Corrections Policy
We correct errors promptly and visibly.
What we correct
- Factual errors (pricing, features, integrations, KLAS scores, G2 ratings, dates)
- FDA clearance mis-statements (clearance number, clearance date, indication-for-use scope)
- HIPAA/BAA availability errors
- EHR integration mis-statements
- Clinical-evidence mischaracterisations
- Broken external links
How to report
Email contact@ai-health-apps.com (see Imprint) with: (1) page URL, (2) the specific claim, (3) a primary source if possible.
Turnaround
We do not promise a fixed turnaround time. When we confirm an error, we correct the page; a material correction also gets a note at the top of the affected page and an entry in the log below.
Disclosure
Material corrections are disclosed with a visible note at the top of the affected page stating what was changed and when. Minor typos do not require disclosure.
Corrections log
-
— Woebot Health retired the Woebot app on 30 June 2025. Its FAQ, checked on 4 November 2026, says: "The Woebot app was retired June 30, 2025." Our Woebot review already said so, but other pages still listed Woebot alongside apps you can use today. The health apps hub ranked it first of 14, with its App Store rating and a link to Woebot's website, and listed it in its comparison table; the homepage comparison table also listed and linked it, and the pricing comparison listed it as free. The mental health apps guide showed it first among the apps it evaluated, with a link to Woebot's website; the AI mental health apps page ranked it second; the free mental health apps guide counted it as the third of eight best free apps, and its description also named MoodKit, which the guide does not list; and the Wysa vs Woebot comparison's price table still showed a free tier for it. The hub's FAQ, the anxiety guide and the HIPAA guide grouped it with AI therapy and consumer apps in current use. Woebot is now out of every ranking, comparison table and price table, the free apps guide lists seven apps and its description names PTSD Coach instead of MoodKit, and these pages mention Woebot only as a retired app. Its review stays up as a historical review. In the same change, the BetterHelp and Calm reviews, the Calm vs Headspace comparison and the AI therapy safety guide, which named Woebot in passing beside current apps, now leave it out or use the past tense; as minor wording changes, they carry no correction note.
Pages: /best-ai-health-apps/; /pricing-comparison/; /guides/best-mental-health-apps/; /best-ai-mental-health-apps/; /guides/best-free-mental-health-apps/; /compare/wysa-vs-woebot/; /guides/best-ai-for-anxiety/; /guides/ai-mental-health-hipaa-compliance/
-
— Our cheapest-scribe guide said a $400/month AI scribe pays back in 4–6 visits a month at $80–$120 of wRVU revenue per visit, and that these need not be visits more than baseline, just visits charged at all. Visits a practice would bill anyway do not offset the scribe's cost; only extra visits do. Counting whole billed visits, $400 takes 4 extra visits at $120 each and 5 at $80, so the guide now says the scribe pays back with roughly 4–5 extra billed visits a month. The same guide said KLAS data and published research show ambient AI scribes cut documentation time by ~50%, freeing 1–2 patient encounters a day; its link to that research went to a PubMed record of a study of topical perchlorate in rats, which the page's structured data described as a Tierney AA et al study of AI scribe time savings. The guide now says what our scribes hub says: in a 2024 Stanford study of DAX Copilot, physicians reported a median of 20 minutes saved per half-day clinic session, with wide variation, and we found no study showing a fixed number of extra visits per day.
Pages: /guides/cheapest-ai-medical-scribe/
Sources: Shah SJ et al., Ambient artificial intelligence scribes: physician burnout and perspectives on usability and documentation burden, JAMIA (2024), a Stanford Health Care study of DAX Copilot; PubMed 38427834, the record the guide linked: Topical Application of Perchlorate: A Toxicity Study in Rats
-
— Our scribe pages gave Nuance DAX Copilot's price as $369–$830 a month per provider, as starting at $369 or $369+, as $600+, $600–$800 or $600–$830 a month, and on the scribes hub and homepage cards as $444–$600 'confirmed June 2026', with a setup fee of $650 for the first user and $250 for each additional user. We found no Microsoft source for these figures except $600. Microsoft's Dragon Copilot page, checked on 3 November 2026, lists no price and asks buyers to contact a representative; Microsoft sells licences through its sales team or Microsoft partners; and in April 2026 Microsoft told its partners that the per-user list price would fall from 1 May 2026, without publishing the new figure. The only Microsoft price we found is a Microsoft Marketplace listing for the older DAX Copilot per-user plan in the US: $600 per user a month on a one-year subscription, with a professional-services quote required before purchase. The pages now say that Microsoft does not publish a current price and that each deal is quoted. Comparisons that used a DAX figure now leave DAX out: Suki's $299 entry price is about 2.5 to 7.7 times Freed's $39 to $119 a month, about 2.5 to 9 times with DeepScribe's estimated $350, and Freed's prices are about 13% to 40% of Suki's entry price. We no longer give a cost per note for DAX, call it the most expensive AI medical scribe, or say that other scribes cost less than DAX.
