detectiondrama.com exists for one moment: the one where a detector puts a number on something you wrote, and that number is wrong.
Maybe it was a student essay you spent two weeks on. Maybe it was a client article you drafted, edited, and rewrote by hand. Either way the screen says a percentage, somebody upstream treats that percentage as evidence, and suddenly you are arguing about your own writing.
This site is where I write down what actually happens when you run text through these tools — the detectors, and the humanizers people reach for after a detector flags them.
The only tool that bypassed Pangram and Turnitin in 2026.
Most humanizers clear one detector and get caught by the other. StealthWriter is the one that gets past both — run Ghost 5.2 Pro at level 7–8, section by section, and re-check before you submit.
Try StealthWriter →Who runs this site
Vlad Ivanov — I build and rank content sites, and I have been publishing about AI writing and SEO since well before it was a category.
My main channel, @wordsatscale, has 26,000+ subscribers and 466 videos on AI-assisted content and search. detectiondrama.com is the site where the AI-detection half of that work lives.
I am not an academic-integrity office, a university, or a detection vendor. I have no contract with any of the companies I write about, and I am not the person who decides whether your paper passes. I am a practitioner who runs the tools, records what they output, and publishes it.
What gets published here
- Detector behaviour. What GPTZero, Turnitin, Originality.ai, Pangram and the rest actually return on specific text — including text written entirely by a human.
- Humanizer reviews. Does a given tool change the score, and what does it do to the sentence while it is at it? A tool that beats a detector by turning your prose into mush has not solved anything.
- Head-to-head comparisons. Same input, several tools, same day, results side by side.
- Free tools. Things you can run in the browser without paying anyone — a Turnitin score predictor, a text humanizer, a watermark checker, a false-positive defence kit.
How the testing works
Nothing here is a vibes review. The method is dull on purpose:
- The same input text goes into every tool in a comparison, on the same day.
- Scores are recorded as screenshots, not paraphrased from memory.
- The date is on the page, because a result from March may be wrong by September. These models get retrained without announcements.
- When a re-test contradicts something already published, the post gets updated and says what changed.
The honest limitation. Detectors are probabilistic, they change without notice, and no result on this site is a guarantee about the result you will get tomorrow on your text. Anyone promising a permanent bypass is selling something. Treat every score here as a dated observation, not a promise.
How the site makes money
Affiliate links. If you buy a tool through one of them, I earn a commission at no extra cost to you. That is disclosed on the pages where it applies and covered in full on the disclaimer.
What that does not buy: a good review. Tools that failed a test get written up as tools that failed a test — including ones that pay well. The site is worth nothing to you, and therefore worth nothing to me, the first time it reads like a brochure.
Where I stand on academic integrity
Two different people land on this site. One is trying to prove work they genuinely wrote is theirs. The other wants to run an AI draft past a checker.
I write for the first. Passing off generated work as your own where you have agreed not to is cheating, and no tool reviewed here changes that. Your school’s rules are your school’s rules, and a percentage on a screen is not a defence in either direction.
Get in touch
Corrections, test results that contradict mine, guest post pitches, link insertions: the contact page has a form and it comes straight to me. Corrections get priority — if something on this site is wrong, I would rather hear it from you than leave it up.