Home › AI Detector for France

AI Detector for France, built for Sorbonne, HEC, and Sciences Po.

Detecteur d'IA pour les redacteurs francais. Pre-scan your English memoire, thesis, or article before Compilatio or Turnitin sees it. Calibrated for the international academic register taught at HEC, INSEAD, Polytechnique, and the Sciences Po English-medium tracks, GDPR-compliant for the CNIL-aware French procurement bar, and honest about French-language scope. Free to try. No card. Your first scan in about six seconds.

Start free, no card See pricing
3 scans/day free No signup required Sentence-level highlights
France in 2026

Why AI detection became urgent in French higher education.

French academic culture is famously strict on plagiarism and undisclosed AI use. The Grandes Ecoles, the Universites, the Ministere de l'Enseignement superieur et de la Recherche, and HCERES form a coordinated spine that every accredited institution now points at in its academic integrity handbook.

France has one of the strictest academic cultures in Europe. The Grandes Ecoles run assessment at a level of formality with no real equivalent in most countries, and the older Universites sit close behind. Undisclosed copying has always carried serious consequences inside that system, and undisclosed generative AI is now treated through the same lens. Universites de France and HCERES coordinate quality expectations across the sector, and the Ministere de l'Enseignement superieur sets the policy frame.

1. The Grandes Ecoles tightened first

Sorbonne Universite, Universite PSL (Paris Sciences et Lettres, which includes ENS Paris, Mines ParisTech, and Dauphine), Sciences Po, HEC Paris, ESSEC, ESCP, EM Lyon, Ecole Polytechnique (X), Centrale, and INSEAD have all published AI-use guidance across 2024 and 2025. Specific rules vary by department, but the working assumption across most coursework is that any submission can be scanned. Sciences Po in particular tends to require pure-author drafts for graded written work. Submitting AI-assisted work as fully original is treated as the same category of offence as undisclosed copying.

2. Compilatio and Turnitin run by default through the LMS

Compilatio is the French-built check used widely across French universities, and Turnitin sits alongside it in the international and Grandes Ecoles cohorts. The Moodle and Canvas integrations across the sector mean a French student rarely submits coursework that has not passed through one of these AI checks. The student does not see the institutional AI report; the examiner does. Pre-scanning is the only way to see what the examiner will see before they see it.

3. GDPR, CNIL, and the EU AI Act raise the data-handling bar

France applies the GDPR alongside the Loi Informatique et Libertes of 1978, with the CNIL as one of the strongest enforcement authorities in Europe. The EU AI Act is phasing in through 2025 and 2026 with France as an active voice in shaping it. A detector that ships without a clear GDPR posture does not clear procurement at Sorbonne, PSL, or the Grandes Ecoles. TextSight is GDPR-compliant on retention, export, and deletion, and the CNIL-aware bar is a feature for French users rather than friction.

Local context

The French institutional landscape.

Who runs AI detection across the Universites, the Grandes Ecoles, and the international business schools, what the policy looks like in practice, and where the freelance and French Tech pressure is coming from.

The Universites and the regulatory frame

Sorbonne Universite, Sorbonne Nouvelle, Universite Paris-Saclay (now Paris-Saclay group), Universite PSL, Universite Paris Cite, Aix-Marseille Universite, the Lyon group (Lyon 1, Lyon 2, Lyon 3), Toulouse, Bordeaux, Strasbourg, Lille, Grenoble Alpes, Nantes, and Montpellier form the research-intensive Universite spine. Education runs Baccalaureat at high-school end, Licence over three years, Master over two years, and Doctorat thereafter. Mid-degree work is the Memoire (master's thesis), end-degree the These (PhD). HCERES handles evaluation, the CNRS coordinates national research, and the Ministere de l'Enseignement superieur et de la Recherche sets the policy frame. Compilatio and Turnitin AI coverage are essentially universal across the group by 2026.