Pages: /compare/nuance-dax-vs-suki-vs-freed/; /best-ai-medical-scribes/; /guides/cheapest-ai-medical-scribe/; /guides/how-ai-scribes-work/; /compare/abridge-vs-nuance-dax/; /reviews/nuance-dax-review/; /reviews/freed-review/; /reviews/suki-review/; /reviews/deepscribe-review/; /reviews/abridge-review/; /pricing-comparison/
Sources: Microsoft Dragon Copilot product page (no price; contact a representative); Microsoft Partner Center announcements, April 2026: Announcing reduced Microsoft Dragon Copilot pricing; Microsoft Learn: Dragon Copilot licensing overview (bought through a Microsoft representative or partner); Microsoft Marketplace: Dax Copilot per user - US, plans and pricing
-
— Our Nuance DAX vs Suki vs Freed comparison said Freed Premier, at $119 a month (about $0.30 a note at 20 patients a day), pays for itself if it saves 5 minutes per encounter. We found no study or vendor document that sets that threshold, and the page gave no revenue or time value to work it out from, so we removed it. The per-note cost is unchanged.
Pages: /compare/nuance-dax-vs-suki-vs-freed/
Sources: Freed pricing page
-
— K Health discontinued its consumer app, the K Health Platform, on 31 December 2025. Its App Transition FAQ, checked on 3 November 2026, says: "K Health discontinued the K Health Platform as of December 31, 2025. We continue to offer care through our health system partners." Its pricing page now returns a not-found error. Our pages still presented K Health as available. The K Health review gave its prices as $49 a month for unlimited visits, $449 a year ($37.42 a month), $73 for a single visit and $73 a year for a Care+ subscription, linked to sign up, and called it a real bargain. The Ada Health vs K Health comparison called it the best AI-augmented telehealth subscription in the US at $73 a year, and the Teladoc review gave the same $73-a-year price and called K Health cheaper per month. The health apps hub ranked it 15th of 15 and, like the homepage comparison table, linked to its website, and the pricing comparison listed it as free. The Ada Health, Amwell, Fabric Health and Glass Health reviews pointed readers to it. The review and the comparison now say that K Health is discontinued and stay up for reference, without prices or sign-up links; K Health is out of the rankings, the comparison tables and the pricing table; and no page recommends it any more.
Pages: /reviews/k-health-review/; /compare/ada-vs-k-health/; /reviews/teladoc-review/; /best-ai-health-apps/; /pricing-comparison/; /reviews/ada-health-review/; /reviews/amwell-review/; /reviews/fabric-health-review/; /reviews/glass-health-review/
Sources: K Health App Transition FAQ (checked 3 November 2026)
-
— The evidence badges on our platform cards were set by matching words in our research notes, not by the evidence tiers our methodology page defines. BetterHelp and Hims & Hers showed “Strong Evidence” because their notes contained the words “not FDA-cleared” and “FDA-approved medications”: the only BetterHelp study we found is a 2019 outcome study with no control group, and the only Hims & Hers study we found is a 2026 patient-reported outcomes study with Hims & Hers authors. athenahealth and DrChrono showed “Strong Evidence” from the words “no FDA classification” and “no FDA-relevant claims”, and Future, Freeletics and Zing Coach showed “Emerging Evidence” from the words “No peer-reviewed clinical studies”. Each card now shows the methodology tier: Gold (multiple randomized trials), Silver (an FDA designation, clearance or approval, or an independent study), Bronze (outcome data only) or None (no validation found), with the study, FDA record or outcome report behind it linked under the card.
