AI Content Detector — Turnitin Alternative

Check English prose for AI-writing signals. Submit text and receive a provider signal, review band, optional sentence highlights, and human-review guidance in one report per text. Results are screening cues, not proof of authorship. For editors and review teams, each dataset row is one text-screening report with a review band and provider provenance.

Workflow: put the results to work

Submit intact English prose within the supported length range and keep each document's identifier. Use the review band to organize an editorial queue, inspect any highlighted sentences in context, and let a human reviewer decide the next step. The result does not identify which model wrote a passage.

What you get

Field	Meaning
reviewBand	no-signal, low-signal, review, or high-signal routing cue for a human queue
reportedAiPercentage	Cautiously displayed provider signal; intentionally omitted for 1-19% low-signal pre-screen results
classification	Threshold-based human, mixed, or ai label for workflow routing, not a finding
reviewGuidance	Mandatory reminder that a human reviewer and the applicable policy decide next steps
aiSentences[]	Optional provider-highlighted sentences, stored as review cues rather than evidence
provider, detectedAt, textId	Provenance that lets a reviewer trace every saved report

The default turnitin-pre-screen profile accepts only intact 300-3,000-word sections. It declines short, oversized, or truncated submissions before calling the provider, so an ineligible input creates no provider request and no report usage event.

When to use it

• Triage a queue of authorised essay, article, or application sections before assigning human review.
• Add a bounded AI-writing signal to an editorial or academic workflow that already stores source text elsewhere.
• Call an API and receive one durable dataset row per eligible section, with an OUTPUT / RUN_SUMMARY contract.
• Keep text preview disabled when a score and provenance are sufficient for the review queue.

Keep every disciplinary, employment, admissions, or other high-impact decision with an authorised human reviewer. Obtain permission to process the text, apply the relevant policy, and provide an appeal or review path where required.

AI tools and writing sources covered by the workflow

The Actor is model-agnostic: it screens writing patterns rather than calling or reverse-engineering the tool that may have produced the text. Typical review inputs include content associated with:

• ChatGPT and OpenAI GPT models
• Claude
• Gemini
• Microsoft Copilot
• DeepSeek
• Grok
• Meta AI and Llama-based assistants
• Mistral and Le Chat
• Perplexity
• Jasper, Copy.ai, Writesonic, and Rytr
• QuillBot and other AI paraphrasing or rewriting tools
• Other current or future LLM-based writing assistants

These names describe possible text sources users may want to screen. ZeroGPT and this Actor return a general signal rather than named-model attribution, and performance can vary across models and versions. Human editing, translation, paraphrasing, text length, genre, and model updates can all affect the signal. The listed companies are unaffiliated with this Actor.

Workflow: from authorised text to a human queue

An editor starts with an authorised 800-word article section, pastes it into texts, and starts the default profile with includeTextPreview: false. The Actor first checks the word range and truncation rule. It then sends only an eligible section to the provider, saves one report row, and turns the row into a review-queue item using reviewBand and reviewGuidance. Next, the editor reads the source in its original system alongside any optional highlights and records the human decision there. For recurring work, schedule the same saved task and export the dataset to the downstream review system after each completed execution.

Input

Field	Default	Purpose
texts	one prefilled section	Up to 10 pasted sections, one report per eligible item
text	empty	Single-section API shortcut; combines with texts if both are supplied
reviewProfile	turnitin-pre-screen	Safe default with 300-3,000-word validation and cautious low-signal display
maxTexts	1	Per-execution report and cost cap; maximum 10
maxCharacters	20000	Safeguard cap; an otherwise eligible pre-screen is declined if this would truncate it
includeSentences	true	Include provider-highlighted sentences as review cues
includeTextPreview	false	Add a short source preview only when the workflow needs it
aiThreshold / humanThreshold	80 / 20	Routing thresholds that leave the provider result unchanged
maxConcurrency	1	Stable provider pacing; maximum 3

Example: one pre-screen

{
  "texts": ["Paste one intact 300-3,000-word English prose section here."],
  "reviewProfile": "turnitin-pre-screen",
  "maxTexts": 1,
  "includeSentences": true,
  "includeTextPreview": false,
  "maxConcurrency": 1
}

Example: a general non-academic score

{
  "text": "Paste a short product description or support reply here.",
  "reviewProfile": "general",
  "maxTexts": 1,
  "includeSentences": false,
  "includeTextPreview": false
}

Output and execution contract

Each eligible section yields one dataset row. The row preserves the raw aiPercentage for auditability and supplies reportedAiPercentage, reviewBand, requiresHumanReview=true, and reviewGuidance for a safer queue experience.

OUTPUT and RUN_SUMMARY contain the terminal outcome, queued and saved counts, profile eligibility, provider attempts and retries, status codes, warnings, errors, and usage-event counters. Treat an UPSTREAM_FAILED or PARTIAL execution as a retry/review condition; an absent row is a signal to review the terminal summary.

Dataset view	Use it for
reviewQueue	Sort eligible sections by cautious review band and human-review requirement
evidence	Inspect provider signals, highlights, feedback, and timestamped provenance

API and automation

Start the Actor through the Apify API, saved tasks, schedules, webhooks, or an Actor chain. A successful execution writes its dataset and summary to Apify storage, so downstream systems can read results by identifier without scraping a user interface.

curl -X POST "https://api.apify.com/v2/acts/khadinakbar~turnitin-ai-detector-alternative/runs" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "texts": ["Paste one intact 300-3,000-word English prose section here."],
    "reviewProfile": "turnitin-pre-screen",
    "maxTexts": 1,
    "includeSentences": true,
    "includeTextPreview": false
  }'

Retrieve dataset rows at https://api.apify.com/v2/datasets/{datasetId}/items?clean=true and the execution summary at https://api.apify.com/v2/key-value-stores/{storeId}/records/OUTPUT. Use ordinary dataset offset and limit pagination for downstream review queues.

