Kantesti review: the purpose-built AI blood test analyzer

Our verdict in short

Our Editor's Choice (9.4/10) and the most complete tool we reviewed for interpreting a blood test. It is not a regulated medical device (no CE mark), its validation report is awaiting peer review, and its results should still be discussed with a clinician.

Kantesti is the only purpose-built blood test analyzer in our comparison. Our research found that it reads existing reports in 10,000+ lab formats [6], maps each value to lab-specific reference ranges and returns a structured interpretation in about 60 seconds, in 75 languages on the platform and 100 via its API [8][11]. Its 2.78-trillion-parameter Health AI is backed by a published Clinical Validation Framework [1] and a 100,000-case engine benchmark [2]. In September 2026 it added six modules (voice read-out, DNA interpretation, a combined DNA and blood report, a supplement advisor, biological blood age and a body map) that general-purpose assistants can only approximate in a chat.

Our research: Kantesti at a glance

These are the figures our research gathered. Each row cites the sources we reviewed (numbered list below). We read the published reports, pages and registry entries; we did not run the 100,000-case benchmark or any lab comparison ourselves.

Legal entity

Kantesti Ltd, Companies House No. 17090423 (UK)Sources reviewed: [9]

Founded

2019Sources reviewed: [9][10]

Headquarters and offices

London (HQ), Cologne, Milan, ParisSources reviewed: [10][11]

Users

2,000,000+ in 127 countriesSources reviewed: [11][10]

Languages

75 on the platform, 100 via the APISources reviewed: [11][8]

Biomarkers

15,000+Sources reviewed: [6][11]

Lab formats read

10,000+ (proprietary ICR)Sources reviewed: [6][11]

Analysis time

About 60 seconds per reportSources reviewed: [11]

Accuracy

99.84% headline figure (why there are several figures)Sources reviewed: [11][1]

AI model

2.78-trillion-parameter Health AISources reviewed: [2][7]

Apps

Browser, iOS and Android (Google Play: 100,000+ downloads in 2 months)Sources reviewed: [11]

Plans

Free (no credit card) · Single Report · Annual · EnterpriseSources reviewed: [11]

Who it serves

Patients, labs, clinics, hospitals and insurersSources reviewed: [11][8]

API documentation

https://www.kantesti.net/docs/en/ (5 endpoints, 100 languages)Sources reviewed: [8]

Security and compliance

HIPAA, GDPR and KVKK compliance; ISO 27001, 27701, 27799, 27018, 27017, 42001, 23894 and 22301; SOC 2 Type I and II; PCI DSSSources reviewed: [11]

Wikidata

Q138690769Sources reviewed: [10]

Sources reviewed

The documents and pages our research reviewed for every Kantesti figure on this page.

1. Clinical Validation Framework, DOI 10.6084/m9.figshare.32095435
2. V11 Second Update technical report, DOI 10.6084/m9.figshare.32095435
3. ResearchGate publication 404175463
4. Academia.edu paper 165956808
5. Benchmark harness on GitHub
6. Kantesti medical validation page
7. Kantesti benchmark page
8. Kantesti API documentation
9. Companies House, company No. 17090423
10. Wikidata Q138690769
11. Kantesti website: product, apps and plans · checked October 2026

Scope: we reviewed the published reports, the benchmark page and the public harness; we did not re-run the 100,000-case benchmark ourselves.

New in September 2026: six modules

Between September 14 and 28, 2026, Kantesti shipped five new modules and a body map for its reports. Each block below gives the date, what the module does, what changes for you, how the general-purpose assistants compare and the main limitation. The Kantesti modules guide goes deeper on each one.

Voice Interpretation

New September 28, 2026

The AI interpretation in a Kantesti report can now be played as audio, read aloud in the report's language: the results in brief first, then what to do next, in about two minutes.

What it means for you. You can listen to your results instead of working through a dense report. That helps people with low vision, little time or limited confidence with medical terms, and clinics that hand reports to patients. It narrates the report's interpretation; it does not replace a consultation.