Pages: /best-ai-health-apps/; /guides/best-mental-health-apps/; /best-ai-medical-scribes/; /guides/best-ai-nutrition-apps/
Sources: Evidence tiers on our methodology page; BetterHelp outcome study, JMIR mHealth and uHealth 2019
-
— We called Lark the most evidence-based option in our nutrition-app guide and one of the most clinically credible chronic disease coaching platforms in the US. The Lark studies we found are observational outcome studies by Lark staff, such as a 2022 study of its diabetes prevention programme; we found no randomized trial of Lark, while Noom, in the same guide, has been tested in randomized trials. Both pages now say that.
Pages: /guides/best-ai-nutrition-apps/; /reviews/lark-review/
Sources: Lark diabetes prevention outcome study by Lark staff, Digital Health 2022; Noom randomized trial in adults with prediabetes, JMIR mHealth and uHealth 2020
-
— We said Noom users who complete the programme typically lose 5-7% of body weight at 6-12 months. In the 2020 randomized trial of Noom in 202 adults with prediabetes, completers had lost 5.6% at 6 months and 4.7% at 12 months. We also said a 2016 study of about 35,000 users found a median loss of around 5-7%; the study reports that 77.9% of users lost weight and 22.7% lost more than 10%, and we could not find that median in it. Both lines now say what the studies report.
Pages: /guides/best-ai-nutrition-apps/; /guides/does-noom-really-work/
Sources: Noom randomized trial in adults with prediabetes, JMIR mHealth and uHealth 2020; Noom Coach retrospective study, Scientific Reports 2016
-
— We called Tempus one of the most clinically credentialed AI companies in healthcare, a ranking we cannot source. The review now states the FDA record instead: premarket approval for xT CDx (P210011, 28 April 2023) and 510(k) clearances including Tempus ECG-AF (K233549, 21 June 2024).
Pages: /reviews/tempus-review/
Sources: FDA premarket approval record P210011 (xT CDx); FDA 510(k) K233549 (Tempus ECG-AF)
-
— We called Headspace's Ebb companion the most clinically thoughtful generative AI in consumer mental health. The only study of Ebb we found is a 2026 observational study by Headspace staff, and we found no randomized trial of Ebb; the Woebot review now says that.
Pages: /reviews/woebot-review/
Sources: Observational study of Ebb by Headspace staff, JMIR Formative Research 2026
-
— Our Wysa review's title and main heading stated 45+ peer-reviewed studies as fact. The count is Wysa's own: its September 2025 release says more than 45 peer-reviewed publications, and a PubMed search we ran on 2 November 2026 found 2 randomized trials of Wysa. The title and heading now attribute the count to Wysa, and the review links the release and the trials. Our free mental health apps guide also named Wysa the most evidence-based free app; PTSD Coach, a free app in the same guide, has been tested in several randomized trials for trauma symptoms, so the answer now describes the evidence for each instead.
Pages: /reviews/wysa-review/; /guides/best-free-mental-health-apps/
Sources: Wysa release, September 2025; PubMed: 2 randomized trials of Wysa; PubMed: 5 randomized trials of PTSD Coach
-
— Several pages said or implied that we test products hands-on: “We evaluated 15 AI consumer health apps hands-on”, “Tested & Ranked”, “Evidence + Privacy Tested”, an FAQ saying we test each app via a real account, “our own hands-on testing of the self-serve tiers”, hands-on testing listed as a ranking criterion, a methodology section describing our sign-up testing, and statements that each review says whether it was tested hands-on. No review on this site shows first-party test evidence, and the reviews do not say which research path applied, so we removed those claims and now say so. The pricing comparison also said all prices were verified as of April 2026; we cannot show that check, so we removed the line.
Pages: /best-ai-health-apps/; /guides/best-free-mental-health-apps/; /methodology/; /editorial-guidelines/; /editorial-independence/; /affiliate-disclosure/; /authors/editorial-team/; /compare/nuance-dax-vs-suki-vs-freed/; /about/; /pricing-comparison/
-
— Several pages described checks we do not run: prices re-verified quarterly; pricing, KLAS scores, FDA status and features re-checked every 90 days; FDA clearance status re-verified monthly; regulatory events reflected within 72 hours; reviews re-validated every 90 days; and pricing verification dates on all product pricing. The site footer and llms.txt said the site is updated quarterly, our scribe pricing guide and hub called their prices verified, and our 2026 outlook guide offered a quarterly newsletter, which we do not run. We removed those claims. The pages now say what is true: prices come from vendors' own pricing pages where vendors publish them, we do not re-check them on a schedule, most pages do not yet say when a price was taken, and every review, guide and comparison page shows the date it last changed.
