AI Content Detector

Score pasted essays, applications, ads, and SEO drafts with ZeroGPT and get one dataset row per document: an AI percent, a human or mixed label, word counts, and the sentences the detector highlighted. Built for teachers, editors, recruiters, content teams, and MCP agents that need Copyleaks/QuillBot-style detection as structured JSON instead of a screenshot from a website.

Designed for people who already have the wording in hand and need a repeatable score they can store, compare, and bill per document. After you score a brand's generated copy, continue with AI Search Brand Monitor to see whether AI answers mention that brand.

What you get

Output	What it answers
classification	Label — ai, mixed, or human from the ZeroGPT percent
aiPercentage + humanPercentage	Score — ZeroGPT's AI share and human share, 0–100
textWordCount + aiWordCount	Volume — how much of the document was counted as AI-written
aiSentences[] + feedback	Highlights — the sentences ZeroGPT marked, plus the plain-language verdict
textPreview, provider, detectedAt	Provenance — which document, which engine, and when

Who it is for

• Teachers and academic integrity staff — score submitted essays and keep a dated row per student document.
• Editors and SEO teams — check drafts and ad copy before publish, then keep the highlighted sentences for revision notes.
• Recruiters and HR coordinators — score cover letters and take-home writing with a consistent threshold.
• Agency QA — batch-check a folder of client drafts in one run.
• AI agents — call one tool, read classification + aiPercentage, and route the next step from OUTPUT.outcome.

When to use this Actor

• Paste one essay, cover letter, or ad and get a ZeroGPT score plus highlighted sentences.
• Batch-score a class set or a week of blog drafts, one row per document.
• Give an agent a Copyleaks-style JSON contract: label, percent, sentences, provenance, and cost.
• Keep includeTextPreview on when you need to match a score back to the source wording later.

Best fit & sibling routing

Use this Actor when the job is score wording you already have. Pair it with ChatGPT GPT Store Scraper when you first need listing copy from the GPT Store, then pass that copy here to score it. For whether Google's AI Overview cites a brand, start with Google AI Overviews Scraper, then score any generated follow-up copy with this Actor. This Actor is designed as a focused standalone workflow for ZeroGPT text scores; file-upload and URL-crawl modes stay with dedicated crawlers.

Input

Field	Type	Default	Description
texts	array	1 sample document	Up to 50 strings, one document each
text	string	empty	Optional single-document shortcut for agents
maxTexts	integer	50	Cap how many documents this run scores
maxCharacters	integer	4000	Per-document cap; longer input is truncated and flagged
includeSentences	boolean	true	Add ZeroGPT highlighted sentences
includeTextPreview	boolean	true	Store a short preview of the scored wording
aiThreshold	integer	80	classification=ai at or above this percent
humanThreshold	integer	20	classification=human at or below this percent
maxConcurrency	integer	2	Documents scored in parallel (1–5)

Example input — one essay

{
    "texts": [
        "Artificial intelligence has transformed the way organizations approach content creation, enabling teams to produce high-quality written materials at unprecedented speed and scale. By leveraging large language models, businesses can generate drafts, summarize research, and personalize communications while maintaining a consistent brand voice across every customer touchpoint. This technological shift requires robust governance, including human review, disclosure policies, and authenticity checks, so stakeholders can trust the information they consume."
    ],
    "includeSentences": true,
    "maxTexts": 1
}

Example input — classroom batch

{
    "texts": [
        "Artificial intelligence has transformed the way organizations approach content creation, enabling teams to produce high-quality written materials at unprecedented speed and scale. By leveraging large language models, businesses can generate drafts, summarize research, and personalize communications while maintaining a consistent brand voice across every customer touchpoint. This technological shift requires robust governance, including human review, disclosure policies, and authenticity checks, so stakeholders can trust the information they consume.",
        "The morning market opened mixed as traders digested overnight futures and a handful of earnings prints. I walked down to the corner shop, bought coffee, and sat on the bench watching buses. Nothing about this paragraph is trying to sound corporate. It is just a record of an ordinary Tuesday with a few numbers mixed in so the detector has enough words to score."
    ],
    "maxTexts": 50,
    "includeSentences": true,
    "aiThreshold": 80,
    "humanThreshold": 20
}

Output

One dataset row per scored document.

