Original AI-detection report on your draft, page by page. Tuned for academic English. Structured AI-probability output with a score on every flagged sentence, delivered to your inbox in under ten minutes.
A single number on a cover page tells you almost nothing. The report that arrives in your inbox is built to be read in four passes — number, heatmap, per-sentence score, distribution. In that order.
"24% AI" doesn't mean a fourth of your work was generated. It means 24% of sentences read as AI-like to the model. Your sentence cadence — long, even, hedged — can match that pattern even when no tool was used.
If 24% spreads thinly across the document, it's likely register. If it concentrates in two paragraphs, those two are worth a manual look. Read the highlighted sentences before you panic.
Each flag shows a score, not a tier: a clean sentence stays at 0, anything from 1–20 shows as an asterisk (*) — statistical noise — and only above 20 gets an exact number. A 30% report built from asterisked 1–20 sentences is a different conversation than a 12% report built from above-20 flags.
The cover summary breaks the score down by band — how many sentences are asterisked (1–20) versus scored above 20 — and points to where flags concentrate in the document. If above-20 flags cluster in a couple of paragraphs, those are the ones worth a manual look. If the score is made up mostly of asterisked 1–20 sentences spread thinly, your style is the issue — not synthesis. Don't act on the number alone.
Most AI detectors output a percentage and stop. Ours scores every sentence and shows only the number that matters: a clean sentence stays at 0, anything in the 1–20 range — the statistical noise that produces false positives — is marked with an asterisk (*) instead of an alarming figure, and only scores above 20 are reported as an exact numeric value. A passage that looks human-written but pattern-matches as AI is the false-positive that ruins a viva — collapsing that low-range noise to an asterisk is what lets you tell stylistic register from likely synthetic prose at a glance.
Most detectors are tuned on web text — blog posts, marketing copy. Methods chapters, citation-dense paragraphs and equations trip them constantly. Our pipeline excludes references, equations and direct quotes before scoring, the same way a standard similarity report excludes the bibliography. The number you see is the number that matters.
You've read the report. Here's how the rest works.
Drop your document, pick the page slab, and submit. The slab price shows instantly — takes a minute.
One payment for the AI report at the listed slab price. No quotation wait — report services are priced upfront, so you pay and we begin straight away.
PDF report with cover summary, sentence-level highlights, and a per-sentence score breakdown — sent to the email on file and kept in your My Account section (and the mobile app) to download any time.
Got a question on a flag? Reply to the report email — a senior editor explains how the score and the per-sentence flags were composed, at no extra fee.
If your report comes back above your institution's threshold, we'll point you to AI Draft Editing — manual rewriting by 27 PhD editors that improves academic quality and clarity while keeping voice, meaning, and citations intact. The check itself never auto-converts to editing work. That's your call.
I had used a chatbot once for grammar, panicked, and ran every detector I could find. Each gave a different number. Their report had sentence-level highlights and a score on every sentence. That's the only one my supervisor accepted.
Wrote my methods chapter myself and got 32% on a free detector. Theirs came back at 6% — the score breakdown showed the formal register was the trigger, not synthesis. I needed that sanity check.
For a Springer submission I needed an independent AI report on the side. Twenty-minute turnaround, structured cover summary that the desk editor accepted at face value. Worth the fee.
Submitted at 11 PM, report at 11:14 with sentence-level highlighting. The two flagged paragraphs were exactly the ones I knew were ChatGPT. Used AI Draft Editing the next day.
Free detectors gave me numbers between 4% and 78% on the same document. This was the only report with sentence-level reasoning and a score on every sentence. The numbers were defensible.
We send batch AI checks for every M.Phil. cohort now. Two PhD desks rejected our cohort in 2024 over AI use — they haven't since we started this. The structured score breakdowns are the reason.
This is the single most common situation we see. Academic prose — especially methods chapters and dense citation paragraphs — uses long, even, hedged sentences. That cadence overlaps with how language models write, so detectors flag it. The per-sentence scoring on the report is built precisely for this case: if your overall score is high but the flags are scattered thinly across formal-register sentences and most fall in the 1–20 range — shown as an asterisk rather than a number — the signal is style, not synthesis. The sentences that score above 20, especially where they cluster in two or three paragraphs, are the ones worth a manual look.
We deliberately don't name the engine in marketing copy because we tune our reports to academic English with our own pre- and post-processing. The engine is only one input. The structured cover summary, the per-sentence scoring, and the citation/equation exclusion are what make the report defensible to a committee — not which black box generated the raw number.
If your university runs a standard AI detection report on submission day, ours will be visually similar — a cover summary with overall percentage, then page-by-page highlighting with a per-sentence score on every flag. We use this format because viva committees are used to reading it. If your university uses a different system, the substance (sentence-level scores, exclusions, the 0 / asterisk / above-20 breakdown) carries over.
Look at the score breakdown first. If most flags fall in the 1–20 range — shown as an asterisk — and are spread across the document, your style is the issue and a rewrite won't help much — submit as-is or change a handful of sentences yourself. If above-20 flags concentrate in 2–3 paragraphs, those paragraphs need rewriting. You can do that yourself or use our AI Draft Editing service, where PhD editors revise flagged passages for academic quality, clarity, and your own voice. We will not auto-convert your check into editing work; the choice is yours.
Your document is processed for the report and stored on our servers for 30 days for support purposes, then deleted. It is never used to train detection models, never shared with third parties, and never enters any public index. The report stays in your account for as long as you keep it.
AI detection in non-English academic prose is currently unreliable across all detectors — the training data simply isn't there. We accept English documents only for AI reports. For Hindi, Tamil, Bangla and other Indian-language manuscripts, our team can advise on alternative checks.
Tuned for academic English. Sentence-level highlights. A score on every flag. Delivered before you've made tea.
Takes you back to the upload form at the top of this page.