Glossary · area 8 of 19
Showing a record existed at a date without asking anyone to trust us.
24 terms. Each opens its own page with what it means, what it does not show, how CrawlCheck measures it, and where it comes up in the guides.
An append-only, publicly readable log in which every signed answer is recorded as a leaf in a Merkle tree, so the issuer cannot quietly change or delete what it said. Anyone can check that an answer is in the log and that the log only ever grew. Certificate Transparency (RFC 9162) is the best-known example; CrawlCheck logs its Resolve answers and business records the same way. A log proves what was published and when, not that it was true.
The issuer's signature over the root hash and size of a transparency log at a point in time. Inclusion proofs are checked against it, and two signed tree heads can be compared to prove the log only grew. A signed tree head is a commitment; the protection comes from outside parties keeping and comparing them.
The short list of hashes that lets anyone recompute a Merkle tree's root from one leaf, proving that the leaf is in the log without downloading the whole log. If the recomputed root matches the signed tree head, the record was logged. It proves presence, not correctness of the record's content.
An independent party that checks a transparency log's new tree heads and co-signs the ones that are consistent with what it saw before. A witness makes it much harder for a log operator to show different histories to different readers. CrawlCheck's log is co-signed by a workflow that runs outside its own infrastructure; a witness checks consistency, not the meaning of the entries.
A log operator showing one history to some readers and a different history to others, so each group sees a consistent but different record. It is defeated by readers exchanging the tree heads they were shown (gossip) and by outside witnesses; two validly signed heads that disagree are themselves the proof.
An HTTP Link header with rel="cite-as" (RFC 8574) that gives the permanent URL a reader should use when citing a resource, instead of the address it happened to fetch. Every CrawlCheck Resolve answer carries one pointing at the as-of URL of that exact answer. The link names the right address to cite; it does not make the content any more correct.
The answer that was in force about a domain at a given past moment, returned exactly as it was signed then, with the window it was in force and its place in the transparency log. It lets anyone audit a decision an agent made last week against the answer the agent actually had. History only exists from the moment answers began to be archived.
Measured: ANSWER_ENGINE_REFUSED 5.8%
A signed record that a specific passage appeared on a specific page at a specific time: the URL, a link that opens the page at the passage, hashes of the passage and of the page text, and the time. It can be rechecked later and reports present, no longer present or unreachable. An anchor proves the text was there, not that the text is true.
The first place a piece of text was published, as opposed to the copies that syndicate or quote it. It is established from evidence in order of trust: pages naming another as rel=canonical, archive captures, declared publish dates and first sightings. The answer states its basis and how strongly the evidence agrees; when the evidence is thin it says so rather than guessing.
A signed statement that something was looked for and not found, naming exactly what was requested, the answer received, from where and when, and what the absence does and does not mean. A timeout, an error or a challenge page is never recorded as an absence. It turns 'there is no llms.txt' from an assumption into a checkable claim with a scope.
A statement that observers on at least k different networks independently signed the same reading of a page, so no single vantage point or operator can produce it alone. Each signer's own signed reading is included, so the count can be recomputed offline. It proves agreement across networks; if one organisation runs all the observers, it does not prove independence of operators.
A scheme that anchors a hash into the Bitcoin blockchain so a document can be proved to have existed at a time. It proves existence and integrity; it proves nothing about whether the content was correct.
Committing a hash to a public ledger so it cannot be backdated. Useful for a report whose value depends on the date it was taken.
A deterministic digest of a payload. Two parties can confirm they read identical bytes by comparing one short string instead of the whole document.
Editing a change log or metadata record after the fact. Hash-chained records make it detectable: the edit no longer matches the digest that was sealed at the time.
The Coalition for Content Provenance and Authenticity specification: a signed manifest bound to an image, video or audio file recording how it was created and what was done to it, often surfaced as Content Credentials. It establishes provenance, the chain of edits, and nothing about rights: a manifest does not prove copyright ownership, and an asset with no manifest is not thereby unowned or synthetic. Most assets on the web carry none, so absence is the ordinary case, not a finding.
An assertion from the Creator Assertions Working Group that travels inside a C2PA manifest and states the creator's preference about AI training and data mining for that specific asset. Because it is signed into the file, it survives copying in a way a robots.txt rule cannot. Attaching it requires signing the asset, which almost no publishing pipeline does, so adoption rather than validity is the limit; it also binds only readers who look inside manifests, which crawlers largely do not.
Storage where an object's identifier is derived from a hash of its bytes, as in an IPFS CID, so the address itself proves what it delivers: change one byte and the identifier changes. It is a stronger immutability claim than a change receipt, which proves a difference between two dated observations. It says nothing about authorship or time: anyone can store any bytes, and the identifier does not say when. Paired with a timestamp proof it becomes a dated, verifiable copy; alone it is only a verifiable copy.
A structure in which many hashes are paired and hashed upward until one root remains, so that a single root commits to every leaf and any one leaf can be proven to belong with a short path of sibling hashes. It is how one Bitcoin transaction can timestamp thousands of daily records at once.
The hash function that turns any input into a fixed 256-bit digest, such that changing one byte of the input changes the digest unpredictably and no two different inputs are known to share one. A published digest commits to a record's exact bytes without publishing the record.
The once-a-day act of hashing every report produced that day into one root and anchoring that root in the Bitcoin blockchain through OpenTimestamps. It proves a record existed by that date in that exact form; it does not prove the record was correct, or that anyone read it.
The named author on a page, and the machine-readable author node that should match it. Engines and readers use it as one proxy for accountability; a byline that names a person no other page corroborates carries little, and a page with none has told a reader nobody stands behind it.
The published and modified dates on a page, visible and in structured data, and whether they agree. A modified date that changes on every build is discarded as noise; a page with no date at all cannot be judged for freshness, which for a fast-moving subject is a reason to be passed over.
A page stating what a service collects, how long it keeps it, which headers and email protections it sends, which outside services it depends on, and how to reach it. Every line should be a checkable fact with the path to check it; a trust center of adjectives is marketing under another name.