Talk:Dementia

Notes

I'm experimenting with using AI to analyse paragraphs against sources with the help of a friend. I'll note some of the issues it found below, with the caveat that this isn't a comment from an expert, but from AI and a random person on the internet.

Lead

• "There is no known cure for dementia." is sourced to a review on palliative care, not on treatments for dementia. Can we have a more specific source.
• We note that the benefit of "Acetylcholinesterase inhibitors" "may be minor" but not that the modest effect is symptomatic, not not disease‑modifying.

Signs and symptoms

• Question whether "memory plus one other cognitive region" is truly "typical" or just a feature of some sub-types.
• Question whether "can often fall behind on bill payments" is true. Indeed all uses of the word "often" should be queried, as "dementia" is wide in scope and worldwide.
• Question whether "The symptoms progress at a continuous rate over several stages" is true, as sources suggest rate and pattern may vary.
• Is it true to say "hallucinations" are "most often visual" for all types of dementia, or should this be specific to some?
• "empowering carers" is vague and unclear how this may improve symptoms.
• "potentially as many as 79.6% of people with dementia in nursing homes may experience pain" is a ridiculously accurate figure and doesn't accurately describe this as just one prevalence figure in one study and at the upper range (so probably unlikely). For example the source says "prevalence varied from 8.6% to 79.6%" which says to me really that there's more we don't know about prevalence than we do. To be honest, I think the fairest use of that source is to recommend further research rather than quote a random number to three significant figures. It is really saying "we don't know".
• "there can sometimes be as many as four of these comorbidities" indicates this is some kind of upper bound. The source was in fact just making the point that people with dementia have more additional health conditions than those without: "Patients with dementia are more likely to have multiple health conditions. A total of 22% with 3 or more comorbidities and 8% with 4 or more comorbidities, compared to 11% and 3% respectively in the all patient group." There really is no upper bound.

Stages

• There's no universal agreement on how many stages or how to describe them. We are over stating the "four stages" as though it is a standard model.
• "numeric scales" isn't very informative.
• The grouping is confusing.
• "more accurately" is an unsupported claim and not neutral.
• GDS and FAST seems to be intermingled here.
• Sources and are not ideal.
• We mention these numeric scales but don't go into any detail about them. Should we?

Pre-clinical

• Claiming "the sense of smell is lost" overstates this.
• "a loss of appetite leading to poor nutrition" is too strong a claim.
• "It is suggested that" is mentioning a plausible hypothesis and doesn't belong here. Either in causes or even in research, assuming it has weight, but this isn't an established "fact".

Prodromal

• "Of those diagnosed with MCI, 70% later progress to dementia." Seems unlikely to be generally true with this degree of exactness. I can't verify the source. Should be removed or qualified.
• Source is not acceptable (entire website, not MEDRS).
• The "DSM-5 and ICD-11" claim only cites DSM.
• "regulates microbiome signaling" implies proven causal control whereas the literature describes modulation and associations and ongoing investigation. This hypothesis also doesn't belong in this section.

Early

• The "24 to 30" range doesn't indicate that cutoffs vary with age, education and language. Source for this paragraph is an entire and old book, so needs replacing.
• The "memory difficulty" and "word finding" issues are specific to some dementia's early phases but not all. This article topic has a challenge for presenting when things sometimes/usually occur in all dementias and when they are specific.
• The "dementia with Lewy bodies and frontotemporal dementia" types seem to needlessly mix the two and cite a source only for the former that isn't an ideal source for this claim.

Middle

• "MMSE scores between 6 and 17 signal moderate dementia." is disputed. Sources may use 10-17 for moderate and 0-9 for severe, so this range given here doesn't seem consensus.

Late

• Again question whether "usually" or "typically" are appropriate, rather than degrees of risk and variability.
• "terminal lucidity" should be described as rare and poorly understood rather than the vague "in some cases".
• We mention pureeing or thickening food but only for "ease of feeding" rather than to manage aspiration risk.

Alzheimer's disease

• Initial two paragraphs lack sources. The 60-70% figure is not stated to be an estimate, rather than a hard fact. The "Trouble with" sentence mixes one symptom with a parenthetical explation with three others that all "fail", which is hard to parse. Also the insight issue is expanded on in the next sentence, and so maybe one sentence should cover that entirely. As a general point: sometimes we should include the standard terminology (it isn't plagiarising our articles of a term is standard) alongside an explanation. If our attempt at "dumbing down" technical terms is is too novel and lacks power, the reader might not appreciate we are talking about the same thing they are reading elsewhere.
• "Little is known" sentence is bad grammatically and not actually true. There is extensive research but little agreement we have a settled explanation.
• "This is due to the fact that" is superfluous. Brain tissue can still only be studied after death. A scan isn't really a tissue examination in the way people think of such. The claim about brain scans contradicts the earlier claim that a scan was insufficient for diagnosis.
• "Extracellular senile plaques" is nearly 100% technical and offers no help to the general reader.
• The "Amyloid causes" sentence presents an unproven hypothesis as fact.
• "Several articles" likely does not belong here. Hypothetical explanation of one possible risk factor is undue, and the reader cares not about "several articles".

Vascular

• Question whether the 20% figure is up-to-date and/or needs qualified.
• Question whether all DLB symptoms are "more frequent, more severe, and earlier presenting than in the other dementia subtypes" vs some (e.g., visual hallucinations, fluctuations, RBD).

Parkinson's disease

• I can't make sense of this sentence.

Frontotemporal

• Unsupported non-standard claim about six main types of FTD.
• "mostly they have a difficulty coordinating the muscles they need to speak" is disputed as it implies poor muscle coordination vs a problem with planning and programming speech movements.
• The existence of a logopenic variant (usually associated with Alzheimer pathology) should be noted for completeness
• "FTD‑ALS" question whether the sources support this. "tau associated" is pretty meaningless to the general reader.

More to follow. -- Colin°^Talk 19:54, 19 January 2026 (UTC)

So one thing is not clear from your post: did you check each of these items yourself? Johnjbarton (talk) 00:22, 20 January 2026 (UTC)

Yes, I'm going through the list of issues found and I've posted here what I think is a genuine problem, plus some comments of my own as I read the paragraphs. I'm only making a caution about AI in the sense that a few of the assertions it makes are at a level I can't confirm, but seem reasonable to me. I think all of the above issues are worth addressing. I've got more to follow too when I get time. -- Colin°^Talk 13:38, 22 January 2026 (UTC)

Sources

• "Memory loss: a practical guide for clinicians" is used 16 times. It is 15 years old and no page numbers. We should replace that and follow the practice in FA's of giving specific pages for multi-page works.
• "Dementia care at a glance." is a 15 year old book and no page numbers.

-- Colin°^Talk 13:35, 14 January 2026 (UTC)