Skip to content
FonteumThe Graph
Fonteum · Compare

Stop shipping answers you can't stand behind

Answer-first capsule: AI and RAG builders do not need another plausible paragraph. They need a dated public-records answer that names the government source and carries the available limitation and observation metadata. A separate snapshot attestation does not sign individual facts. Fonteum turns the same question an LLM might answer from memory into a source-backed answer across

the documented federal-source catalogSource: Fonteum methodology · As of 2026-07-12
.

Published June 24, 2026 · Last reviewed June 2026 · Capability comparison — public facts only

9-capability comparison

From plausible text to a source-cited, re-checkable public-records answer

From plausible text to a source-cited, re-checkable public-records answer
CapabilityUngrounded LLM answerFonteum
Same questionWhich public records show whether a provider or vendor appears on an exclusion list?Which public records show whether a provider or vendor appears on an exclusion list?
Example answerA plausible paragraph says the organization is probably clear, cites no source, gives no checked date, and may be stale by the time it is read.The OIG LEIE snapshot contains aggregate exclusion records. Fonteum answers expose source, snapshot-date, limitation, and evidence fields where the response supplies them; no individual record is named in this aggregate example.
SourceMay blend training data, web memory, or retrieval snippets without a field-level citation.Government record first — OIG, SAM.gov, CMS, HRSA, or another registered source family named in the answer.
DateOften no as-of date, or a generic phrase like recently.Responses expose snapshot and last-checked fields where the route and source supply them; missing dates remain explicit.
SignatureNo signed evidence package travels with the text.A separate snapshot attestation can expose a digest and signature for its identified object; it does not sign the answer's individual facts.
Time-machine edgeAnswers the present-tense question only, and may not know what changed since training or retrieval.As-of reads only for retained rows from named sources. Production history is concentrated in UK sanctions, with small procurement, OFAC, UN, and provenance sets; it is not universal replay.
Re-check lineA user can rerun the prompt, but cannot re-create the exact upstream record behind the sentence.Re-fetch the government file, compare the digest, and re-check the evidence at /verify.
RAG handoffThe answer is a text blob the application has to defend after the fact.The answer ships with source, date, limitation, and evidence pointers that a RAG app can show to the user.
Failure modeConfident, unsourced, and possibly stale output that sounds useful until someone asks for the record.A dated public-record answer that can be quoted, audited, and re-checked without naming nonessential individuals.

This comparison is aggregate and FCRA-safe. The worked example cites an OIG LEIE aggregate count and names no individuals; it compares answer mechanics, not any provider, contractor, or person.

Why grounded answers matter

The evidence path is the product

Answer-first capsule

Stop shipping answers you can't stand behind. If an AI product answers a compliance, credentialing, diligence, or sourcing question, the answer needs a source, an as-of date, a limitation, and an evidence trail at the moment it leaves the model.

The time-machine edge

Most LLM answers collapse time into now. Fonteum can answer an as-of question only when the named source has a retained row version. Production history is source-specific and sparse outside UK sanctions; current records cannot be treated as universal replay.

The re-check line

A Fonteum answer can expose source and observation metadata. When a concrete snapshot attestation is supplied, its digest can be inspected at /verify; that object-level check does not sign each fact in the answer.

FAQ

Common questions

Why is an ungrounded LLM answer a problem for public-records data?
The problem is not fluency; it is accountability. A public-records answer can affect compliance, credentialing, diligence, or user safety. If the answer has no source, checked date, limitation, or evidence trail, the builder cannot show what record supports it when a user, auditor, or customer asks.
What does Fonteum add to a RAG or agent stack?
Fonteum adds the evidence layer: with source, date, limitation, digest, or attestation fields where the response supplies them. A RAG or agent application can cite the government record instead of asking the model to defend an unsupported answer.
What is the worked example on this page?
The example is aggregate and FCRA-safe: the OIG LEIE snapshot carries exclusion records. The point is not to name an individual; it is to show the difference between an unsourced answer and a dated answer that points back to OIG as the government record.
How does the time-machine edge help AI builders?
Only retained row versions can support an as-of answer. Production history is concentrated in UK sanctions, with small retained sets for procurement, OFAC, UN, and provenance; other sources may have no past row to replay.
Can I use this through an API?
Yes. Fonteum exposes public-records data through API surfaces for builders, plus research downloads for aggregate work. Source and evidence metadata varies by route and can be null, so applications must preserve only the fields actually returned.
Request access

Ground your next answer in the public record.

Start with /for/rag, connect agents at /for/ai-agents, inspect the API, re-check evidence at /verify, or browse aggregate work at /research.

See also
  • /for/rag How RAG builders ground provider-data answers in source-cited records.
  • /for/ai-agents How agents call Fonteum with evidence metadata intact.
  • /api API entry point for public-records data with source and date metadata.
  • /verify Inspect a concrete snapshot artifact and digest where supplied; individual facts are not signed.
  • /research Aggregate public-records research with downloads and methodology.

What’s on file, by the numbers

Platform snapshot · 2026-07-22

13.4Mproviders & companiesProviders, organizations, owners, and facilities on file
26.2Msource-linked factsSource-linked field facts in the dated platform snapshot
90sources with dataDistinct snapshot source IDs with at least one positive record count
79fresh sourcesDistinct source IDs whose latest positive-data snapshot falls within the preceding 45 days
111sources integratedActive registry rows; integration does not establish a load
13state Medicaid jurisdictionsDistinct states represented in the state-exclusions serving table

Integrated, with-data, and fresh-observation counts are separate. No platform-wide source-completeness count is published. Completeness is source-specific and must be evaluated against the named source's expected scope. State coverage is a separate jurisdiction measure.

Source authority is record-specific

Use the issuer named on the record.

Fonteum spans federal, state, and global public publishers. A source page or returned record identifies its issuer and dataset where that metadata is available. A platform registry count does not assign every page to one authority or establish loaded, fresh, or complete coverage.

Browse source records and their stated limitations →

Reproducible by design

Inspect the evidence each published figure actually supplies.

Source and date

Research pages expose the named public file and observation date where those fields are available. Source-file SHA-256 coverage is separate; facts do not currently link deterministically to signatures.

Available derivation

Studies with a retained release and committed derivation link the SQL or method used. Other studies state the evidence and reproduction limits they actually have.

Daily observations

Dated table row-count observations can detect local drift. They do not imply that an upstream publisher released or Fonteum ingested new data that day.

Named medical review

Reviewed by Jennifer Montecillo, MD, medical reviewer. Non-practicing medical reviewer.

Read the full provenance and attestation methodology →

Request access