Koncile detects prescriptions, doctor's notes and medical certificates edited in the PDF or the image, and those generated or edited by AI, then returns a fraud score from 0 to 100. Prescriber, dosage and dates are checked too, handwritten included.
Create a free account to run the analysis.
Create a free accountFor document volumes or an API, talk to an expert.
The document checker used by insurers, banks and large companies
4.8 / 5
Yes, and the checks are the same on both sides of the Atlantic: a prescriber who exists and can be reached, the mandatory details, a drug, dose and quantity that make sense, and extra care on controlled drugs. What pharmacies rarely check is the file: the software that made the PDF, editing traces, a retouched image.
Prescription, doctor's note or medical certificate, printed or handwritten: the checker runs four levels of checks. An altered quantity or a changed date leaves traces, in the pixels or in the computer file.
Prescriber, registration number, drugs, strengths and directions are extracted, with 94.7% accuracy published on handwritten prescriptions. When the doctor crosses out a printed value and writes the right one next to it, we keep the correction.
A test prescription rewritten with iLovePDF days after it was created scored 80 out of 100. Software, dates and hidden versions of the PDF tell what the screen does not show: a quantity or a date changed after the fact.
On a photo or a scan, K-Detect, our image analysis model, isolates figures written over, text laid over the original and documents generated in whole or in part by AI.
The registration number is read and its format checked, the dose and duration compared with your plain language rules, the dates matched with the rest of the file.
The three documents come from the same clinic but fool different people: the pharmacist, the employer, the insurer or the school. The traces they leave are alike.
A quantity written over, a longer duration, a drug added under the last line, or a whole prescription built to obtain a drug that gets diverted. On the payer side, the same document backs a claim for medication or care that was never dispensed.
A sick note made up or with extended dates, sent to an employer as a photo or a PDF. In the UK, a fit note can be issued by a doctor, nurse, pharmacist, occupational therapist or physiotherapist since July 2022: check who signed it and whether that professional is registered.
For an insurer, a school, a sports club or a travel claim: a certificate of convenience, or a real one with a new date and name. As with prescriptions, the file and the image keep the traces of the edit.
The registers and the rules differ, the logic is the same: who prescribed, what, and is the document the one that was issued.
Most prescriptions in England travel through the NHS Electronic Prescription Service; paper FP10 forms and private prescriptions still circulate. Prescribers can be checked on the GMC, NMC or GPhC registers. Schedule 2 and 3 controlled drugs need the total quantity written in words and figures.
Every prescriber has a 10 digit NPI listed in the public NPPES registry, and controlled substances also require a DEA registration. State prescription drug monitoring programs show a patient's controlled substance history, and electronic prescribing is the norm for controlled drugs.
Home care referrals, private clinics, insurance and benefit claims, employer files and telehealth uploads still carry prescriptions as scans, photos and PDFs. That is where a document check catches what the electronic circuit never sees.
Before checking, you have to read, and a prescription is often handwritten. Each document then gets a fraud score from 0 to 100 with the detail test by test.
A model trained on prescriptions reads the prescriber, registration number, drugs, strengths and directions, handwritten or printed: 94.7% accuracy published on handwritten prescriptions, and 96.4% end to end on a medical lab project in July 2026.
The prescriber's name, address and registration number are extracted and the number's format checked, ready to match against the GMC register or the NPI registry. A prescription with no identifiable prescriber comes out as an exception, with the missing field named.
Your rules are written in plain English: "the dose must stay within the reference range", "the drug must fit the declared condition". They run on every prescription, without development.
When a PDF is edited, it often keeps its original version. Koncile rebuilds it: the earlier quantity or date reappears, before and after side by side.
Native PDF, scan, photo or hybrid file. Typed text laid over the scan of a real prescription, a drug added in another font: K-Detect, our image analysis model, also flags documents generated by AI.
Patient name and date of birth are matched with the ID and the other documents in the same file, so a prescription in one name does not support a claim in another.
Koncile's API returns the prescription fields, a fraud score from 0 to 100 and detailed explanations of manipulation risks, right inside your dispensing or practice software.
Prescriber, registration number, patient, drugs, strengths and directions, handwriting included.
High, medium or low level, with a one sentence explanation.
High risk from 80.
A one line summary ready for your information system.
curl -X POST https://api.koncile.ai/v1/upload_file \ -H "Authorization: Bearer $API_KEY" \ -F "files=@prescription.pdf"requests.post("https://api.koncile.ai/v1/upload_file/", headers={"Authorization": f"Bearer {API_KEY}"}, files=[("files", open("prescription.pdf", "rb"))]){ "status": "DONE", "General_fields": { "Prescriber_NPI": {"value": "1000000000", "confidence_score": 0.97}, "Medication": {"value": "Example drug 500 mg", "confidence_score": 0.95} }, "Fraud_detection": { "risk_score": 80, "summary": "PDF rewritten with iLovePDF after creation; dosage outside reference", "signals": [ {"name": "Editing software", "severity": "high"}, {"name": "Dosage consistency", "severity": "medium"} ] }}Wherever a prescription triggers a delivery or a payment, it is read and checked before it reaches your systems.
