Free AI detectors are being used to judge students, freelancers and job applicants. So we put four of them to the same test, on the same day, with no sign-ups. One of them says the Declaration of Independence is 100% AI.
The test
We pasted the same four texts into four free AI detectors that anyone can use without creating an account. Every test was run on 17 October 2026, and we re-ran several of them later the same day to confirm the scores didn’t change. They didn’t.
- Human: the opening of the Declaration of Independence (1776).
- Human: the Gettysburg Address (1863).
- Pure AI: a paragraph about remote work generated by an AI model, untouched.
- AI + a 2-minute edit: the same paragraph, rewritten by hand in plain, casual language.
The results
| Detector | Declaration (human) | Gettysburg (human) | Pure AI | AI + 2-min edit | Correct |
|---|---|---|---|---|---|
| ZeroGPT | ❌ 100% AI | ❌ 93.9% AI | ✅ 100% AI | ❌ 17.9% (“human”) | 2 / 4 |
| QuillBot | ✅ 0% AI | ✅ 0% AI | ✅ 100% AI | ❌ 0% AI | 3 / 4 |
| GPTZero | ✅ 99% human | ✅ 100% human | Locked | Locked | 2 / 2 |
| Sapling | ✅ 1.2% AI | ✅ 0.4% AI | ✅ 99.7% AI | ✅ 99.3% AI | 4 / 4 |
ZeroGPT: 2 out of 4
ZeroGPT flagged both historical documents as AI, including a 100% score for the Declaration of Independence, then called the lightly edited AI text “human written”. Under every result sits a big “Humanize Text” button: the detector is also selling the tool to beat the detector.
QuillBot: 3 out of 4
QuillBot got the easy cases right, but two minutes of rewriting took the AI paragraph from 100% to 0%.
GPTZero: 2 out of 2 (then locked)
GPTZero correctly called both historical texts human. On the third scan it asked us to create an account to keep going, so we couldn’t test the AI texts without signing up. We didn’t count that against it, but it’s worth knowing that “free” means about three checks.
Sapling: 4 out of 4
Sapling got everything right in this test, including flagging the rewritten paragraph at 99.3%. That’s impressive, or worrying, depending on whether you think rewriting something in your own words makes it yours.
The ugly truth
This was four texts, checked on one day. It’s not a scientific study, and detectors change their models all the time. But if a tool can call the Declaration of Independence AI even once, its score should never be the only evidence against a real person.
- If you get flagged: ask which tool was used, and show your work: version history, drafts and notes.
- If you’re the one checking: treat a detector score as a question, not an answer.
No sponsors and no affiliate links in this post. We’re not affiliated with any of the tools tested. The video narration is AI-generated; the testing, script and fact-checking are done by the Ugly Truth team.