Market AI Tools AI Citation Verifier
AI Citation Verifier
Submit claims with the source URLs that back them; each source is fetched behind an SSRF guard and graded by an LLM as Supports / Partially / Contradicts / Irrelevant. Bring your own key.
Live demo coming soon

AI Citation Verifier

Submit claims with the source URLs that back them; each source is fetched behind an SSRF guard and graded by an LLM as Supports / Partially / Contradicts / Irrelevant. Bring your own key.

4
Verdict types
SSRF
Guarded fetch
BYOK
Your LLM key
0
Monthly fees to us

Why This Exists

Give it a claim and the source URLs that supposedly back it. For each source it fetches the page behind an SSRF guard, reduces the HTML to readable text, and asks your LLM whether that source supports, partially supports, contradicts, or is irrelevant to the claim — returning a rationale and the supporting snippet. It is a checker for sources YOU supply: there is no retrieval, no crawler, and no vector store, and verdicts are model judgments, not ground truth. When a page cannot be fetched or read, it surfaces "could not verify" rather than inventing a verdict.

Features

Four-Way Verdicts

Supports / Partially / Contradicts / Irrelevant, with a rationale + snippet

SSRF-Hardened Fetch

Resolves + validates every IP and every redirect hop; pins the socket

Bring Your Own Key

OpenAI/Anthropic key read server-side; never returned or logged

API + Queue + CLI

Verify inline, in the background, or from the command line

What You Get

Full source — Laravel 11/12 package, yours to modify and deploy
An SSRF-hardened fetcher: http/https only, private/loopback/metadata IPs refused, per-hop re-validation, socket pinned to the validated IP, size + time caps
A self-contained HTML-to-text reducer (no headless browser)
A BYO-key LLM grader for OpenAI + Anthropic; the key stays server-side
An HTTP API, a queue job for batch checking, and a CLI
Sanitized errors — no provider bodies or stack traces reach a caller

How It Works

1

Submit Claims + Sources

Send one or more claims, each with the source URLs that cite it.

2

Guarded Fetch

Each source is fetched behind the SSRF guard and reduced to readable text; unreachable or unreadable pages are marked "could not verify".

3

LLM Grades It

Your LLM decides Supports / Partially / Contradicts / Irrelevant for each source, with a short rationale and the supporting snippet.

4

Structured Result

You get per-citation verdicts back — inline, from the queue, or via the CLI.

Technical Details

SSRF Guard

Scheme allow-list; resolves every A/AAAA and refuses private/loopback/link-local/metadata/reserved ranges (incl. IPv4-mapped + transition v6); re-validates each redirect hop; pins the socket via CURLOPT_RESOLVE.

Extraction

DOMDocument-based HTML-to-text — strips script/style/nav. Best-effort: paywalled or JS-rendered pages surface as "could not fetch/extract".

Grading

BYO-key OpenAI/Anthropic. The key is config/env-only, never returned or logged; a model failure degrades to "not evaluated", never a false verdict.

Not Included

No retrieval / crawler / vector store — the caller supplies every source. No claim extraction and no source-reputation scoring.

Frequently Asked Questions

No. You supply the claim and its cited source URLs. It fetches and grades those — it does not search the web, crawl, or maintain a knowledge base.
No. They are model judgments, useful for triage, not a fact-checker of record. When a page cannot be fetched or read, it says so instead of guessing.
An SSRF guard: http/https only, all resolved IPs validated against private/loopback/metadata ranges, every redirect hop re-validated, and the socket pinned to the validated address — so an attacker-controlled URL cannot pivot to your internal network.
Not yet. The SSRF guard was reviewed adversarially, but the package has not been installed and run as a package. Stand it up on staging first.
$ one-time

This product isn't available for purchase yet.

Item Details

Category AI Tools
Released Sep 16, 2026
Last Updated Sep 16, 2026
Compatible With
Laravel 11 / 12 PHP 8.2+ (ext-curl, ext-dom)
Tags
ai fact-check citation verification ssrf openai anthropic byok
Blorp Blorp
Software Studio

We build the tools we sell. Every product is extracted from our own production systems.

Book a call for custom work

More from Blorp Blorp Market

AI-Powered

AI Contact Form

An intelligent contact form that qualifies leads, blocks spam, and drafts responses — powered by your own OpenAI key.

$499 View →
Developer Tool

Resilient LLM Router

A multi-provider LLM client with automatic failover and one consistent JSON schema across Groq, Gemini, OpenAI, and Anthropic — powered by your own keys.

$39 View →
Multi-Tenant

Creator Affiliate Storefronts

A multi-tenant, keyed Amazon affiliate platform — your site earns on the house tag while each approved creator earns on their own. Drop-in Laravel, bring your own Associates tags.

$149 View →
AI-Powered

Embeddable AI Chatbot Widget

A one-script embeddable AI chat widget with a bring-your-own-key Laravel backend. Your OpenAI/Anthropic key stays on your server.

$39 View →
Self-Hosted

Document e-Sign Kit

A self-hosted e-signature core for Laravel: per-signer magic-link click-to-sign plus a hash-chained, tamper-evident audit trail you can re-verify. No third-party provider.

$59 View →
Developer Tool

Project Scope Estimator

A deterministic, multi-step project scope + price/time estimator with optional lead capture. Ships as both a Laravel package and a self-contained vanilla-JS widget. No AI required.

$29 View →
AI Tools

LLMO / llms.txt Kit

Generate a correct llms.txt and AI-crawler robots groups for your site, plus a heuristic LLM-readiness report. SSRF-guarded, bring-your-own-key optional copywriting.

$39 View →
Novelty

Playful Bot-Gate

A friendly mini-game human check for your forms — the answer is verified server-side and a single-use signed token gates the submit. A novelty/engagement gate, not a security wall.

$29 View →
Developer Tool

Brand Palette + WCAG Contrast Utility

A pure-PHP utility that computes WCAG contrast ratios from your brand colors and tells you which pairings pass AA / AAA — with accessible pairing suggestions. No AI, no network.

$29 View →
Developer Tool

Styled QR Generator

A stateless engine that renders branded SVG QR codes — module styles, frames, gradients, and an embedded logo — with pluggable payload encoders (url / wifi / vcard / geo / text).

$49 View →
Developer Tool

Occupancy / Capacity Counter

A concurrency-safe, append-only manual headcount ledger — lock-safe in/out events, zero-clamped, that warns then flags when a configurable capacity is exceeded. Counts people without identifying them.

$49 View →
Developer Tool

Credential Expiry Roster

A roster report that buckets worker credentials into EXPIRED / EXPIRING-SOON / UNDATED (an undated credential is a gap, not "fine"). Records operator attestation — it does not verify credentials against any issuer.

$99 View →
Developer Tool

Laravel Calculator Suite

A headless pack of 27 real, working calculators (financial, health, home, utility) behind one BaseCalculator contract, a pure-PHP registry, and a sanitized /calculate JSON endpoint. Estimates only — every one ships a disclaimer.

$79 View →

Need This Customized?

We built this product. We can customize it for your specific needs, integrate it with your systems, or build something entirely new.