The Grandes Ecoles and the international tracks

Ecole Polytechnique (X), HEC Paris, ESSEC, ESCP, EM Lyon, INSEAD (Fontainebleau), Centrale, ENS Lyon, ENS Cachan, Sciences Po, and Mines ParisTech run MBA, MSc, and executive programmes with heavy international intake. Smurfit-equivalent international relevance sits at INSEAD and HEC, and English-medium assessment dominates the international tracks. Academic integrity expectations there are calibrated to professional-school norms rather than typical undergraduate ones. ESL false-positive risk is real in those cohorts and a calibrated detector matters.

The Memoire and the These

French degree work commonly culminates in a substantial written artifact: the Memoire at master's level and the These at PhD level. A Compilatio or Turnitin AI flag on a Memoire chapter is not just a grade question; it is a jury-trust question, and French departments are small enough that reputation carries inside a discipline. A pre-scan before chapter handover has become standard practice across French Master 2 and Doctorat cohorts working in English-medium tracks.

The Paris tech belt and the French content economy

French Tech has pulled hundreds of startups into Paris, Lyon, and Toulouse over the last decade. The Paris tech belt now spans Doctolib, Backmarket, Mirakl, Spendesk, Qonto, Lydia, Pennylane, Alan, Sorare, Aircall, PayFit, Veepee, and Vestiaire Collective, with luxury houses LVMH, Kering, and Hermes commissioning heavy English copy for international audiences. Aerospace Toulouse (Airbus, Thales, Safran) and banking (BNP Paribas, Societe Generale, Credit Agricole, AXA) round out the high-rate professional market. Media writing at Le Monde, Le Figaro, Liberation, Les Echos, L'Express, and L'Obs sits alongside the freelance market. Upwork and Fiverr both added AI-content review to dispute resolution in 2025, and most in-house French content teams now run detection on incoming deliverables.

Plans & pricing

Pick the plan that fits your work.

Use the free tier today, no email needed. Paid tiers billed in USD; French cards from BNP Paribas, Societe Generale, Credit Agricole, La Banque Postale, Boursorama, Hello Bank, Credit Mutuel, plus N26 France, Revolut, and Wise all handle the charge with the standard one to two percent FX margin. Full details on the pricing page.

Free
$0/forever

 

Try the detector. No card, no email.
  • 3 scans / day
  • 5,000 chars per scan
  • Sentence-level highlights
  • Plagiarism Risk indicator
Start free
Starter
$7.49/month

Billed $89.88/year — Save $30

For students & light writers. $7.49/mo ~€6.95 today.
  • 20 scans / day
  • 20,000 AI rewriter words/mo
  • Chrome extension
  • Email support
Get Starter
Business
$29.99/month

Billed $359.88/year — Save $120

For agencies & small teams. $29.99/mo ~€27.95 today.
  • 100,000 AI rewriter words/mo
  • REST API access
  • 5 team seats
  • White-label PDFs
Get Business

Yearly billing saves 25%. France charges 20% VAT on consumer digital services; TextSight invoices from outside France in USD, so French VAT is not collected at checkout on B2C tiers. B2B customers with an EU VAT-ID see reverse charge on the invoice. View full pricing →

For French students

The French student workflow across the Memoire and These cycle.

French degree work culminates in the Memoire at master's level and the These at PhD level, with Licence coursework feeding in. Most French students working in English-medium tracks settle into one of three TextSight patterns by their second semester.

Pattern 1: Pre-LMS submission scan

Paste the essay or coursework into TextSight thirty minutes before the Moodle or Canvas deadline. Read the Authenticity Score and the sentence-level colour map. If the score is below 70, rewrite the red sentences and re-scan. Submit once above 75. This catches both genuine ChatGPT residue and the false-positive flag that highly structured academic English at Sciences Po, HEC, INSEAD, and the Sorbonne international tracks sometimes triggers, particularly when the student has been taught to write in a formally precise English register.

