amige. vs
Winston AI.
| amige. | Winston AI | |
|---|---|---|
| founded | 2026 | 2022, by Thierry Lavergne (Montréal, bootstrapped) |
| modalities | image · video · text | text + images / deepfakes (no video, no audio) |
| free tier | 3 scans / day | 2,000 credits over 14 days, then paid |
| entry paid plan | $9 / month (Pass) | $10 / month (Essential, annual) |
| API | no public API (yet) | yes, credit-metered, sales-led |
| self-reported accuracy | no single headline number | 99.98% |
| plagiarism + OCR | no | yes, bundled at the credit level |
| model attribution | yes | no |
| per-detector breakdown | yes (panel of classifiers shown per scan) | no (single verdict, 99.98% claim) |
| LMS integration | none | Canvas, Google Classroom, Moodle, Brightspace |
looking for a Winston AI alternative?
Winston AI is the bootstrap-stage Canadian text-detection product positioned as the GPTZero alternative for schools, with a 99.98% accuracy claim on the homepage. founded 2022, ~$121K disclosed funding, small team, profitable as a lifestyle SaaS rather than a venture trajectory. the wedge is clean: tighter LMS integration than GPTZero, plagiarism bundled at the credit level.
amige. reads image, video, and text for consumer one-off checks, routing each scan to the detectors strongest for it and naming the likely maker. Winston covers text and images for classroom submission workflows, with LMS integration and plagiarism bundled. the use cases don't overlap much. different jobs.
where Winston is stronger
LMS integration for educators. Canvas, Google Classroom, Moodle, Brightspace, plus a Chrome extension for teachers grading inline and a WordPress plugin for publishers. amige. has none of this; we're consumer-first.
plagiarism bundled at the credit level. 2 credits per word for plagiarism + 1 credit per word for AI detection on the same submission. amige. doesn't do plagiarism.
OCR + multilingual support. 14+ languages including Simplified Chinese, plus PDF/DOCX with OCR. handles the practical reality of submitted student work better than amige.'s English-focused text classifier.
HUMN-1 site certification. a badge website owners can embed to signal “Winston scanned this site, verified human-authored.” useful for publishers worried about Google's content-authenticity signals. amige. doesn't ship a site-certification surface.
Winston says “99.98%”. independent reviewers say it over-flags non-native English writing.
where amige. is stronger
image and video. Winston detects AI images and deepfakes (Midjourney, Stable Diffusion, ChatGPT, Nano Banana) but doesn't do video at all. amige.'s panel treats image, video (including deepfakes), and text as equal primary surfaces.
model attribution. Winston reports “AI” or “human” with a single percentage. amige. reports “looks like Midjourney v6” or “looks like GPT-4o” with per-model confidence. for explaining a verdict to a skeptic, the difference is large.
per-detector transparency. Winston's 99.98% is a single black-box number. amige. ships the per-detector breakdown next to every verdict so you can see when classifiers agree and when they don't. a trained model routes each scan to the detectors strongest for it, the panel's reads are fused and calibrated, and the verdict abstains rather than guess when the signals conflict. see how it routes, fuses, and names the maker end to end in the machine.
consumer-grade pricing and free tier. Winston free is 14 days then paid. amige.'s 3 scans / day is permanent. for a curious user with one question, the friction delta matters.
accuracy claims
Winston markets 99.98%. independent reviewers report it over-flags non-native English submissions, consistent with the category-wide structural bias documented in the 2023 Stanford / Cell Patterns study (61-97% TOEFL essay false positives on legacy detectors). Winston shares the broader detector failure mode where formal academic prose gets misread as machine-written.
we're not aware of public lawsuits against Winston, and the 99.98% headline is a self-reported number with no independent benchmark to back it. the broader category, including Turnitin and GPTZero, sits inside an active public conversation about wrongful student accusations; structural false-positive risk on ESL and formal-prose writing is documented across detector vendors and applies here too.
amige. doesn't headline a single accuracy number on purpose. most detectors in the category claim 99% and measure 65-90% in independent tests. we'd rather show our panel arguing than one confident percentage that overpromises.
it reads handwriting
Winston OCRs scanned and handwritten work, so a teacher can photograph a physical paper and check it, and it bundles a plagiarism checker. amige. scans digital files and doesn't do plagiarism or handwriting; its edge is image and video, with model attribution.
who each is for
pick Winston AI if you teach K-12 or higher ed, run admissions, or moderate a writing program. you batch documents. you want LMS integration (Canvas / Classroom / Moodle / Brightspace). you need plagiarism + AI in one product. you care about OCR for scanned submissions.
pick amige. if you want to check whether an image is AI in twenty seconds. you want to know which model produced a deepfake video. you want the verdict broken down by detector and attributed to a specific model. you want a $9 plan with a free tier and a shareable permalink.
for the classroom essay case, Winston is the right shape (with the false-positive caveats). for the social-media is-this-real case, Winston isn't in the game.
questions
Does Winston AI detect AI images and deepfakes?
Yes, Winston detects AI-generated images and deepfakes from tools like Midjourney, DALL-E, and ChatGPT Image, but it doesn’t cover video. amige. handles image, video, and text, and names the likely generator family rather than returning a single AI-or-human percentage.
Is amige. or Winston AI more accurate?
Winston markets a 99.98% figure, which is a self-reported number rather than an independent benchmark, and independent reviewers report it over-flags formal and non-native English writing. amige. doesn’t publish one number; it routes each scan to the detectors strongest for it, runs a panel of independent detectors per scan, and flags an uncertain middle band instead of forcing a binary call.
amige. vs Winston AI for teachers?
Winston is the stronger classroom fit, with Canvas, Google Classroom, Moodle, and Brightspace integrations, bundled plagiarism, and OCR for scanned submissions. amige. ships none of that and targets one-off checks rather than grading workflows.
Is Winston AI free?
Winston offers a 14-day trial of 2,000 credits, then it’s paid, with entry plans around $10 to $12 per month. amige. gives 3 free scans per day with no time limit and a $9 paid plan.
sources.
- 01
- 02Winston AI — Crunchbase profileFounded 2022 by Thierry Lavergne; ~$121K disclosed funding (bootstrap).
- 03AI Busted — Winston AI detector review 2026Independent review documenting Winston over-flagging formal and non-native English writing.
- 04Stanford TOEFL bias study (Liang et al., Cell Patterns 2023, summary)Underlying ESL false-positive problem affecting Winston and the category.