Existing-Feature and Prior-Proposal Review
2 November 2026
1. Review Findings
A review of publicly accessible Perplexity API Forum discussions did not identify an existing proposal encompassing the complete architecture outlined here.
2. Distinct Contribution
This proposal introduces an interoperable framework combining institution-published verification records, permission-aware evidence registries, cross-platform retrieval, structured provenance and independently maintained correction histories.
3. Potentially Related Discussions
- Hierarchical Multi-Agent Reasoning System with Integrated Fact-Verification and Self-Correction Loops (December 2025).
- Truthfulness Color Indicator Toggle for Comet Add-In/Perplexity Sources (March 2026).
- Able to Specify Sources Similar to the Perplexity UI? (July 2025).
4. Possible Overlap
These discussions address potentially adjacent topics. The extent of any substantive technical or conceptual overlap has not been established and should not be presumed.
5. Review Limitations
Findings are limited to publicly accessible materials examined and do not establish the absence of unpublished work or prove exclusive technical novelty.
6. Authorship
Grace R. Nettey — Manifold Grace. Human-originated conceptual research and architectural proposal, with AI-assisted drafting and presentation.
Here is the complete Perplexity API feature request, organised in exactly the order of the forum fields. You can copy the content directly into the submission form.
I’ve preserved your central proposal: independently published institutional evidence, interoperable verification records, provenance, controlled retrieval, and consistent verification across different AI platforms.
1. Pre-flight checklist
This is a Perplexity API feature request, not an app, Comet or web UI request.
A developer workflow and API use case are described below.
Existing documentation and forum topics should be checked for overlapping proposals before ticking the second box.
2. Feature type
New endpoint or API capability
New model, preset, or model behavior
New parameter or request option
New response field or metadata
SDK improvement
Dashboard, API key, or auth improvement
Billing, credits, rate limit, or usage reporting improvement
Docs or examples improvement
Other API improvement
Recommended selections are preselected. If the forum only permits one, choose New endpoint or API capability.
3. Affected API area
Agent API
Search API
Sonar API
Embeddings API
SDKs
Dashboard, API keys, or auth
Billing or credits
Not sure
Agent API and Search API are the strongest proposed integration areas. Sonar API may also benefit, but it is not essential to the initial request.
4. Summary
Introduce interoperable, institution-published evidence-verification records accessible through Perplexity’s APIs, allowing developers to retrieve claim-specific evidence, provenance, corroboration status, limitations, contradictions and correction histories consistently across AI platforms.
5. Problem
AI retrieval systems can discover and cite webpages, but the existence of a retrieved source does not necessarily establish that the specific claim has been verified.
Credible evidence may be held by research institutions, regulators, public authorities, accredited professionals, archives and other accountable organisations. These organisations may be able to assess underlying evidence and publish appropriately approved verification statements, including records that cannot be fully disclosed publicly.
At present, developers may need to construct separate retrieval pipelines, verification rules, provenance structures and correction mechanisms for different AI providers.
This creates risks of inconsistent interpretation, duplicated verification effort, incomplete attribution, stale evidence and uncertainty about whether apparently independent AI responses rely on the same underlying sources.
An interoperable evidence-verification capability would help separate the act of retrieving information from the determination of what credible published evidence actually supports.
6. Proposed API behavior
Introduce an optional evidence-verification integration layer accessible through the Agent API and Search API, using a common, versioned evidence-record schema.
The capability would allow Perplexity-powered applications to query approved verification records published by participating institutions or independent evidence-review organisations.
Proposed workflow
- Institutional publication: An authorised organisation evaluates selected evidence and publishes an approved verification record containing a defined claim, supporting evidence, reviewer identity or role, methodology, publication date, limitations and verification status.
- Record registration: The published record receives a persistent identifier, version number, source attribution and correction history. Private supporting documents remain protected unless specifically authorised for disclosure.
- API retrieval: Developers can request a verification record using a claim identifier, public URL, entity reference or structured query.
- Provenance and status: The API returns relevant evidence metadata, including the publishing organisation, assessment method, corroborating sources, unresolved contradictions and current verification status.
