Is Creyos Accurate? What the Research Says

“Accurate” hides several questions: does Creyos give similar results twice, does it measure what it claims, are its comparison norms sound, and can it tell people with a condition from people without? (For what Creyos is and how a visit works, see the Creyos guide.) The honest answer: it is a reasonably well-grounded measure of thinking skills, with moderate reliability on several tasks, decent evidence that online results track in-person testing in adults, and weak evidence for telling diagnoses apart. It is not, and does not claim to be, a diagnostic test.

The verdict in one box

Where the tasks come from

The tasks grew out of Adrian Owen’s research group at Cambridge and later Western University. Two papers frame the lineage. In 2010, Owen and colleagues ran a six-week online brain-training study with 11,430 people and found no evidence that training transferred to untrained tasks, which is also why Creyos is a measurement tool rather than a training program (see does brain training work?). In 2012, Hampshire and colleagues analyzed more than 44,000 people on the online tasks and found that performance reflected at least three distinct components (short-term memory, reasoning and verbal ability) rather than a single “IQ.”

Most individual tasks are modern versions of tests with long track records: digit span, the Corsi block test, the Stroop effect, the Tower of London. That gives them face validity and decades of related research. It does not by itself validate the online versions or their use in any particular clinical group. Our task-by-task explainer covers each one.

The norms: big, but not a clinical sample

Creyos says its normative database now includes more than 85,000 people. That is large compared with many paper tests. Three caveats matter more than size:

• Self-selected volunteers. The sample was recruited online. Among adults, about 41% were in North America, 40% in Europe and 10% in Oceania.
• No health screening. Participants were not asked about or excluded for health conditions, so the comparison group includes people with ADHD, depression and other conditions in unknown proportions.
• Thinner data at the edges. The original Neuron study covered ages 12 to 70; norms were later extended to age 6. Creyos notes that demographics for children under 12 are not characterized, and norms for the SART (6,395 people) are not split by age or gender.

None of this makes the norms useless. It means “average” is average for a broad, internet-using population, and that comparisons for younger children and the oldest adults rest on less information.

Reliability: moderate for several tasks

Reliability asks whether you would get a similar score if you took the task again with nothing changed. Creyos’s own data, comparing first and second self-chosen plays on its old public website, show correlations from 0.45 (Paired Associates) to 0.92 (Double Trouble), averaging 0.69. Feature Match and Number Ladder (formerly Monkey Ladder) sit at 0.57. A task with a correlation near 0.5 can rank the same person quite differently on two days.

The best independent check is Kochan and colleagues (2022), an Australian university team: 52 adults aged 55 to 75 took Cambridge Brain Sciences and Cogstate tasks at home three times. Individual measures ranged from 0.20 to 0.83 (intraclass correlation), but combined global scores exceeded 0.8. The practical reading: patterns across several tasks deserve more weight than any one score. Retesting with Creyos shows the full task-by-task table and what it means for repeat sessions.

Does it measure what it claims? Validity evidence

The strongest support is in healthy and aging adults. In the same Kochan study, a combination of the online measures correlated strongly with a full in-person paper-and-pencil battery (canonical correlation R = 0.87). Thienel and colleagues (2024), with no declared conflicts of interest, studied 141 mid-life and older adults and found the online Creyos battery was about as strongly associated with brain structure on MRI as a comprehensive in-person assessment.

Creyos also says it has not seen systematic differences between online and in-person testing, citing a 2019 paper by Sternin, Burns and Owen. That paper is a review of computerized testing in aging, not a head-to-head comparison, and one of its authors is Creyos’s chief scientific officer.

Clinical groups: where the evidence is thin

For ADHD, Creyos’s own science guide is candid: its ADHD protocol “is not a standalone diagnostic tool,” provides no cutoffs, and so “sensitivity and specificity cannot be calculated.” People with ADHD can score in the typical range on most markers. For depression, we found no Creyos-specific peer-reviewed validation; a validation study in older adults with depression or mild cognitive impairment (NCT06780917) is recruiting. Comparisons with other tools are in Creyos versus other ADHD tests and Creyos versus neuropsychological testing.

