Editorial Policy

Our guides and comparison pages are read by students, writers and editors making real decisions. This policy states how that content is produced, verified, kept current and corrected.

How we write

• Direct answers first. Every page opens with a 2-3 sentence answer to its title question, then details. You should never scroll past filler to find the point.
• Measured claims over marketing. When we cite our own tool's behavior, we use measured before/after values from our detection engine, dated. When a rewrite didn't help, we say so.
• Competitor fairness. Comparison pages state where competitors are better. Pricing and free-tier facts reflect the competitor's own site copy at the noted verification date, and change — always confirm before paying anyone.
• No fake testimonials. We publish no user quotes we cannot verify.

How we verify

• Tool claims (word limits, login requirements, pricing) are checked against the tool's own public pages on the date noted in the page.
• Claims about detector behavior and accuracy reference public documentation or research where available (e.g., vendor accuracy pages, independent studies on false positives).
• Our own numbers come from the documented methodology — the same engine, the same signals, reproducibly.

Update cadence

• Guides and how-tos: reviewed at least quarterly; sooner when a detector update materially changes the advice.
• Comparison pages: re-verified monthly, with the verification date updated on page.
• This policy and the methodology: updated whenever practice changes, with changes noted on page.

Corrections

Found something wrong or stale? Email [email protected] with the page URL and the issue. Substantive corrections are made within five business days and noted on the page; minor typos are fixed silently. We don't quietly delete criticism — if a claim of ours was wrong, the correction says so.

Independence

We recommend competitor tools where they're the better fit, and say so on the relevant pages. No page on this site is sponsored, and we accept no payment for placement in any comparison.

How a page gets written

Every guide follows the same pipeline, and the steps are deliberately boring:

Our source hierarchy

When sources disagree, we rank them and say which one we followed:

1. First-party documentation — a vendor's own accuracy page, a university's own policy PDF, Google's own Search Central guidance.
2. Peer-reviewed or independent research — published studies on detector false-positive rates, for example.
3. Documented community experience — recurring patterns in public forums, cited as anecdote, never as fact.

Marketing pages sit outside the hierarchy: we quote their pricing (it's a fact about their product) but not their accuracy claims (it isn't verifiable by us).

Gray areas, and how we handle them

• Competitor prices drift. We date every comparison and re-verify monthly. If you find one stale, the correction link below wins you our gratitude and a fixed page.
• Detectors update silently. Advice like "ZeroGPT is jumpy on short texts" can age. Where a claim depends on a detector version, we date it and revisit quarterly.
• Our own tool. Same rules, applied to ourselves — our methodology page publishes weights, calibration results and known over-flags, including the ones that make us look worse.

What we never do

• No fabricated testimonials or review scores — if we ever show user quotes, they'll be verifiable.
• No paid placement, sponsorship or affiliate links that influence rankings in our comparisons.
• No scraping competitors' claims into "facts" without their own pages as source.
• No quietly deleting criticism of our tool — corrections are noted on the page.

Readers make this page better

The fastest corrections we've received came from readers — a stale price, a changed policy, a detector that behaves differently in the wild. Email [email protected] with the URL and what's wrong; substantive fixes land within five business days with a note on the page. Topic suggestions go to the same address, and questions we can't yet answer honestly become research items rather than confident paragraphs.

Related: the tool methodology · about the team · privacy policy.

How we use AI in our own writing

Fair question, honest answer: we use AI assistance for research organization and first-draft compression, the same way many professional newsrooms now disclose. Every page is human-edited, human-verified and human-owned — the facts, the judgments and the mistakes. Where AI assistance touched a claim, the claim still has to survive the source hierarchy above or it doesn't ship. We hold our own content to the policy on this page, including the parts that make us look imperfect.

Page types and their rhythms

Corrections in practice

Two examples of how this works, so it's not abstract. When a comparison page said a competitor offered a free tier that they had since gated behind an account, the fix was same-day: the row changed, the verification date updated, and a sentence noting the change stayed on the page for the month — because readers who remembered the old claim deserved the bridge. When our own engine's mixed-text over-flag came to light in calibration, it went straight onto the methodology page as a known limitation with the numbers attached. Slow corrections erode trust faster than the original mistake; the five-day window exists to keep us honest about that.

What this site deliberately does not cover

Focus is an editorial decision too. We don't publish news churn ("detector X shipped an update"), gossip about competitors' funding, or predictions about AI policy — none of it helps you fix a draft tonight. We don't run user-submitted essays as case studies unless the writer explicitly offers one, and we don't cover detection of images, audio or code: text is what our tool does, and a page we can't verify adds noise, not value. When a topic sits at the edge of that scope — say, a school changing its AI policy — we cover only the part that changes what a writer should do, link the primary source, and leave the commentary to people closer to it.