HealthNet Homecare, a UK home care provider, has its prescriptions read by Koncile with a success rate close to 100%, including an exam code not present on the document. The lesson: when the doctor crosses out a printed value, the correction is what must be kept.
Uploaded prescriptions are read and checked before dispensing: prescriber details, dose, file and image. An incomplete prescription comes out as an exception, not a guess.
Prescriptions, invoices and statements arrive in reimbursement claims. Each document is checked on receipt and matched with the member's ID in the same file.
Doctor's notes, fit notes and medical certificates received as photos or PDFs are checked before they reach payroll, with a score the HR team can explain.
Without a tool, these three checks stop clumsy fakes. They work for doctor's notes and medical certificates too.
Prescriber, registration number, date, signature, patient, drug, strength, dose, quantity. On a controlled drug, the quantity in words must match the figures.
Search the GMC, NMC or GPhC register in the UK, the NPPES registry in the US. If in doubt, call the practice on a number found elsewhere than on the document.
With our free PDF metadata viewer, look at Creator, Producer and the dates. A prescription comes out of clinical software. An online PDF editor, or a file modified days after its date, are red flags.
Verifying a prescription is not judging a treatment. We say so clearly.
Koncile checks that the prescriber can be identified and that the prescription fits your rules. It does not say whether the treatment is right: that is the pharmacist's and the doctor's job.
It has no original metadata. The check then relies on the image, the reading of the handwriting and the consistency of the content.
The NHS Electronic Prescription Service and US e-prescribing run their own secure circuit. The checker is for prescriptions that arrive on paper, as photos or as PDFs.
A neatly altered quantity or a generated prescription often leaves no visible error. The checker relies on signals invisible to the naked eye.
| Element of the prescription | What Koncile checks | What gives an edit away |
|---|---|---|
| Issuing software | Creator and Producer compared with clinical software | iLovePDF, Word, Canva or an online PDF editor |
| File dates | Creation and modification, compared with the prescription date | A file modified days after its date |
| Saved versions | Versions appended at the end of the file, dormant objects | The earlier quantity or date stays in the file |
| Handwriting | Reading of the handwriting, prescriber's corrections | Handwriting that changes on a single line, a figure written over |
| Prescriber | Name, address, registration number format | A missing or malformed number, a name that does not match |
| Prescription | Drug, strength, dose and duration against your rules | A dose outside the reference range, an unusual duration |
| Image | Edited areas and AI generated parts, with K-Detect | Text laid over the scan, a drug added in another font |
A prescription tells what a person is treated for. Founded by a former lawyer, Koncile was built with data protection at the core of its architecture: third party SOC 2 audit, GDPR compliance, secure hosting. Documents are encrypted, deleted after analysis and never used to train a model.
Yes. Pharmacists check the prescriber, the mandatory details and whether the drug, dose and quantity make sense, with extra rules for controlled drugs. In the US they can also check state monitoring programs. What they rarely check is the file itself, which is where an edited PDF or a generated image shows.
You measure a risk. It becomes high when the prescriber cannot be found, a detail is missing, the dose is outside the reference range, or the file shows editing: an online PDF editor, a late modification, an earlier version hidden in the PDF, text laid over the image. The checker at the top of this page runs these checks in about ten seconds.
Check that the clinician exists and is registered (GMC, NMC or GPhC in the UK, NPPES in the US), that the dates fit, and call the practice on a number found independently. On a photo or a PDF, look for retouched dates and for the software that produced the file.
The pharmacy label on the medication, the dispensing receipt or a printout from the pharmacy usually serve as proof. In England, the NHS App lists repeat prescriptions. A copy of the original prescription is the strongest proof, and the one most worth checking.
In the UK, presenting a forged prescription can be fraud by false representation under the Fraud Act 2006, punishable by up to 10 years in prison. In the US, obtaining a controlled substance with a forged prescription is a federal offence under 21 U.S.C. § 843(a)(3), and states have their own laws.
Yes. Koncile reads handwritten prescriptions with 94.7% published accuracy, prescriber's corrections included, then runs the consistency checks. On a photo, metadata is missing: the image and the content carry the check.
It can flag an odd dose. It lacks reliable reading of handwriting, models trained on edited or generated images, and knowledge of the traces each editing tool leaves. A language model is only one part of the machine.
Yes. Koncile complies with SOC 2 and GDPR, and an independent firm certifies our security and data protection. Prescriptions are encrypted, deleted after analysis and never used to train a model.
Each document gets its own method: a proprietary ML model for photos, known metadata and forensics for native PDFs, and contextual cross checks.
Inflated balances, deleted lines, a PDF rewritten with software other than the bank's.
See the detection
Net pay changed, contributions that do not match gross pay, a layout that does not match the payroll software.
See the detection
Photo replaced, date of birth or number changed, an MRZ that no longer matches the fields.
See the detection
Edited amounts, wrong VAT, duplicates and made up suppliers.
See the detectionUpload your bank statements, pay stubs or IDs: Koncile extracts the data and flags fakes in seconds.
Talk to an expert