Independent versus company-affiliated research

Creyos says its tasks appear in more than 400 peer-reviewed studies (some of its documents say over 300). A PubMed search on 19 October 2026 found 30 abstracts naming “Cambridge Brain Sciences” or “Creyos”; many more papers mention the platform only in their methods. Both numbers are real; they count different things.

Affiliation matters when reading the literature. A 2026 conflict-of-interest statement lists Adrian Owen as a Creyos shareholder and two frequent co-authors, Conor Wild and Michael Battista, as Creyos employees. That doesn’t make their work wrong; much of it is careful and large. It does mean the independent studies above (Kochan, Thienel, Hennessy, Streicher) carry extra weight, and they paint a consistent picture: useful in adults as a measure, modest as a classifier.

What Creyos has not been shown to do

• Diagnose any condition. The federal rule for this device type states that such aids do not identify the presence or absence of a diagnosis.
• Track medication response. No peer-reviewed study shows Creyos-guided medication adjustment.
• Tell ADHD subtypes apart from cognitive data.
• Replace a full neuropsychological evaluation, or meet school, disability or legal testing requirements.
• Improve cognition. It is a test, not a training program.

Creyos is registered and listed with the FDA as a Class II, 510(k)-exempt computerized cognitive assessment aid; it is not FDA-cleared or approved and does not diagnose any condition. The exemption itself applies only to aids not used to diagnose specific diseases or conditions.

How we use it, given this evidence

At Willow & Stone, Dr. Stacey Forbes, DNP, APRN, PMHNP-BC, uses Creyos as one input inside a clinical evaluation, alongside your history, questionnaires and, when relevant, labs. A diagnosis comes from that whole picture. If results suggest something outside our scope, such as a need for neuropsychological testing, we say so and refer out; see psychiatric versus psychological evaluation and screening versus diagnosis. For adult ADHD questions specifically, the full process is on our adult ADHD evaluation page.

Which protocol you would take, whether at home or in the Arlington office, and the fee are confirmed when you book; ask our team, or request a consultation.

Common Questions

Sources & Further Reading

• Hampshire et al. 2012, Neuron — Fractionating human intelligence (PMID 23259956)
• Owen et al. 2010, Nature — Putting brain training to the test (PMID 20407435)
• Creyos — The Creyos normative database (blog, Feb 2026)
• Creyos — Understanding the Creyos Health Normative Database (PDF)
• Creyos — Science Overview (PDF), test-retest table
• Kochan et al. 2022, J Alzheimers Dis — remote unsupervised reliability (PMID 36314208)
• Thienel et al. 2024, Brain Imaging Behav (PMID 39243354)
• Hennessy et al. 2024, Clin Child Psychol Psychiatry (PMID 38282296)
• Al-Saoud et al. 2025, Child Neuropsychol (PMID 38863216)
• Streicher et al. 2026, Neurotrauma Rep — concussion test agreement (PMID 42539374)
• Stafford et al. 2020, J Neurol — concussion history and task performance (PMID 32172379)
• Wild et al. 2022, Cell Rep Med — cognitive sequelae of COVID-19 (PMID 36103880)
• Urian et al. 2026, J Alzheimers Dis — two-task screener (PMID 41848390)
• Creyos — ADHD Report Science Guide (PDF)
• ClinicalTrials.gov — NCT06780917 (Creyos-MD validation study)
• FDA — Product classification PTY, computerized cognitive assessment aid, exempt

This article is general education, not a diagnosis or advice for your situation. Creyos results support, and never replace, a clinical evaluation; only a clinician who has evaluated you can diagnose a condition or recommend treatment. Do not start, stop, or change a medication without talking to the clinician who prescribes it.