- Interoperability: The same approved record can be retrieved by other participating AI platforms through documented read-only interfaces, without exchanging private chat histories.
- Grounded generation: Applications can incorporate these records into answers while distinguishing verified evidence from unsupported claims, competing interpretations and unresolved uncertainty.
- Corrections: Updated, challenged, superseded or revoked records are reflected through versioned status responses and appropriate cache invalidation.
Illustrative endpoint design
GET /v1/evidence/claims/{claim_id}— retrieve an approved evidence record.POST /v1/evidence/search— search available verification records.GET /v1/evidence/claims/{claim_id}/status— check current status and version.
These are proposed endpoint names, not claims about existing Perplexity functionality.
Suggested response fields
claim_id, version, supported_statement, publisher, reviewer, evidence_basis, corroboration, verification_status, conflicts, limitations, citations, reviewed_at, expires_at, permissions and correction_history.
Suggested verification statuses
Supported, independently corroborated, disputed, contradicted, insufficient evidence, superseded and withdrawn.
Verification must remain claim-specific. Institutional credibility alone should not automatically establish factual truth.
Implementation approach
Begin with one bounded corpus and an institution-operated registry. Use authenticated HTTPS/JSON endpoints, scoped access permissions and read-only tool adapters.
Existing tool-calling or MCP-compatible infrastructure could be used where appropriate. This should not require model retraining or the creation of a centralised global repository of private records.
7. Example usage
Example developer workflow
A developer builds a research assistant using Perplexity’s Agent API.
A user asks whether a particular public claim has been independently substantiated.
The application queries a participating evidence-verification registry, retrieves the approved claim record and provides the result to the model alongside other relevant search evidence.
The final answer identifies the supporting institution, its published assessment, any independent corroboration, important limitations and unresolved disagreements.
Illustrative API request
POST /v1/evidence/search
Content-Type: application/json
Authorization: Bearer <API_KEY>
{
"query": "Has the published claim been independently verified?",
"locale": "en-GB",
"requested_scope": "public",
"include_provenance": true,
"include_conflicts": true
}
Illustrative response
{
"records": [
{
"claim_id": "example-001",
"version": 2,
"supported_statement": "Example claim",
"publisher": "Participating verification organisation",
"verification_status": "independently_corroborated",
"evidence_basis": [
"Published institutional assessment",
"Independent supporting publication"
],
"reviewed_at": "2026-11-01",
"limitations": [
"Assessment applies only to the stated claim"
],
"conflicts": [],
"citations": [
"https://example.org/evidence/example-001"
]
}
]
}
The request and response demonstrate proposed behaviour only; they are not existing Perplexity endpoints or verified production records.
Cross-platform test
Two independently configured AI applications, potentially operating in different languages and countries, query the same approved claim.
Both should obtain the same underlying record identifier, published version, verification status and source provenance, even if their natural-language explanations differ.
This would test evidence consistency without requiring identical model-generated responses.
8. Workarounds tried
Existing approaches considered
- Conventional web search and citation retrieval.
- Retrieval-augmented generation using indexed documents.
- Custom tool/function calling to external evidence services.
- MCP-compatible integration where supported.
- Application-managed source metadata and verification rules.
- Manual comparison of institutional publications and independent corroborating sources.
These mechanisms provide useful building blocks, but they do not automatically deliver a common, independently maintained, interoperable verification-record standard.
A developer can potentially build a custom registry and connect it through available retrieval or tool interfaces. However, doing so requires separate implementation of record schemas, permission enforcement, source authentication, verification statuses, version tracking, correction handling and cross-platform consistency.
Research undertaken
I reviewed publicly documented integration approaches and circulated the proposed architecture to multiple AI providers for technical consideration.
The original proposal was submitted to Perplexity on 1 November 2026, with subsequent correspondence concerning API-team referral.
Testing status
This is a research-led architecture proposal. I have not yet built or benchmarked an operational Perplexity API integration or completed a production proof of concept.
The proposal therefore distinguishes documented existing capabilities from the additional functionality being requested.
9. Impact
Beneficiaries
- AI developers: Reusable evidence-verification interfaces rather than independently rebuilding the same governance mechanisms.