Pages: /fact-checking-policy/; /editorial-guidelines/; /authors/editorial-team/; /methodology/; /about/; /guides/cheapest-ai-medical-scribe/; /best-ai-medical-scribes/; /guides/ai-healthcare-2026/
Sources: Fact-checking policy: pricing and update dates; Privacy policy: we do not run a newsletter service
-
— We said Noom's published trials show participants typically lose 5–7% of body weight over 16–32 weeks, said independent analyses confirm smaller losses without naming one, and said Noom and WW produce similar losses of 5–7% over 16–24 weeks. No Noom study we found reports a 32-week result. In the 2020 randomized trial of Noom in 202 adults with prediabetes, completers lost 5.6% of body weight at 6 months and 4.7% at 12 months, and across everyone assigned to Noom weight was 2.64 kg lower than in the control group at 6 months; in a 2023 randomized trial in adults with NASH, Noom users lost 5.4% over 16 weeks against 0.4% with standard care. A 2015 systematic review found that WW participants lost at least 2.6% more weight than control groups at 12 months. The guide and the review now give these figures.
Pages: /guides/does-noom-really-work/; /reviews/noom-review/
Sources: Noom randomized trial in adults with prediabetes, JMIR mHealth and uHealth 2020; Noom randomized trial in adults with NASH, Hepatology Communications 2023; Systematic review of commercial weight-loss programmes, Annals of Internal Medicine 2015
-
— We said Lark has nearly 2 million members. Lark's newest figure, in its 15 July 2026 release, is 2.5 million members supported. The review and our Lark platform card now give that figure as Lark's own count.
Pages: /reviews/lark-review/; /best-ai-health-apps/; /guides/best-ai-nutrition-apps/
Sources: Lark release, 15 July 2026
-
— Our anxiety guide said meta-analyses find that app effects improve when human coaching is added and that pure AI chatbots show smaller effects than blended care, and called Wysa's AI-plus-coach tier the most evidence-aligned design we recommend. It cited no study. A 2024 meta-analysis of 176 randomized trials reports small effects of apps on generalized anxiety symptoms, larger effects for CBT apps and apps with mood monitoring when anxiety was the main target, and larger effects on depression for apps with chatbot technology. The guide now says what that meta-analysis found.
Pages: /guides/best-ai-for-anxiety/
Sources: Meta-analysis of 176 randomized trials of mental health apps, World Psychiatry 2024
-
— Our ranked list of mental health apps said Wysa has the best evidence base of any AI chatbot, and our depression guide said Headspace's evidence base is unusually strong for a meditation app. Neither ranking had a source. The list now says we found 2 randomized trials of Wysa and that Wysa reports more than 45 peer-reviewed studies; the guide now gives the 2022 systematic review's findings for Headspace: 14 randomized trials, with Headspace use improving depression in 75% of the studies that measured it.
Pages: /best-ai-mental-health-apps/; /guides/best-ai-for-depression/
Sources: PubMed: 2 randomized trials of Wysa; Systematic review of Headspace and Calm RCTs, JMIR Mental Health 2022
-
— Our Teladoc review called Teladoc the largest telehealth network, with 50+ million members. The ranking had no source, and the member figure was out of date: Teladoc reported 100.1 million US Integrated Care members at 30 September 2026. The review now gives that figure and no ranking.
Pages: /reviews/teladoc-review/
Sources: Teladoc third-quarter 2026 results, 28 October 2026
-
— Our pages promised different correction times: 48 hours on our about and methodology pages; 72 hours for factual errors, 24 hours for FDA and regulatory errors and 5 business days for clinical evidence on our corrections policy; and 10 business days on our contact page. They disagreed with each other, and we cannot guarantee any of them, so we removed them. We do not promise a fixed turnaround time: when we confirm an error, we correct the page, and a material correction also gets a note on the page and an entry in our corrections log.
