4 to 46
Seizures in the records before the strongest and the weakest model first raised the possibility.
AI Evaluation
In dementia, seizures are common and easily taken for the dementia itself. Seagull shows whether your AI recognizes them, and what it recommends when it doesn’t.
Given the same weeks of records, some models raised the possibility early. Others let dozens of seizures pass, or went along with medicines that can make them worse.
4 to 46
Seizures in the records before the strongest and the weakest model first raised the possibility.
2 of 7
Models that went along with an antipsychotic for a woman whose morning episodes were seizures.
63 of 72
Weeks a simple keyword search raised a false alarm. Speed without judgment isn’t safety.
A daughter writes that her mother “zoned out at breakfast.” A clinic note calls it anxiety. A fall goes unwitnessed. No single note gives it away. The pattern does, and AI is now the one reading those notes.
Seagull began with the founder’s mother. Her episodes of confusion and unresponsiveness were taken for her dementia and not evaluated neurologically. They were later understood as seizures.
Seagull asks whether your AI would read her records the same way.
Read how Seagull beganHow soon it noticed, how often it was wrong and what it recommended, measured against an expert reading of the same records.
Every finding quotes your AI’s own words and the records it relied on, so your team knows where to look.
Independent results on one of the most easily missed risks in dementia care, ready for the question your customers will ask.
Assistants that families and clinicians ask about a person’s changing behavior.
Chart summaries, ambient documentation and risk alerts.
Caregiver chat, visit notes and care-plan tools.
Alerts that combine sensor events with notes about falls, sleep and movement.
Record screening for trials and safety monitoring, where a missed seizure is a missed exclusion or adverse event.
Charting assistants and incident review in memory care and skilled nursing.
Most AI tests score a single answer. Seagull follows a person’s records as they build, the way the pattern really emerges.
Half the people have episodes with another cause. An AI that raises an alarm at every confusing moment doesn’t pass.
Your AI reads Seagull’s synthetic patients, written the way families, aides and nurses actually write.
No. Your AI reads Seagull’s synthetic patients. No real patient data enters the test.
Any AI that reads or writes about a person with dementia: assistants, documentation and summary tools, monitoring alerts and research screening.
No. Your AI passes by raising the possibility and sending it to a clinician. It is never asked to diagnose.
No. A DSEF license puts DSEF into your product. An AI evaluation tests the AI you already have. You can do either, or both.
About two weeks. The free first test takes about a week.
Only with your permission.
No. It shows how your AI handled these cases. It isn’t certification, and the report says so.
Your AI reads the records of four synthetic patients. You get a one-page result and a conversation about what it means for your product.
Seagull evaluates how AI products handle possible seizures in dementia. It does not diagnose seizures, certify products or replace clinical judgment.