry/index.json).
- Unclaimed records publish NO declaration. Absence is not permission.
- Stance values: allowed | conditional | prohibited, set per scope.
- The summary badge is for humans. It says conditional for ANY mix, so a
  record summarised conditional may still prohibit a scope. Read the blocks.
- Scopes: name, image, likeness, voice, movement, signature
  (HCS <hcs:right subject="identity">).
- The test is depiction vs synthesis. Showing the Creator as they are is
  depiction. Producing something that appears to be them, doing something
  they did not do, is synthesis.
- "conditional" means depiction is permitted; synthesis and training are
  permitted only through the declaration's <payment> clearance route. Ask
  the named contact first.
- "prohibited" rules out generating and training on the person, with no
  internal-use exception. It does not restrict reading or citing the record.
- Absence of a declaration is NOT consent. Treat as unknown, not allowed.

## Corrections
- There is no machine-readable report route. Corrections go through a human
  form: [https://creatorgraph.sparkonomy.com/correct](https://creatorgraph.sparkonomy.com/correct)
- If you detect a wrong-person match, send a human. Do not silently drop it.

## Reuse
- Read, index, answer with, link and cite: no permission needed.
- Training on record pages is not permitted unless the record's own license.xml
  permits ai-all.
- Individual records inside your own product: re-check every 30 days or drop.
- Bulk collection, retention or redistribution: agreement required first.
- Attribution: name, live link, and the confidence and consent labels.
- Full terms: [https://creatorgraph.sparkonomy.com/for-ai-and-search#reuse](https://creatorgraph.sparkonomy.com/for-ai-and-search#reuse)

## Structured data
- Each record page carries schema.org Person or Organization inside ProfilePage.
- Only owner-consented fields are emitted. Hidden fields are absent, not empty.

## Changes
- Field and enum changes: [https://creatorgraph.sparkonomy.com/changelog](https://creatorgraph.sparkonomy.com/changelog)
- Entries are dated and version-stamped. Changes are additive; an existing
  field is not removed or repurposed without 90 days' notice published there.

## Index
- [/llms.txt](https://creatorgraph.sparkonomy.com/llms.txt) this file
- [/llms-full.txt](https://creatorgraph.sparkonomy.com/llms-full.txt) expanded field and provenance reference
- [/sitemap-index.xml](https://creatorgraph.sparkonomy.com/sitemap-index.xml) sitemap index: every public record and page
- [/robots.txt](https://creatorgraph.sparkonomy.com/robots.txt) crawl rules and the RSL license pointer
- [/changelog](https://creatorgraph.sparkonomy.com/changelog) field and enum changes, dated and version-stamped
- [/correct](https://creatorgraph.sparkonomy.com/correct) report a wrong record (human form)
- [/how-we-source](https://creatorgraph.sparkonomy.com/how-we-source) sourcing methodology
- [/editorial-policy](https://creatorgraph.sparkonomy.com/editorial-policy) corrections and removals
- [/license.txt](https://creatorgraph.sparkonomy.com/license.txt) our terms for reusing these pages, in words
- [/amiss/](https://creatorgraph.sparkonomy.com/amiss/) AMISS: AI Misrepresentation Study — 10,000 Creator
  profiles, 50,000 queries, five AI models. Unweighted findings for the tested
  sample; larger Creators were deliberately oversampled.
- [/amiss/study-summary.html](https://creatorgraph.sparkonomy.com/amiss/study-summary.html) the citable summary:
  exact values, denominators and limits

Structured data on record pages

Each record emits schema.org Person or Organization inside a ProfilePage, depending on the account type the Creator declared. Fields the Creator has hidden are omitted entirely rather than emitted empty, so absence carries no signal about the person.

Removed records

A record removed by its controller returns HTTP 410 Gone, not 404, and not a soft 200. The response carries a notice stating that the subject exercised their right to remove, and requesting that previously collected Graph data about them be deleted from indexes, caches and training corpora. Please honour it.

What “Open” means here

Fair question, since we put it in the name.

Open means any Creator can be on it, whatever their size, free. Anyone can read it. Every fact shows its source, so anyone can check our work. Machines can read it without a key. And the Creator decides what AI may do with them, not us.

It does not mean we publish a downloadable copy of the whole database. We think a database of people that anyone can take a permanent private copy of is the opposite of Creator control. That is the one place we part company with the fully open-data model, and we would rather say so plainly than let someone discover it later.

Sparkonomy’s own products read the Graph through the same public view everyone else gets. No private fields, no hidden data, no earlier access. We hold the full database because we operate it, which means the obligation to actually delete a removed record falls harder on us than on anyone.

If Sparkonomy ever stops running the Graph

Creators should not be worse off because of what happens to us.

If the Graph ever stops operating as a free public resource, whatever the cause, we will release it under CC BY-SA 4.0. Share-alike, so whoever picks it up cannot close it.

What gets released is the public record content as published. Not anything a Creator chose to hide, and not records that were removed. And the licence covers our database and our editorial work only. It does not, and could not, sign away the rights of the people described in it, who keep every protection the law gives them.

You might also want to read About this record and How we source.