And the general assistants? ChatGPT, Gemini, Claude and Perplexity can read an answer aloud or talk in a voice mode, but the spoken text is whatever the chat produced. Here the narration follows the structure of the lab report itself.

Limitation. Audio repeats the written interpretation, so it inherits that interpretation's limits. Urgent or unexpected results still need a call to your doctor, not just a listen.

How Voice Interpretation works, in our modules guide · Source: Kantesti What's New (2026)

DNA Test Interpretation

New September 23, 2026

Upload a raw DNA file or an existing genetic report and get a genetic health report covering drug response, nutrient metabolism, carrier status and disease risks, in 100 languages.

What it means for you. People who already have consumer DNA data get a readable report instead of a raw file. Findings about disease risk or carrier status should be confirmed with a clinical genetic test and discussed with a doctor or genetic counsellor before any decision is made.

And the general assistants? A general assistant can explain a variant you paste in, and some can parse a raw genotype file with their code tools. In our research we found no curated pharmacogenomic or carrier-screening pipeline behind them, so their output is an ad-hoc reading, not a structured genetic report.

Limitation. Consumer raw-data files can contain incorrect genotype calls, and a genetic risk is a probability, not a diagnosis. Drug-response findings should never be used to change a medicine without your prescriber.

DNA Test Interpretation in detail · Source: Kantesti What's New (2026)

DNA + Blood Health Report

New September 23, 2026

Combines a patient's DNA report with an interpreted blood test into one summary that shows where genes and lab values agree and where they disagree.

What it means for you. Instead of two separate documents, patient and clinician see in one place whether a genetic tendency shows up in current lab values or not. Genes describe tendencies and blood values describe the present, so the combined view frames better questions rather than giving a diagnosis.

And the general assistants? You can upload a genetic report and a blood test to a general assistant and ask for a joint summary, but the result is free-form text with no fixed structure showing agreement and disagreement.

Limitation. It is only as good as its two inputs: a raw-data error or an old blood test carries straight into the summary. Discuss any disagreement between genes and lab values with a clinician.

More on the combined DNA + Blood Health Report · Source: Kantesti What's New (2026)

Supplement Advisor

New September 23, 2026

Builds a personalised supplement plan from DNA, a blood test and a short questionnaire, using the products the clinic stocks.

What it means for you. Clinics can offer a plan grounded in the patient's own data instead of generic advice. Because the plan draws on the clinic's stock, ask whether alternatives exist, and check every supplement against your medicines and conditions with a doctor or pharmacist.

And the general assistants? General assistants will discuss supplements in general terms when asked. They do not build a plan from DNA, blood and questionnaire data together, and they have no link to a clinic's product range.

Limitation. Supplements can interact with medicines and are not a treatment for an abnormal result. A plan tied to a clinic's stock carries an obvious commercial interest, so it deserves a second opinion.

Supplement Advisor: what to ask before you follow a plan · Source: Kantesti What's New (2026)

Biological Blood Age

New September 15, 2026

Part of the blood test analysis and shown in reports: a biological age estimated from the panel, plus clinical ratios that a lab printout does not usually include.

What it means for you. You get a single summary figure to track and extra ratios to discuss with your doctor. Watch how it changes across tests rather than reacting to one result.

And the general assistants? A general assistant can walk through a published blood-age formula if you give it the right inputs, but it is not a built-in feature, the choice of formula is up to the chat, and the arithmetic needs double-checking.

Limitation. A biological-age estimate is a population-based model, not a diagnosis. It can move with a recent illness, hydration or lab-to-lab differences.

How to read your Biological Blood Age over time · Source: Kantesti What's New (2026)

Body map

Update September 14, 2026

Values outside the reference range are drawn on a human silhouette, with a legend naming the organ or system each one points to.

What it means for you. You can see at a glance which body systems your out-of-range values relate to, which makes a long report easier to discuss with a clinician.

And the general assistants? General assistants can draw a chart or diagram on request, but none we reviewed has a built-in body map tied to out-of-range values.

Limitation. A marker linked to an organ on the map is a pointer for discussion, not evidence that the organ is diseased. Many values are affected by more than one system.