AI-agent review prompt

Screen this authorised 1,000-word English essay for AI-written-content signals. Return the dataset row's reviewBand, reportedAiPercentage, reviewGuidance, provider, and timestamp; then read back OUTPUT.outcome. Leave source-model attribution unspecified. For PARTIAL, UPSTREAM_FAILED, or INVALID_INPUT, surface the diagnostic or eligibility warning for a human.

The Actor is designed for an AI-agent workflow that needs bounded input, a saved dataset result, explicit outcome semantics, and pagination. Decision authority remains with the writer-review process and its human reviewer.

Evidence and freshness

Source metadata, input rules, output fields, and provider-cost policy are recorded in the Actor source. Release verification, delayed storage readback, and usage evidence stay in the release log rather than becoming timeless marketing copy. Refresh provider pricing against the owner account before changing public price or production volume.

Design note

I designed this workflow around a narrow question: how can a review team receive one explicit signal without confusing that signal with a decision? I chose a bounded long-form input, provenance fields, a cautious low-signal display, and mandatory human-review guidance so the output fits a real queue rather than a hidden scoring system.

How this alternative compares with Turnitin

This comparison covers a narrow AI-written-content screening workflow. Turnitin's AI-writing feature is documented within its enhanced Similarity Report and institutional licensing context; this Actor provides a separate ZeroGPT-backed pre-screen for API-led queues. Suite and detection parity remain outside the comparison.

Decision dimension	This Actor	Turnitin workflow	What that means
Scope and workflow	Pasted, intact 300-3,000-word English prose becomes a dataset-backed pre-screen report.	Turnitin documents AI-writing detection in its enhanced Similarity Report.	Use this Actor for the bounded API pre-screen; retain Turnitin when the institutional report workflow is required.
Billing and cost	The live Pricing tab shows the per-saved-report event and additional Apify platform usage.	Turnitin's documented feature is available within a licensed product environment; this comparison has no matching public self-serve same-job price.	The Actor exposes a per-report event; a price comparison requires a current Turnitin quote.
Effective efficiency	Eligibility validation avoids sending short, oversized, or truncated default-profile inputs to the provider.	Turnitin's suitability depends on its licensed report flow and policy configuration.	Both require human judgment; this Actor is intended for a small, pre-defined review step.
Output contract	Structured dataset fields, terminal outcomes, provider label, and timestamp are available to API, schedule, webhook, and Actor-chain workflows.	Turnitin's enhanced Similarity Report is a broader institutional report surface.	The API contract supports automation, but it does not replace institutional dashboard, policy, or collaboration capabilities.
Integration and automation	Apify runs can be started and read programmatically with storage-backed results.	Turnitin may be the better fit where its licensed integrations and institutional workflow are needed.	Select based on the actual operating workflow, not a score alone.
Reliability and provenance	The Actor records provider attempts, retries, status codes, and a timestamped report row when it succeeds.	Turnitin documentation cautions that AI-writing detection should not be the sole basis for adverse action.	Keep a human reviewer and the relevant policy in every decision path.

Turnitin is a trademark of Turnitin, LLC. This independent Actor is not affiliated, associated, or endorsed by Turnitin, LLC.

Scope, boundaries, and responsible use

• This Actor is an independent provider-backed signal, not Turnitin and not a reconstruction of Turnitin's proprietary detection method.
• It returns an estimated signal, never a verified authorship determination.
• The default pre-screen profile is intentionally narrow: English prose, 300-3,000 words, intact source text, and human review required.
• Send only text you are authorised to process. The text is sent to the selected detection provider for scoring.
• Consult applicable academic, workplace, privacy, and anti-discrimination requirements before using the output.

FAQ

Why does the default profile reject text outside 300-3,000 words?

The profile is designed for long-form pre-screening. Enforcing a visible range prevents a short phrase or silently truncated document from being treated like an evidence-bearing report.

Does an ai classification prove the text was written by AI?

No. It is a routing label based on a provider score and your thresholds. Every saved default-profile report sets requiresHumanReview to true.

Can it detect ChatGPT, Claude, Gemini, Copilot, DeepSeek, Grok, or Llama?

It can screen text that may come from those and other AI writing tools, while returning a general ZeroGPT AI-writing signal. Originating-model identification is outside the report contract, and performance can vary as models and editing techniques change.

Is a ZeroGPT key required from the caller?

No. The Actor uses an owner-managed provider key. You are still responsible for making sure you may submit the text for processing.

What if the provider is unavailable?

Retryable provider interruptions receive backoff retries. When no row can be saved, RUN_SUMMARY records UPSTREAM_FAILED; when only some rows succeed, it records PARTIAL. The primary report event applies only to saved rows.

Pricing and run costs

This Actor uses Pay per event plus Apify platform usage. The Pricing tab lists the current event rates and billing terms.

Event	Billing unit	When it applies
apify-actor-start	Actor Start	Charged when the Actor starts running. Number of events charged depends on Actor memory (one event per GB, minimum one event).
ai-writing-pre-screen-completed	AI-writing pre-screen report	One text scored by ZeroGPT and saved as a JSON report for human review. Not an authorship or misconduct finding.

Run cost combines the charged events and Apify platform usage. Review the run charge limit and requested result count before starting.

Connect an AI agent

Use the Apify MCP configurator to choose an available client connection. Inspect this Actor’s current input schema and required credentials before running it.