Field	Type	Meaning
textId	string	text-1 or a caller-supplied id
classification	string	ai / mixed / human
aiPercentage	number	ZeroGPT AI share, 0–100
humanPercentage	number	ZeroGPT human share, 0–100
textWordCount	integer	Words counted
aiWordCount	integer	Words counted as AI
aiSentences	string[]	Highlighted likely-AI sentences
feedback	string	Plain-language ZeroGPT verdict
textPreview	string	Leading characters of the document
provider	string	Always zerogpt
detectedAt	string	UTC timestamp

Example output — one persisted item

{
    "textId": "text-1",
    "classification": "ai",
    "aiPercentage": 100,
    "humanPercentage": 0,
    "textWordCount": 68,
    "aiWordCount": 68,
    "characterCount": 548,
    "truncated": false,
    "feedback": "Your Text is AI/GPT Generated",
    "aiSentences": [
        "Artificial intelligence has transformed the way organizations approach content creation, enabling teams to produce high-quality written materials at unprecedented speed and scale."
    ],
    "aiSentenceCount": 3,
    "textPreview": "Artificial intelligence has transformed the way organizations approach content creation",
    "provider": "zerogpt",
    "providerResultId": 58839863,
    "detectedAt": "ISO-8601 timestamp"
}

OUTPUT and RUN_SUMMARY record the run outcome (COMPLETE, PARTIAL, VALID_EMPTY, INVALID_INPUT, UPSTREAM_FAILED, CONFIG_ERROR), itemsPushed, classification counts, and chargedEventCounts.

Workflow: from a draft folder to a review queue

A content lead starts with 20 blog drafts in a spreadsheet, pastes each body into texts, and runs this Actor with includeSentences: true. Rows with classification=ai go to a human editor with the highlighted sentences as revision notes. Mixed rows stay in a second queue. Human rows ship. Schedule the same saved input weekly for new drafts. Then export the dataset to Sheets and filter on aiPercentage.

Pricing — Pay per event + platform usage

Event	Price	Charged when
Actor start	$0.00005	Once per run (scaled by memory)
text-detected	$0.02	One document scored and saved

Cost scenarios (typical runs)

Run	Texts saved	Event cost (approx.)
One essay	1	~$0.02
10-draft batch	10	~$0.20
50-document max batch	50	~$1.00

Only saved rows are billed. Then check the live Pricing tab for current Pay per event plus platform usage on your plan.

API example

curl -X POST "https://api.apify.com/v2/acts/khadinakbar~ai-content-detector/runs?token=$APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "texts": ["Artificial intelligence has transformed the way organizations approach content creation, enabling teams to produce high-quality written materials at unprecedented speed and scale. By leveraging large language models, businesses can generate drafts, summarize research, and personalize communications while maintaining a consistent brand voice across every customer touchpoint."],
    "includeSentences": true,
    "maxTexts": 1
  }'

Results land in the run's default dataset: https://api.apify.com/v2/datasets/{datasetId}/items?clean=true. Paginate large batches with offset/limit on that endpoint. The same actor is exposed through Apify MCP as apify--ai-content-detector for agent clients.

Use with AI agents through Apify MCP

Score this pasted essay with ZeroGPT. Return classification, aiPercentage, highlighted sentences, dataset rows, OUTPUT outcome, provenance (provider + detectedAt), and confirm the $0.02 per-text cost. Read back the default dataset when the Actor completes.

Agents should pass texts (or text for one document), keep maxTexts bounded, and treat OUTPUT.outcome as the machine contract. Scope is pasted wording only; pagination is offset/limit on the dataset items endpoint.

Best results guidance

• Give ZeroGPT 60+ words per document so sentence highlights have something to mark.
• Keep one student, ticket, or URL's wording per array item so textId maps back cleanly.
• Use aiThreshold 80 / humanThreshold 20 unless your review policy needs a stricter AI band.
• Split long articles at 4,000 characters for more precise highlights and a predictable $0.02 row.
• Turn includeTextPreview off for classrooms that should store scores without retaining the original essay in Apify.

Builder's note

I built this after calling ZeroGPT's Business detectText API against both a corporate AI-style paragraph and a messy diary paragraph: I found the official payload already returns fakePercentage, isHuman, feedback, and highlighted sentences in h[] — the same shape Copyleaks and QuillBot show in a UI, only as JSON. In my testing, wrapping that engine as an Apify PPE actor was more useful for agents than scraping the public ZeroGPT website, because the score, the highlights, and the billing unit stay on one row.

FAQ

How is classification decided? ZeroGPT returns fakePercentage. This Actor labels ai at or above aiThreshold (default 80), human at or below humanThreshold (default 20), and mixed in between. The raw percents stay on the row.

Does a user need their own ZeroGPT key? No. The Actor uses the owner's managed ZeroGPT Business API key. You paste text and receive scores.

What happens if ZeroGPT is briefly unavailable? The Actor retries retryable HTTP statuses, then records that document in OUTPUT. A run that saved at least one row finishes PARTIAL; a run where every document stayed unavailable finishes UPSTREAM_FAILED. Only saved rows are charged.

Can an AI agent use this? Yes. It is MCP-ready: texts in, one structured row per document, with outcome readback, provenance, cost, and dataset pagination.

Legal

Submit only text you are authorized to process. You are responsible for complying with applicable laws, ZeroGPT's terms of service, classroom or employer policy, and Apify's terms when using the scores. AI-detection percentages are estimates for review workflows, not a legal finding of authorship.