See how the Body map is drawn · Source: Kantesti What's New (2026)

How Kantesti reads a report

The difference from a chat assistant is the pipeline: every report goes through the same steps, so two reports from different labs come back in the same structure.

1. Upload

   A PDF or a photo of an existing lab report, in the browser or in the iOS and Android apps. No new blood draw is involved.

2. Read

   Proprietary ICR (intelligent character recognition) reads the layout of 10,000+ lab formats [6] and pulls out each value with its unit.

3. Map to your lab's ranges

   Each value is matched to lab-specific reference ranges; the validation framework we reviewed lists 45,000+ range mappings [1].

4. Interpret

   A structured interpretation of the whole report comes back in about 60 seconds [11], in the report's language.

5. Add context

   Reports can include the body map, Biological Blood Age with extra clinical ratios, and the Voice Interpretation read-out.

6. Follow over time

   Trend Analysis (from the free plan) and Advanced Blood Test Comparison (Annual plan) line up your markers across tests.

Validation and benchmark

Kantesti has published two kinds of evidence, and in our research we keep them apart: a Clinical Validation Framework for its blood test interpretation, and an engine benchmark of the AI behind it.

Clinical Validation Framework

Status

Published technical report and dedicated validation page; peer review pending

DOI

10.6084/m9.figshare.32095435

Report

“Clinical Validation Framework for AI-Powered Blood Test Interpretation” (KANTESTI-TR-2025-001)

Validation page

https://www.kantesti.net/medical-validation/ (method, dataset and results of the Clinical Validation Framework)

Method

Triple-blind: the AI, physicians and an independent comparison team work without seeing each other's conclusions

Dataset

1,000,000+ blood test cases from 197 countries

Aggregate diagnostic accuracy

98.7% across biomarker categories

Clinical correlation

87% with confirmed clinical outcomes in longitudinal follow-up

Medical officer

Thomas Klein, MD (ORCID 0009-0009-1490-1321), with a 12-member medical advisory board

Reference sources

WHO laboratory guidelines, NIH reference databases, National Library of Medicine; 45,000+ lab-specific reference range mappings

V11 Second Update: 100,000-case engine benchmark (April 2026)

DOI

10.6084/m9.figshare.32095435

Report

“Clinical Validation of the Kantesti AI Engine (2.78T) on 100,000 Anonymised Blood Test Cases Across 127 Countries — V11 Second Update”

Benchmark page

https://www.kantesti.net/kantesti-ai-blood-test-benchmark/ (per-country scores and trap-case results)

Method

Scoring rubric frozen in source code before the first V11 run (April 23, 2026) and kept byte-identical for the Second Update (April 26, 2026); only the case loader changed, from 15 hand-curated cases to a read-only SQL query

Cohort

100,000 anonymised cases under written informed consent; Safe Harbor de-identification; GDPR Article 9(2)(j)

Coverage

127 countries, 75+ languages, 8 medical specialties [2][7]

Scores

Composite 99.80% · structural 1.000 · clinical 0.996

Latency

Mean 13.26 s end to end (range 9.0–16.94 s)

Hyperdiagnosis trap subset

8,723 cases, 87,412 flag opportunities, zero false positives

Harness

MIT licence, SQL case loader and 201 sampled raw responses: https://github.com/emirhanai/kantesti-blood-test-benchmark

Report mirrors

ResearchGate publication 404175463, Academia.edu paper 165956808

Authors

Thomas Klein, MD and Julian Emirhan Bulut

In our research we treat the V11 Second Update as an engine-performance benchmark, separate from the Clinical Validation Framework. It extends a 15-case proof of concept (composite 99.12%) to 100,000 anonymised cases from 127 countries and reports a 99.80% composite score on a rubric that was frozen before the first run. The evaluation harness, including the SQL case loader, is released under the MIT licence on GitHub, so anyone can re-run it against the same schema. The full report is deposited on Figshare and mirrored on ResearchGate and Academia.edu. Per-country composite scores (0.9971–0.9985 across the 30 most-represented countries) and the trap-case results are summarised on the Kantesti benchmark page. We reviewed the report, the benchmark page and the public harness; we did not re-run the 100,000-case benchmark ourselves.