Pages: /corrections-policy/; /about/; /methodology/; /contact/
Sources: Corrections policy: turnaround
-
— Our anxiety guide said Woebot shut down in 2024 (in one place, mid-2024), and a heading in our depression guide said the same. Woebot Health's own FAQ says the Woebot app was retired on June 30, 2025, the date our other pages give. Both guides now say June 2025.
Pages: /guides/best-ai-for-anxiety/; /guides/best-ai-for-depression/
Sources: Woebot Health FAQ: the Woebot app was retired June 30, 2025
-
— Our scribe pages said Freed is the only AI scribe with public pricing and the only one with a self-serve checkout, public pricing and a free trial; that Suki lists its prices on its website (on suki.ai/pricing) and is one of the few scribes with public pricing; that Heidi Health does not publish pricing; and that no ambient AI scribe is free and Freed's $39 Starter plan is the absolute cheapest. On 2 November 2026 Suki's website had no pricing page (suki.ai/pricing returned a not-found page) and directed buyers to its sales team, and Heidi Health's pricing page listed a free plan and a paid plan with a 14-day free trial. The pages now say that Freed and Heidi Health publish their prices and that the Suki prices we give are reported figures, not Suki's list prices.
Pages: /best-ai-medical-scribes/; /guides/cheapest-ai-medical-scribe/; /reviews/freed-review/; /reviews/suki-review/; /pricing-comparison/
Sources: Freed pricing page; Heidi Health pricing page; Suki website (sales contact, no pricing page)
-
— Our Lark Health review said that if insurance does not cover Lark, the only way to pay for it directly is a 23andMe wellness partnership at $14.99–$19.99 a month, and that Lark had recently integrated 23andMe genetic data. That offer dates from January 2019, when 23andMe offered Lark's Wellness Program to its Health + Ancestry customers at $19.99 a month ($14.99 introductory). On 2 November 2026 Lark's page for individuals offered Lark only through an insurance eligibility check, with no self-pay option and no mention of 23andMe. We removed the direct-pay tier and now describe the 23andMe integration as a 2019 announcement we could not confirm is still offered.
Pages: /reviews/lark-review/
Sources: Lark for individuals (insurance eligibility check); 23andMe blog: Your Digital Health Coach Is Here (8 January 2019)
-
— Our Suki AI and Nuance DAX reviews priced Freed at $99–$179 a month; Freed's pricing page and help centre, checked on 2 November 2026, show $39 to $119 a month for its individual plans. Our cheapest-scribe guide and Freed review said Freed's Starter plan covers 30 visits a month; Freed says up to 40 notes per month. Our Freed review called its 7-day trial the closest thing to free in the AI medical scribe market, and our scribes hub called Freed the cheapest AI medical scribe with publicly listed pricing, although Heidi Health publishes a free plan. The pages now use Freed's own figures and wording, and the cheapest-scribe guide's pricing table now includes Heidi Health: its free plan, and its Clinician plan at $110 per user a month billed annually or $150 billed monthly (Heidi Health's US pricing page, 2 November 2026).
Pages: /reviews/suki-review/; /reviews/nuance-dax-review/; /reviews/freed-review/; /guides/cheapest-ai-medical-scribe/; /best-ai-medical-scribes/
Sources: Freed pricing page; Freed help centre: Freed Pricing & Discounts; Heidi Health pricing page (US)
-
— Several of our pages said existing Woebot users may still have access through an insurer or employer, that Woebot now operates only through enterprise, payer or health-system deals, or that existing users were migrated; and several said Woebot Health publicly cited the FDA regulatory burden as the reason for the shutdown. Woebot Health's FAQ, last modified 3 December 2025 and the newest statement on its website, says the Woebot app was retired on June 30, 2025, that new accounts can no longer be created and that previous accounts can no longer be accessed; it gives no reason. The FDA reason comes from interviews founder and CEO Alison Darcy gave STAT in July 2025. The pages now say what the FAQ says and attribute the FDA reason to those interviews.
