ZeroGPT (zerogpt.com) is an AI text detector and writing toolkit founded in 2022.[1] It offers a variety of writing services including humanizing, plagiarism checking, grammar checking, and paraphrasing.
Pros
- Scored 7/7 in LLM detection
- 5/7 AI Humanized text detection
- Offers a variety of products
Cons
- 4 false positives found
- Unclear credit system
- Very limited standard and free plans
- Slow, inaccurate plagiarism checker
Features
- AI detector: Detects AI-written text.
- Plagiarism checker: Catches copied writing.
- Humanizing: Makes AI text read as human.
- Email writer: AI-drafted emails.
- Translator: AI translation services.
- Images: Both AI image detection and background removal.
- Grammar checker: Spellcheck human-written content.
- API
Pricing Plans
| Free | Pro | Plus | Max |
| No credit card needed | $9.99/month or $95.88/year | $19.99/month or $179.88/year | $26.99/month or $227.88/year |
| 15,000 characters per AI detection | 100,000 characters per AI detection | 100,000 characters per AI detection | 150,000 characters per AI detection |
| Not Ad Free. One plagiarism check of 300 words/month. Not included: batch files, PDF reports, history of detections. | Ad Free + PDF reports, 50 batch files, detection history. One plagiarism check of 750 words/month. | Includes all Pro features + 100 plagiarism checks per month (25k characters max). | All Plus features + 150 plagiarism checks + access to advanced chatbot + other advanced features. |
ZeroGPT also offers an “Expert” plan that features everything Max has to offer with significantly higher limits (350k characters, 100 batch files, 100k words per month for plagiarism detection). The cost is $599.88/year.
Credits can be purchased to allow for plagiarism checking beyond the limit of a particular pricing tier. The cost is $1 per 1000 words with the Free and Pro plans, and 70¢ per 1000 words for the Plus, Max, and Expert plans.[2] However, it’s unclear what the value of an individual credit is as it’s not displayed anywhere on the website.
Test #1: ZeroGPT vs. LLMs
For the first test, we put ZeroGPT up against short articles written by leading LLMs (ChatGPT, Gemini, etc.). We prompted each LLM with the below text and then ran the generated content through ZeroGPT’s detector to see what it picked up.
Prompt Used:
I need a 500-word article about Theodore Roosevelt that is 100% accurate. It should also sound as human as possible (with the intent of passing AI detectors) and should not include an em dashes.
Expand
Collapse
| LLM | Test Result | Proof |
| ChatGPT – 5.4 Thinking Model | 88.5% AI
✅ |
 |
| Claude – Sonnet 4.6 Extended | 99% AI
✅ |
 |
| Google Gemini – Thinking | 98% AI
✅ |
 |
| Deepseek – Deepthink | 98.4% AI
✅ |
 |
| Perplexity – Free Model |
98.2% AI ✅ |
 |
| Grok – Auto |
98.9% AI ✅ |
 |
| Meta |
85.7% AI ✅ |
 |
Test #2: ZeroGPT vs. AI Humanizers
For our second test, we took the Meta AI-generated text from the first test and ran it through seven different humanizers. The humanized output was then entered into ZeroGPT to see if AI was detectable.
Expand
| AI Humanizer | Test Result | Proof |
| Humanize.ai (humanized text) |
41.8% AI 🟡 |
 |
| StealthGPT (humanized text) |
52.5% AI
🟡 |
 |
| Undetectable.ai (humanized text) |
98.3% AI
✅ |
 |
| StealthWriter.ai (humanized text) |
43.9% AI
🟡 |
 |
| WalterWrites.ai (humanized text) |
78.1% AI
✅ |
 |
| Grammarly (humanized text) |
95.7% AI
✅ |
 |
| Quillbot (humanized text) |
87.1% AI
✅ |
 |
Test #3: False-Positives + Plagiarism
The plagiarism detector was able to pick up copied text in 6 of the 10 instances.
For the final test, we input human-written articles into ZeroGPT’s AI detector to see if it came back with any false positives. We also ran the same text articles through the plagiarism detector.
Expand
| Article URL | Test Result | Detection | Plagiarism |
| “The Moon is Down”
Steinbeck |
6.9% AI, 100% Plagiarized
✅ |
 |
 |
| Winston Churchill (speech) | 53.9% AI, 10% Plagiarized
❌ |
 |
 |
| AP News (article) | 24.1% AI, N/A
✅ |
 |
Failed to perform test (upstream error) |
| Wikipedia (“elephant”) | 96.5% AI, 100% Plagiarized
❌ |
 |
 |
| The Guardian (article) | 12.9% AI, 96% Plagiarized
✅ |
 |
 |
| NPR (article) | 18.2% AI, 90% Plagiarized
✅ |
 |
 |
| Substack (post) | 9.2% AI, 0% Plagiarized
✅ |
 |
 |
| NASA – Buzz Aldrin (bio) | 47.7% AI, 100% Plagiarized
❌ |
 |
 |
| NIH – How Sleep Works | 88.9% AI, 100% Plagiarized
❌ |
 |
 |
| Harvard Business Review (article) | 15.9% AI, 28% Plagiarized
✅ |
 |
 |
Is ZeroGPT Accurate?
If this question is posed to the majority of Trustpilot reviewers, the answer would be a resounding “no.”
93% of users gave it 1 star, mostly due to general inaccuracy and false positives. Another common complaint is that it was designed to upsell the user on a humanizer.
Frequently Asked Questions
Do unused credits carry over?
Yes, credits can be purchased at any time and never expire (see image).
Are sentence-by-sentence breakdowns provided?
No. While it does highlight the sentences it believes to be AI, ZeroGPT doesn’t offer any explanation or score (see image).
Is document scanning provided?
Yes, both PDF and MS Word documents can be uploaded directly into the AI detector homepage.[3]
Is image scanning provided?
Yes, users can upload images to ZeroGPT to detect whether they were AI generated.[4]
Are reports shareable?
No, not directly, but the results can be exported as a PDF and shared.
Does ZeroGPT provide certifications?
No, there are no certifications attached to the detector results.
Are AI detections used for internal training?
No, on their FAQ page, ZeroGPT states that the inputted text is not used, saved, published or used to train their models.[5]