- Perplexity: Opportunities for differentiated evidence-grounding capabilities, richer provenance metadata and enterprise-oriented integrations.
- Researchers and educators: More transparent identification of published evidence, corroboration, contradictions and uncertainty.
- Regulators and compliance teams: Greater traceability of source records, review methods and correction histories.
- Institutions: Controlled publication of reviewed findings without exposing all underlying private materials.
- End users: Better understanding of which claims have been substantiated and by whom.
Potential platform benefits
The proposed capability could improve answer traceability, source attribution, auditability, reproducibility and consistency across AI applications.
It could also reduce duplicated verification work where multiple developers rely on the same appropriately reviewed public evidence records.
Developer experience
Standardised read-only endpoints and versioned response metadata would simplify integration with external verification services.
Reliability
Explicit verification statuses, timestamps, limitations and correction records would help developers distinguish current evidence from outdated or unsupported material.
Cost and latency
Reusing approved evidence records could reduce some repeated verification work. However, additional registry queries introduce network and operational overhead. Any net cost or latency improvement would need to be measured rather than assumed.
Suggested evaluation metrics
Citation accuracy, retrieval coverage, unsupported-claim frequency, record-version consistency, correction propagation, permission enforcement, multilingual accuracy, latency and cost per verified query.
10. Additional context
Proposal provenance
Conceptual contributor: Grace R. Nettey
Research: Manifold Grace — AI Governance & Research
Original cross-platform proposal circulated: 1 November 2026
Subject: Interoperable evidence verification and global retrieval — data-flow design
This feature request arises from independent, human-led research concerning AI epistemic reliability, institutional evidence, information provenance, interoperability and accountable retrieval.
AI-assisted drafting and technical articulation were used to organise and communicate the research. These tools do not negate the underlying human-originated conceptual contributions, although individual technical elements may overlap with existing technologies or prior research.
The proposal does not claim that all underlying mechanisms are novel. Its principal contribution is the proposed combination of institutional evidence publication, structured verification records, permission-aware retrieval, cross-provider interoperability, correction mechanisms and provenance governance.
Authorship and attribution
I request that the originating contributor’s identity, original submission date and research context remain associated with this proposal.
If identifiable elements subsequently inform product requirements, architecture or implementation, appropriate documentation of the contribution would be appreciated.
This request concerns accurate intellectual provenance and does not imply that Perplexity is obliged to implement the proposal.
Scope and safeguards
The architecture is intended to support evidence-based claims, not establish infallible truth or allow institutional authority to substitute automatically for verification.
It should accommodate disputed records, independently corroborated findings, credible non-institutional evidence, appeal mechanisms and changing evidence.
Private documents, personal correspondence and third-party information should never become publicly retrievable merely because they were submitted for review.
Suggested initial proof of concept
A narrowly scoped pilot involving one participating institution, approved evidence records, read-only verification endpoints and two independently configured AI clients.
The pilot could establish whether the proposed schema and governance model offer measurable advantages over existing retrieval methods.
Requested consideration
I welcome technical feedback concerning:
- Existing Perplexity capabilities that substantially overlap with the proposal.
- Feasibility of integrating externally published verification records with the Agent API or Search API.
- Suitable approaches to interoperability, provenance and correction handling.
- Whether the proposal could be considered by an appropriate API, research or engineering representative.
This is a voluntary proposal for consideration and technical discussion, not a request for an implementation commitment.
Contact: Grace R. Nettey — Manifold Grace
Final checks before posting
Confirm no substantially identical existing feature request
Review the example JSON and replace placeholder evidence only if needed
Confirm the public authorship details you wish to disclose
Save a dated copy of the final submitted post
0 of 4 complete
Open Perplexity forum
I checked the forum’s current feature-request listing, including its API-focused submissions. I did not establish that no similar proposal exists anywhere in the forum, so that checklist item remains for confirmation.
Perplexity API Platform Forum
Suggested post title: Interoperable Evidence Verification API — Institutional Provenance, Corroboration and Cross-Platform Retrieval
This keeps the feature request centred on your proposed technical capability while preserving a clear record of your conceptual contribution.