Skip to content
vs

AI Detector

Pangram vs GPTZero: AI Detector (2026)

Pangram leads with a cited accuracy number: 99.98%, backed by University of Chicago Booth and University of Maryland research, for a reader who wants the evidence up front. GPTZero leads with workflow instead: Canvas and Google Docs integrations that put the check where a teacher already grades.

Pangram and GPTZero are both text-only AI detectors with classroom integrations, but they make different bets on accuracy positioning. Pangram reports a 99.98% accuracy rate and a 1-in-10,000 false-positive rate, citing University of Chicago Booth and University of Maryland research; the free plan gives 4 credits per day, with one credit covering up to 1,000 words. GPTZero focuses on classroom workflow: Canvas and Google Docs integrations and a free plan of around 10,000 words per month. Slop or Not runs text checks on iPhone and Mac on-device.

Privacy, Pricing, and Workflow Matrix

Public product, pricing, documentation, and benchmark pages can change. Source links are listed below the table so the cells stay readable.

FeatureSlop or NotPangramGPTZero
Best fit

On-device iPhone and Mac checks or free browser checks for text and images

AI text checks for educators, publishers, and institutions (vendor claims 99.98% accuracy)

AI text checks for teachers, writers, students, recruiters, and teams

Text detection

Checks writing in the native app and online checker

Labels text as human, AI, or AI-assisted and reports how much reads as AI-generated

AI Detector and Advanced AI Scan products

Image detection

Checks photos, synthetic images, and provenance clues in the app or online checker

No

No

GPTZero is text-only; its "AI Vision" scans for AI-written text, not images

Deepfake detection

Covers deepfake and synthetic-image checks as image detection

No

No

Metadata and provenance signals

Reads SynthID, C2PA, and IPTC signals in native and online image checks when a file carries them

Not publicly clear

Not publicly clear

Manual forensic tools

Reports detector and provenance evidence in native and online checks; not a manual forensic editor

No

No

Evidence level

Probability verdict plus provenance signals, with internal-test caveats instead of 99.9% certainty claims

Returns an AI percentage with interpretability highlights and PDF or DOCX upload

Advanced AI Scan and AI probabilities are listed

Browser checker

Free online checker runs in any browser on Numen's private Mac server; submissions are deleted after processing

Web app plus a Chrome extension and Google Docs integration

Web detector plus Chrome extension

Native Apple app

Native iPhone and Mac app for on-device checks; browser users can use the free online checker

No

No

Web app and browser extension; no native iPhone or Mac app

API workflow

Pro Mac app includes a local CLI and MCP server; no public hosted API

Developer API with pay-as-you-go credits is listed

Developer API integration is listed

Classroom workflow

Useful for checking student writing on-device in the app or through deleted-after-processing online checks

Institutional License adds LMS integrations and unlimited checks for schools

Canvas and Google Docs integrations are listed

Free tier & limits

Free daily checks in the iPhone and Mac app and a free online checker in any browser. Pro removes the native daily cap; the native app has no per-document word cap.

Free plan lists 4 credits per day with no payment method; one credit covers up to 1,000 words

Free plan (~10,000 words/month, 10,000-char per-scan limit); paid Essential, Premium, and Professional plans are listed

Privacy

Native checks stay on the iPhone or Mac and never send submitted text or images to a third-party AI provider. Online checks run on Numen's own Apple hardware, and submissions are deleted after processing.

Text is sent to Pangram and checked in the cloud to return a result

Its privacy policy lists content submitted for detection as a collected data category, alongside account, contact, and payment data; retention for web-submitted content is not stated

Main limitation

On-device privacy requires iPhone or Mac; browser checks run on Numen's private Mac server, not on-device

Text only; regular use beyond the 4 daily free credits moves to paid credit plans

Public pages center on text detection and integrations

✓ = supported, ✕ = not offered, – = not publicly documented. Details reviewed June 2026 from public product, pricing, documentation, and benchmark pages.

Sources reviewed

Where Slop or Not Fits

Pangram's four daily credits run dry around 4,000 words, and GPTZero's free plan caps near 10,000 words a month. Slop or Not skips the credit system entirely — limited daily checks, no per-document word cap, on iPhone or Mac.

  • Slop or Not runs its own detection models, so no third-party AI provider sees your text or images. Native checks run on the Apple Neural Engine; online checks run on Numen's self-hosted Apple hardware, and submissions are deleted after processing.
  • Free daily checks in the iPhone and Mac app and in the free online checker. Pro removes the daily cap, and native checks have no per-document word cap.

How They Compare by Workflow

Slop or Not is not the right answer for every workflow — pick the tool whose strongest workflow matches the check you need to run.

Slop or Not

Use it for people who want native Apple checks or the free online checker for text, images, and provenance clues.

Pangram

Use it for academic and publisher text detection it markets as high-accuracy.

GPTZero

Use it for teachers and teams that want a text detector inside classroom or document workflows.

Related Comparisons

Keep comparing adjacent tools before you pick a workflow.

Privacy tools from the same team

More From Numen

AI can now fake a profile photo or wrap a scam in a link that looks safe to tap. Numen builds the apps that put you back in charge: spot AI-generated text and images with Slop or Not, run the best open-source models uncensored and private with Private LLM, and strip trackers from links and QR codes before they follow you with Clean Links.

One belief runs through all three: your data belongs to you, and no one else.