Pattern 2: Iterative Memoire and These scanning

The standard French Memoire (master's thesis) runs 60 to 100 pages depending on discipline, and the These (PhD) is a much longer multi-year artifact. The same iterative use applies to PhD chapter drafts before jury or supervisor handover at any of the Universites or Grandes Ecoles. Scan after each major revision, not just at the end. The score should trend up as the draft tightens. If it does not, the issue is structural (paragraph templating, sentence-length flatness, formulaic transitions) rather than wordsmithing, and you can fix it earlier in the cycle.

Pattern 3: Concours and admissions essays

Used most heavily by Concours-prep students and international applicants writing supplementary essays for Sciences Po, HEC, ESSEC, ESCP, INSEAD, or the Universite PSL international tracks. Admissions essays are one-shot artifacts; a false-positive flag has a very different cost than a flag on a routine class assignment. A pre-scan is cheap insurance.

All three patterns work on the free tier for occasional submissions. Students with weekly essays or a Memoire summer usually upgrade to Pro at $14.99 yearly (around €13.95). The Chrome extension on Starter speeds up the workflow for students writing inside Google Docs or Word Online.

For French Tech and freelancers

The French Tech, luxury, and Paris content economy.

Upwork and Fiverr both added AI-content review to dispute resolution in 2025. A client who suspects AI-generated work can request a scan, and a high AI determination can hold or void milestone payment. For French freelancers earning EUR 50 to EUR 120 an hour on technical or luxury content, a single voided EUR 1,500 deliverable is a real loss.

France has several strong content-publishing hubs. Paris is the centre for luxury (LVMH, Kering, Hermes), fashion, hospitality, and B2B SaaS. French Tech now spans Doctolib, Backmarket, Mirakl, Spendesk, Qonto, Lydia, Pennylane, Alan, Sorare, Aircall, PayFit, Veepee, and Vestiaire Collective. Lyon covers industrial B2B and biotech; aerospace Toulouse (Airbus, Thales, Safran) and banking (BNP Paribas, Societe Generale, Credit Agricole, AXA) round out the high-rate professional market. INSEAD in Fontainebleau sits adjacent to all of this with executive-education content needs.

Standard freelance workflow with TextSight

Draft the deliverable normally (using ChatGPT, Claude, or Mistral as an outline tool or first-pass research helper is widespread and not the issue), then scan the final deliverable before sending. Authenticity Score above 75 is the working floor for safety. Score below 70 means rewrite the flagged sentences before sending. The integrated AI rewriter is useful for fixing individual flagged lines without restructuring the whole piece.

What gets published in English

Paris luxury, fashion, and hospitality brands publish English content for global travellers and shoppers. French Tech SaaS companies publish English landing pages, blog posts, case studies, and ad copy aimed at North American and broader European buyers. Google's helpful-content signal weights AI patterns against ranking, so English content from a French publisher needs to read as human-written to perform organically. Media writing at Le Monde, Le Figaro, Liberation, Les Echos, L'Express, and L'Obs sits alongside the freelance market and increasingly runs AI checks before publication.

For French in-house content roles

Senior French copywriters and EU-remote freelancers serving Paris-headquartered platforms treat the pre-submission scan as part of the deliverable, not an optional QA step. Starter at $7.49 yearly (around €6.95) handles the volume for most individual contractors. Business at $29.99 yearly (around €27.95) covers small agency teams with bulk upload and five seats.

Local AI ecosystem

Mistral AI, the EU AI Act, and what TextSight competes on.

France has a real AI ecosystem with a serious model lab at the centre of it. We want to be plain about how TextSight sits next to Mistral and how we read the EU AI Act, rather than pretending the competitive landscape is purely American.