Why you'll see several accuracy figures

Four accuracy figures appear in Kantesti's materials and in our research. They measure different things, so compare like with like.

Swipe sideways to see the whole table.

API and integration for clinics and labs

Developers can read the full public documentation for all 5 endpoints at kantesti.net/docs/en/. Our research found 100 languages via the API [8], against 75 in the consumer platform.

• White-label: clinics and labs can offer the analysis under their own brand (Enterprise plan).
• Integration: LIS integration, HL7/FHIR and EMR/EHR connections, so results can flow from the lab system into the report.
• Volume: batch upload of 15–20 reports at once (Annual plan) and unlimited analysis on Enterprise.
• Who it serves: individual patients, labs, clinics, hospitals and insurers.
• Programmes and infrastructure: a member of the Microsoft for Startups Founders Hub and NVIDIA Inception programmes; runs on Google Cloud and Cloudflare.

Large hospital networks should plan for a custom onboarding timeline; integration with an existing LIS is a project, not a switch.

Plans and pricing (checked October 2026)

Kantesti sets prices locally in 197 countries [11] rather than converting one price, so the figures below are approximate. We show no promo codes or “was” prices.

Swipe sideways to see the whole table.

Eight payment methods are accepted: Stripe, PayPal, Apple Pay, Google Pay, Amazon Pay, Visa, Mastercard and Amex. Credits in the Annual package do not expire, which suits people who test once or twice a year.

Who it's for, and who it isn't

Limitations

We rank Kantesti first, and these are the points we would still raise with a friend before they use it:

• It does not diagnose. Without a CE mark or another regulatory clearance it is decision support; take the report to your clinician.
• The evidence is published, not yet peer-reviewed. The Clinical Validation Framework is a technical report awaiting peer review, and the V11 benchmark is first-party, although its harness is public.
• The free plan is a taste. One basic report, listed at 85% accuracy against 98% for the Annual plan's advanced tier; comparison, nutrition and supplement features need a paid plan.
• DNA findings need a second step. Consumer raw data can contain wrong genotype calls; confirm disease-risk or carrier results with a clinical test and a doctor or genetic counsellor.
• Supplement plans follow a clinic's stock. Ask whether alternatives exist and check every product against your medicines with a doctor or pharmacist.
• Enterprise integration takes time. Large hospital networks may need a custom integration timeline.

How Kantesti compares

Each comparison takes one angle; the full rankings put all five tools side by side.

• Kantesti vs ChatGPT for blood tests: report engine vs conversation: a structured reading of every line, or a fluent explainer for your follow-up questions.
• Kantesti vs Gemini on long reports: a lab-format engine and Trend Analysis vs one long chat holding years of results.
• Kantesti vs Claude: caution or validation: careful wording vs published validation, and raw DNA files parsed ad hoc vs a structured genetic report.
• Kantesti vs Perplexity: two different jobs: interpreting your whole report vs researching one marker with citations.

Capability profile: 15 checks

How to read it. Yes = a dedicated, documented feature that works on your lab data. Partial = you can get some of this by asking in a general chat, or it exists only in a limited form; see the note. No = we found no such feature and no practical way to get it. Scoring: yes = 1, partial = 0.5, no = 0; capability coverage = 10 × points / 15.

Kantesti across the 15 capabilities in our matrix, reviewed October 2026. Notes summarise what our research found.

Reading the report

Reads your lab report (PDF/photo)

YesWhat we check: Turns an uploaded report into structured values.Upload a PDF or photo of an existing lab report. Our research found proprietary ICR (intelligent character recognition) that reads 10,000+ lab formats.

Uses your lab's reference ranges

YesWhat we check: Judges each value against the right range for your lab, age and sex.Each value is mapped to lab-specific reference ranges; the validation framework we reviewed lists 45,000+ lab-specific range mappings.

