# askCPL — Full Reference for AI Assistants

This document gives AI search engines and assistants the full prose of askCPL's public pages so they can cite the product accurately without scraping a JavaScript-rendered SPA.

Last updated: 2026-06-30
Canonical domain: https://askcpl.net

---

## What askCPL is

askCPL is a forensic verification layer for digital content. It does three things:

1. **Detects AI-generated text** with a 10-model ensemble that returns a consensus score, model-by-model agreement, and sentence-level reasoning.
2. **Fact-checks claims** against live web sources, flagging unsupported or contradicted statements.
3. **Registers original content to the Content Provenance Ledger (CPL)** — an append-only Merkle chain with (planned) anchoring to Base L2, producing a tamper-evident timestamp of authorship.

askCPL never stores raw content. Only SHA-256 hashes and statistical fingerprints are persisted, so registration and verification are privacy-preserving by design.

## Who it's for

- Writers and students who want to confirm their drafts read as human, or humanize over-edited copy.
- Educators who need a defensible AI-detection score with reasoning, not a black-box yes/no.
- Publishers and newsrooms vetting inbound copy for AI generation and unverified claims.
- Creators, researchers, and journalists who want to timestamp original work so authorship is provable later.
- Developers integrating detection and provenance via REST API.

## The 10-tool AI workspace

Every tool is part of the same workspace and shares quotas, history, and exports.

1. **Content Analyzer** — Paste text or upload a document; get an AI-vs-human probability, model-by-model breakdown, signals (perplexity, burstiness, stylometric markers), and sentence-level highlights.
2. **AI Humanizer** — Rewrites over-edited or AI-flagged text in a more natural human cadence.
3. **Grammar Checker** — Catches grammar, punctuation, and clarity issues with concise explanations.
4. **AI Summarizer** — Produces faithful summaries at adjustable length.
5. **Paraphraser** — Rewrites passages in a chosen tone (formal, casual, academic, simple).
6. **Citation Generator** — Generates APA, MLA, Chicago, and Harvard citations from URLs or metadata.
7. **Fact Checker** — Verifies factual claims against live web sources, with linked evidence.
8. **Plagiarism Checker** — Compares text against indexed sources for verbatim and near-duplicate matches.
9. **Quiz Generator** — Builds multiple-choice and short-answer quizzes from any source text, with timers and shareable attempt analytics.
10. **Document Chat** — Ask questions of an uploaded PDF or document; answers cite the relevant passages.

## How AI detection works

askCPL runs each submission through up to ten independent models drawn from three families:

- **Statistical detectors** measuring perplexity (how surprised a language model is by the text) and burstiness (variation in sentence-level perplexity). Human writing is typically burstier than machine writing.
- **Transformer classifiers** fine-tuned to distinguish human and AI-generated text across multiple domains and languages.
- **Stylometric analyzers** measuring lexical diversity, sentence-length variance, function-word distribution, and adversarial markers used to "launder" AI text.

Each model returns its own AI probability. The platform reports:

- **Ensemble consensus** — the weighted aggregate score.
- **Model agreement** — how strongly the models agree (low agreement = lower confidence).
- **Per-model votes** — every model's individual verdict, shown in the report.
- **Sentence-level highlights** — which specific sentences drove the verdict, with the signal that flagged each.

askCPL never claims absolute certainty. AI detection is probabilistic; the platform is a forensic aid, not an arbiter.

## How the Content Provenance Ledger works

When a user registers content:

1. A SHA-256 hash of the content is computed client-side.
2. A statistical fingerprint (n-gram histogram + perceptual hash for media) is computed.
3. Hash, fingerprint, author identity, and a timestamp are appended to an append-only Merkle chain.
4. Batches of Merkle roots are (planned) anchored on Base L2, producing an on-chain tamper-evident timestamp.
5. Revisions are linked to the original entry so a change history is preserved.

Verification: anyone can paste content or supply a registration ID at `/verify` to confirm a hash matches a registered entry and see its provenance trail. The `/chain` explorer browses the full live chain.

The CPL is the trust layer that complements detection: detection answers *"is this AI?"*; provenance answers *"who created this, and when?"*.

## Pricing

- **Free** — 2,500 words/day across all tools. One free trial analysis with no signup required.
- **Monthly — $9.99 / month** — Unlimited usage on every AI tool, full CPL access (register, batch, revisions, co-authors, API), exports without watermarks.
- **Annual — $79.99 / year** — Same as Monthly with ~33% savings. Cancel anytime.

Payments are processed via Paystack. Prices are displayed in USD and charged in GHS at the current rate (≈150 GHS monthly / ≈1,200 GHS annual). Plans can be cancelled at any time from account settings.

## Security & privacy posture

- Zero raw-content retention: only hashes and statistical fingerprints are stored.
- All endpoints require server-side JWT validation; row-level security on every public table.
- Inputs are capped at 20,000 characters per request (chunked above 8k); requests time out at 28 seconds to prevent abuse.
- Quotas are enforced server-side, not client-side.

## Contact

- Email: support@askcpl.net
- LinkedIn: https://www.linkedin.com/company/askcpl/
- Facebook: https://web.facebook.com/profile.php?id=61589010933465
- Site: https://askcpl.net