Mistral AI and the French model scene

Mistral AI is a serious Paris-based model lab and one of the strongest open-weight model families in the world. TextSight is calibrated against the major English-output generators including ChatGPT and Claude. Mistral output in English shares enough surface patterns with those models that the classifier picks up similar signals, but we do not publish a Mistral-specific accuracy number because we have not run a controlled benchmark at the scale we have for the larger models. For Mistral output in French, the lower-accuracy caveat that applies to French content generally also applies here.

What we do not compete on

Mistral builds frontier models. TextSight is a detection and authenticity tool. We are not in the same category. If you are a French developer evaluating which model to ship with, that is a Mistral question, not a TextSight question. If you are a French student or writer asking whether your output reads as machine-generated, that is the question we answer. Some French detection tools have a stronger French-language posture than we do. We respect that and have said so above. Our calibration is English-first and we are not going to oversell French accuracy because the question came from a French page.

EU AI Act and what it means for a SaaS detector

The EU AI Act is rolling out in phases through 2025 and 2026, and France was an active voice in shaping it. The Act covers AI systems in scope across the EU, including transparency obligations around generated content. We treat that backdrop seriously without making blanket compliance claims we cannot verify on every clause. TextSight processes text you submit to return an Authenticity Score and sentence-level signals, we do not sell user content, and we do not use customer text to train classifiers without explicit opt-in. Free-tier scans without signup are not tied to a user account.

Cloud processing reality

Any cloud AI tool, ours included, processes text off your machine. For confidential Memoire material or NDA-bound client work, follow your school or employer policy on external upload. If your department forbids third-party upload, that policy applies to every cloud tool in the market. The honest framing is that we operate as a third-party SaaS in the same category as most international AI tools used in France today, we behave conservatively with the content we receive, and we will not claim full EU AI Act compliance until we have the documentation to back it up.

Honest framing

English vs French content, and where TextSight fits.

TextSight is built and tuned for English. We say this directly because France is a bilingual writing market and we do not want students arriving with a French-only Memoire and getting a worse result than they expected.

French is the official medium of instruction at most Universites, but English dominates at the master's level, MBA tracks, Grandes Ecoles international programmes, and INSEAD by design. Our honest scope is English content only; Francais is out of scope as the primary evaluation language.

English content

This is where the classifier is calibrated. French Tech writers producing English copy for international audiences, agencies writing English landing pages for luxury or hospitality clients, and students submitting English-medium degree work at HEC, INSEAD, ESSEC, ESCP, Sciences Po, Polytechnique international tracks, and the PSL international tracks all sit in the strongest part of the tool.

French content

Accuracy is meaningfully lower on French-only text. The classifier sees the patterns it learned in English training and tries to apply them, but the result is less reliable than it is in English. We do not recommend TextSight as the primary check for French-only Memoires or Theses. For French submissions, Compilatio's French coverage is the better institutional fit where your school provides it.

Bilingual workflows

A common pattern at French business schools is to draft in French, translate to English for an international audience, and then scan the English version. This works well in practice. The English-language output is what readers, clients, jurys, or examiners reviewing English-medium work will see, and that is what TextSight is built to evaluate. We would rather lose the French-only segment of the market than oversell accuracy we have not measured.

FAQ

French users frequently ask.

