Sourcing
- Factual claims about detector accuracy, research findings or vendor behaviour are linked to a primary source — a peer-reviewed paper, the vendor's own documentation, or original reporting. Not to another blog summarising one.
- Where we cite a number, we name the study, the publication and the year in the text, so it can be checked without hunting.
- We do not cite competitor marketing pages as evidence of anything except what that competitor claims.
What we will not publish
- Bypass guarantees. We do not claim our product, or any product, can reliably defeat a named AI detector. Detectors change without notice and disagree with each other; a promise about a future score is not one anybody can keep.
- Invented statistics. No accuracy figure appears on this site unless we can point at where it came from.
- Fabricated reviews, ratings or testimonials. When social proof appears here it will be from real, identifiable users, and marked as such.
- Advice on evading academic integrity processes. We write about how detection works and what to do about false accusations. That is a different thing, and we intend to keep it different.
Review and updates
- Pages making factual claims carry a "last reviewed" date. That date means a person re-read the page and checked its sources still say what we said they say — not that a word was changed.
- Research-dependent pages are reviewed at least every six months.
- When significant new research changes a conclusion, we update the page and say what changed rather than quietly editing it.
Corrections
If something here is wrong, tell us at support@textsight.ai and we will fix it. Substantive corrections — anything that changes a conclusion — get a dated note on the page rather than a silent edit.
Conflict of interest
We make an AI humanizer and an AI detector. That is a direct commercial interest in this subject, and you should read our writing about it with that in mind.
The commitments that follow from it: our own detector is subject to the same criticisms we make of everyone else's, and we say so on the pages where it comes up. We do not publish comparisons that exist only to make us look good. If a competitor is better for a particular use, we would rather say it than be caught not saying it.
Use of AI in our own writing
It would be absurd for us to pretend otherwise: we build AI writing tools and we use AI tools. What we commit to is that every page is read, edited and fact-checked by a person before publication, that the sources are verified by a person, and that nobody publishes a claim here they have not checked. The byline is accountable for the page regardless of what helped draft it.
Who writes this
Content is produced by Lacewing Technologies, the team behind HumanFlow. Reach us at support@textsight.ai or via the contact page.