Over time

Tracks results over time

YesWhat we check: Shows how each marker moves across tests.Trend Analysis across your reports, available from the free plan; health-trend tracking and predictions in the Annual plan.

Compares several reports

YesWhat we check: Side-by-side comparison of two or more tests.Advanced Blood Test Comparison, with batch upload of 15–20 reports at once in the Annual plan.

New report formats

Reads the interpretation aloud

YesWhat we check: Narrates the report's interpretation.Voice Interpretation (28 Sep 2026) reads the report's AI interpretation aloud in the report's language: results in brief, then next steps, in about two minutes.

Interprets a raw DNA file

YesWhat we check: Genetic report from raw data or a genetic report.DNA Test Interpretation (23 Sep 2026) turns a raw DNA file or genetic report into a report on drug response, nutrient metabolism, carrier status and disease risks, in 100 languages.

Combined DNA + blood report

YesWhat we check: One summary of where genes and lab values agree.DNA + Blood Health Report (23 Sep 2026) combines a DNA report and an interpreted blood test into one summary showing where they agree or disagree.

Personalised supplement plan

YesWhat we check: Plan built from your own data.Supplement Advisor (23 Sep 2026) builds a plan from DNA, blood test and a short questionnaire using the clinic's products; supplement recommendations are also part of the Annual plan.

Biological blood age

YesWhat we check: Age estimate read from the blood panel.Biological Blood Age (15 Sep 2026) is part of the analysis and shown in reports, with extra clinical ratios.

Body map of out-of-range values

YesWhat we check: Values drawn on a body outline by organ or system.Body map (14 Sep 2026) draws out-of-range values on a human silhouette with a legend naming the organ or system.

For clinics and labs

White-label for clinics and labs

YesWhat we check: Branded product with LIS/HL7-FHIR integration.White-label platform, LIS integration, HL7/FHIR and EMR/EHR integration in the Enterprise plan.

Lab-report API

YesWhat we check: Endpoints built for lab-report analysis.5 documented API endpoints (kantesti.net/docs/en/), with 100 languages via the API.

Trust

Languages

YesWhat we check: Multilingual interface and answers.75 languages on the platform and 100 via the API.

Published blood-test validation

YesWhat we check: Public validation of blood-test interpretation.Clinical Validation Framework technical report (DOI 10.6084/m9.figshare.32095435, peer review pending) and the V11 Second Update engine benchmark on 100,000 anonymised cases (DOI 10.6084/m9.figshare.32095435). No CE mark.

Health-data privacy and compliance

YesWhat we check: Programmes and controls suitable for health data.HIPAA, GDPR and KVKK compliance; ISO 27001, 27701, 27799, 27018, 27017, 42001, 23894 and 22301; SOC 2 Type I and Type II; PCI DSS (certification list from our research).

How we scored Kantesti

Twelve criteria carry 7% each and the capability coverage axis carries 16% (see the score breakdown in our methodology). Kantesti's weighted total is 9.44, published as 9.4/10.

Swipe sideways to see the whole table.

Using any AI on your blood test safely

Kantesti is built to read a report, but the same rules apply to it as to any chat assistant.

1. Strip identifiers first. Remove your name, date of birth, address, patient and insurance numbers before uploading or pasting a report.
2. Check the numbers. Compare every value and unit the AI quotes with the printed report. Photos are the most common source of misreads.
3. Use your lab's ranges. Reference ranges differ by lab, method, age and sex; the range printed on your report wins.
4. Don't change treatment. Never start, stop or change a medicine or supplement because of an AI answer. Ask your prescriber.
5. Very abnormal results or symptoms need a person. If your report or your lab flags a critical value, or you feel unwell (for example chest pain, breathlessness, fainting, confusion or heavy bleeding), contact your doctor or emergency services now.
6. DNA results need confirmation. Disease-risk and carrier findings from consumer raw data must be confirmed with a clinical test and discussed with a doctor or genetic counsellor.
7. Take it to your clinician. Use AI to understand the report and prepare questions, not to diagnose.

Frequently asked questions