Do French institutions like Sorbonne, Sciences Po, HEC, or Polytechnique use AI detection?
Major French institutions including Sorbonne Universite, Universite PSL, Paris-Saclay, Sciences Po, HEC Paris, Ecole Polytechnique, ESSEC, ESCP, EM Lyon, INSEAD, Centrale, and ENS have tightened academic-integrity rules around generative AI through 2024 to 2026. Specific tooling varies by school and department, with Compilatio (a French-built check used widely across French universities) and Turnitin being the most common institutional tools. French students commonly use TextSight as a personal pre-submission scan to anticipate what those institutional tools will flag on English-medium memoires, dissertations, and theses.
Does TextSight work for French-language content?
TextSight is English-first. Detection accuracy on French-only content is meaningfully lower than on English and we do not recommend it for primary French workflows. For memoires or theses written in French, the institutional Compilatio or Turnitin check covers French where the school subscribes to it. TextSight is most useful for the share of French students and professionals writing in English at HEC, INSEAD, ESSEC, ESCP, Sciences Po, and the international tracks at Sorbonne and PSL.
How does French pricing work if TextSight bills in USD?
TextSight bills in USD, and French cards from BNP Paribas, Societe Generale, Credit Agricole, La Banque Postale, Boursorama, Hello Bank, Credit Mutuel, plus N26 France and Revolut all process the charge with a standard one to two percent FX margin (Revolut and N26 typically pass the interbank rate with no markup, which is the cheapest option at typical EUR-USD rates). The Pro subscription at $14.99 yearly lands around €13.95 on a typical French card; Starter at $7.49 yearly lands around €6.95; Business at $29.99 yearly lands around €27.95. France charges 20 percent VAT on consumer digital services, but because TextSight is invoiced from outside France in USD on the B2C tier we do not collect French VAT at checkout. For B2B customers with an EU VAT-ID, reverse charge applies on the invoice.
Is TextSight GDPR and CNIL-compliant for French users?
Yes. France applies the GDPR alongside the Loi Informatique et Libertes (1978), with the CNIL (Commission Nationale de l'Informatique et des Libertes) as one of the strongest enforcement authorities in Europe. TextSight is GDPR-compliant: scanned text is processed for the detection workflow and not retained for training, users have full export and deletion rights through the dashboard, and account data is portable on request. The CNIL-aware bar is a feature for French users rather than friction, and Sorbonne, PSL, and the Grandes Ecoles scrutinise vendor data handling closely under that backdrop.
What does the EU AI Act mean for using TextSight in France?
The EU AI Act is rolling out in phases through 2025 and 2026, and France was an active voice in shaping it. We do not make blanket EU AI Act compliance claims. What we can say plainly: we process text you submit to return a score and sentence-level signals, we do not sell user content, and we treat third-party processing the way any cloud SaaS does. For confidential institutional or client work, follow your school or employer policy on external upload. If your department forbids third-party upload, that policy applies to every cloud tool in the market.
Does Mistral AI being French change how I should use TextSight?
Mistral AI is a strong Paris-based model lab and one of the most respected open-weight model families in the world. TextSight is calibrated against the major English-output generators including ChatGPT and Claude. Mistral output in English shares enough surface patterns with those models that the classifier picks up similar signals, but we do not publish a Mistral-specific accuracy number because we have not run a controlled benchmark at the scale we have for the larger models. For Mistral output in French, the lower-accuracy caveat that applies to French content generally also applies here.
I write for a Paris luxury house or a French Tech SaaS. Does pre-scanning matter for English content?
Yes. LVMH, Kering, Hermes, and the wider Paris luxury market publish heavy volumes of English content for international audiences. French Tech companies (Doctolib, Backmarket, Mirakl, Spendesk, Qonto, Lydia, Pennylane, Alan, Sorare, Aircall, PayFit, Veepee, Vestiaire Collective) publish English landing pages, blog posts, and ad copy aimed at North American and broader European buyers. Google's helpful-content signal weights AI patterns against ranking, so English content flagged as high-AI tends to underperform organically. A pre-publication scan is a cheap quality gate.
Is TextSight a Compilatio replacement for French faculty?
No. TextSight is built for individuals doing pre-submission self-checks, not for institutional academic integrity workflows. French institutions running formal AI checks should evaluate Compilatio, Turnitin, or Urkund for LMS integration. TextSight fits the student side of the workflow: predicting what the institutional tool will flag before submission to an examiner, jury, or supervisor.
Related

More guides for French users.

Pre-scan your next submission. Ship clean.

Free to try. No card. Your first scan in about six seconds.

Start free, no card See pricing
Calibrated for English-medium French academic register · GDPR and CNIL-aligned · Sentence-level highlights for Compilatio and Turnitin pre-scans