For most of medical history, Alzheimer’s disease could only be diagnosed with complete certainty after death, through autopsy. That is changing. The 2024 diagnostic criteria from the Alzheimer’s Association now define the disease biologically, using biomarkers that can detect Alzheimer’s-specific brain changes in living patients. Still, post-mortem examination of brain tissue remains the only way to directly observe the plaques and tangles that characterize the disease, and about 10% of people diagnosed clinically during life turn out to have a different condition at autopsy.
What Happens at Autopsy
The gold standard for confirming Alzheimer’s is neuropathological examination after death. A pathologist examines brain tissue under a microscope looking for two hallmark features: amyloid plaques (sticky protein clumps between nerve cells) and neurofibrillary tangles (twisted fibers of a protein called tau inside dying neurons). The diagnosis requires a specific burden of both. Under the standard classification system, pathologists score the density and distribution of neuritic plaques using a scale from none to frequent, and they grade tangle spread across brain regions using a staging system that tracks how far tangles have progressed from the memory centers of the brain outward to other areas. An intermediate or high burden of both features confirms the diagnosis.
This is the benchmark against which every living diagnostic tool is measured. When researchers report that a blood test or brain scan is “90% accurate,” they mean it agrees with what autopsy eventually reveals.
How the Disease Is Diagnosed in Living Patients
In practice, diagnosing Alzheimer’s involves layering multiple types of evidence. No single test gives you a yes or no answer on its own. The process typically spans more than one day and may involve your primary care doctor, a neurologist, and sometimes a psychiatrist or neuropsychologist.
The diagnostic workup generally follows a sequence. It starts with a complete medical history, covering your current health, medications, and family history. A physical exam checks for other conditions that could mimic dementia, including heart, liver, kidney, or thyroid problems. A neurological exam looks for signs of stroke, Parkinson’s disease, or other structural brain issues. Blood and urine tests rule out treatable causes of memory problems like vitamin deficiencies, infections, diabetes, or thyroid dysfunction. Brain imaging with CT or MRI can identify tumors, strokes, or fluid buildup.
Only after these other causes are excluded does the evaluation focus specifically on Alzheimer’s.
Cognitive Testing
Standardized cognitive assessments are a routine part of diagnosis, though they measure symptoms rather than the disease itself. The two most commonly used tools are the Mini-Mental State Exam (MMSE) and the Montreal Cognitive Assessment (MoCA). Both test memory, orientation (knowing the date, where you are), language, and basic math through a series of short tasks that take about 10 to 15 minutes.
These tests don’t have sharp, universally agreed-upon cutoff scores. The MoCA developers have suggested a score of 18 (out of 30) as a possible threshold for Alzheimer’s-level impairment, but no standard dementia cutoff has been formally established. The MMSE similarly uses varying thresholds across different clinical settings. What these tests do well is track decline over time. A single score is less informative than the pattern of change across repeated visits, which is why doctors often repeat them at intervals.
Biomarker Tests That Detect Alzheimer’s Directly
The biggest shift in Alzheimer’s diagnosis is the move toward biomarkers: measurable biological signals that reveal whether the disease’s characteristic brain changes are actually present, regardless of symptoms. The 2024 diagnostic criteria make this explicit: Alzheimer’s disease should be defined by biology, not by clinical symptoms alone.
Amyloid PET Scans
An amyloid PET scan uses a radioactive tracer injected into the bloodstream that binds to amyloid plaques in the brain, making them visible on imaging. When compared against autopsy findings, amyloid PET scans show 88% to 98% sensitivity (meaning they correctly identify people who have plaques) and 80% to 95% specificity (meaning they correctly rule out people who don’t). This makes amyloid PET one of the most reliable tools available for confirming the presence of Alzheimer’s pathology during life.
Tau PET Scans
A separate type of PET scan targets tau tangles rather than amyloid plaques. Tau tangles spread through the brain in a predictable pattern: they begin in memory areas of the temporal lobe, move to regions in the back and sides of the brain involved in language and spatial awareness, and eventually reach the frontal lobes. A tau PET scan shows not just whether you have Alzheimer’s, but how far it has progressed. More tau, spread across more brain regions, means more advanced disease. Under the 2024 criteria, amyloid PET is used to diagnose the presence of Alzheimer’s, while tau PET is used to stage its severity.
Spinal Fluid (CSF) Testing
A lumbar puncture (spinal tap) can measure levels of amyloid and tau proteins in the cerebrospinal fluid that surrounds the brain. Doctors look at ratios between specific forms of these proteins. Two key ratios are used clinically. When the ratio of a phosphorylated form of tau (pTau181) to a specific amyloid fragment (called Abeta42) exceeds 0.023, results are consistent with a positive amyloid PET scan, meaning Alzheimer’s pathology is likely present. A second ratio, total tau to Abeta42, is considered positive above 0.28. These fluid biomarkers can confirm Alzheimer’s biology without the cost of a PET scan, though the procedure is more invasive.
Blood Tests
Blood-based biomarkers represent the most accessible frontier. A protein called phosphorylated tau 217 (p-tau217) has shown roughly 90% accuracy at identifying people who have Alzheimer’s-related brain changes confirmed by PET imaging or autopsy. In early studies from the National Institutes of Health, blood measurements of p-tau217 were about 90% accurate at distinguishing people who later developed dementia symptoms and similarly accurate at matching autopsy findings. These tests are not yet approved for diagnosing Alzheimer’s before symptoms develop, but they are increasingly being incorporated into clinical evaluation of people who already have cognitive complaints.
How Accurate Is a Clinical Diagnosis?
When experienced clinicians diagnose Alzheimer’s using the full combination of cognitive testing, imaging, and clinical judgment, about 90% of those diagnoses are confirmed at autopsy. That 90% agreement rate sounds high, but it means roughly 1 in 10 people diagnosed with Alzheimer’s during life actually had a different form of dementia. Conditions commonly mistaken for Alzheimer’s include Lewy body dementia, frontotemporal dementia, and vascular dementia, all of which can produce overlapping symptoms.
Biomarker testing is narrowing that gap. By directly measuring amyloid and tau rather than relying on symptom patterns alone, doctors can now distinguish Alzheimer’s from look-alike conditions with greater confidence. This distinction matters more than ever because newer treatments specifically target amyloid and would be ineffective, or even harmful, in someone whose dementia has a different biological cause.
What “Definitive” Means Now
The answer to whether Alzheimer’s can be definitively diagnosed depends on how strict your definition is. Autopsy remains the only way to directly count plaques and tangles under a microscope, and it is still considered the ultimate confirmation. But amyloid PET scans, with sensitivity as high as 98%, come remarkably close to that same certainty in living patients. Spinal fluid ratios and emerging blood tests add further layers of biological confirmation.
The 2024 criteria acknowledge this reality by redefining Alzheimer’s as a biological entity that exists on a continuum. The disease begins with invisible brain changes long before memory problems appear, progresses through stages of increasing biological burden, and eventually produces symptoms. Biomarkers can now detect and stage this process at multiple points along the way. For clinical purposes, though, biomarker testing is currently recommended only for people who already have symptoms. Testing asymptomatic individuals is still limited to research settings.