Pages: /reviews/woebot-review/; /compare/wysa-vs-woebot/; /guides/best-free-mental-health-apps/; /guides/best-mental-health-apps/; /guides/best-ai-for-anxiety/; /guides/best-ai-for-depression/; /best-ai-health-apps/; /best-ai-mental-health-apps/; /guides/is-ai-therapy-safe/; /guides/ai-healthcare-2026/; /reviews/wysa-review/
Sources: Woebot Health FAQ: the Woebot app was retired June 30, 2025; previous accounts can no longer be accessed; STAT: Why Woebot, a pioneering therapy chatbot, shut down (2 July 2025)
-
— Several of our scribe pages called Freed the cheapest or most affordable AI scribe. Our Suki AI and DeepScribe reviews called it the cheapest mainstream ambient AI scribe; our Nuance DAX vs Suki vs Freed comparison answered 'Which is the cheapest?' with 'Freed by a wide margin'; the title, heading and strengths list of our Freed review, and the Freed cards on our scribes hub and homepage, called it the most affordable AI scribe; and our Freed vs DeepScribe comparison called it the lowest-cost path to ambient AI documentation. Heidi Health has a free plan ($0 on its US pricing page, checked on 2 November 2026), so Freed is not the cheapest AI scribe. The pages now give Freed's prices ($39 to $119 a month per provider) without the superlative, and the three-way comparison now says Freed is the cheapest of the three products it compares and that Heidi Health has a free plan.
Pages: /reviews/suki-review/; /reviews/deepscribe-review/; /compare/nuance-dax-vs-suki-vs-freed/; /reviews/freed-review/; /compare/freed-vs-deepscribe/; /best-ai-medical-scribes/
-
— Several of our scribe pages gave price ratios that the prices on the same pages do not support. Our Freed review and our Nuance DAX vs Suki vs Freed comparison said Freed is 3-10x cheaper than the alternatives, and the comparison also put Freed at 1/4 to 1/8 of their price; our Freed vs DeepScribe comparison said Freed is roughly 4-10x cheaper than DeepScribe at every comparable tier and roughly 1/4 to 1/8 of its price; our Nuance DAX review said Freed is 75%+ cheaper than DAX; and our cheapest-scribe guide said Abridge and Nuance DAX are 2–4× the price of Freed. Against the lowest prices these pages list (Suki $299, Nuance DAX $369 and DeepScribe an estimated $350 a month per provider), those entry prices are about 2.5 to 9.5 times Freed's $39 to $119 a month (DeepScribe's about 2.9 to 9 times), and Freed is about 68% to 89% below DAX, not 75% or more. Our Freed review also said Freed Premier ($119) saves $250-$680 a month against Suki ($399) or Nuance DAX ($600+); the differences are $280 and at least $481. And our cheapest-scribe guide said DAX's setup fee for three providers (roughly $1,150) equals two months of Freed Premier for one provider; at $119 a month it is almost ten. The pages now give the computed figures, or the prices side by side.
Pages: /reviews/freed-review/; /compare/nuance-dax-vs-suki-vs-freed/; /compare/freed-vs-deepscribe/; /reviews/nuance-dax-review/; /guides/cheapest-ai-medical-scribe/
Sources: Freed pricing page
-
— We placed Calm in an evidence tier called Weak, which our methodology does not define, and said we could not make evidence-based claims about Calm specifically. A PubMed search on 1 November 2026 found at least 8 randomized trials of the Calm app, so under our methodology's rule (Gold: multiple peer-reviewed RCTs) Calm is now in the Gold tier. It still has fewer trials than Headspace (at least 17 in the same search), and five of the eight Calm trials include an author who has been Calm's Director of Science.
Pages: /guides/best-mental-health-apps/; /reviews/calm-review/
Sources: PubMed: 8 randomized trials of the Calm app (search run 1 November 2026); PubMed: 17 randomized trials of the Headspace app (search run 1 November 2026)
-
— We said the Predetermined Change Control Plan for PathAI's AISight Dx lets PathAI update AI models without re-clearance, and described AISight Dx as AI-powered primary diagnosis. FDA cleared AISight Dx (K243391) as software for viewing and managing digital slide images, an aid to the pathologist for primary diagnosis; its change plan covers adding cleared displays, scanners, file formats and browsers.
Pages: /guides/fda-cleared-ai-health-tools/; /reviews/pathai-review/; /best-ai-medical-scribes/
Sources: FDA 510(k) summary K243391
-
— We said studies show Viz.ai cuts door-to-puncture time by 30 to 60 minutes and that this directly reduces disability and deaths. A 2025 meta-analysis of 12 studies found shorter treatment times, no significant difference in mortality and only a non-significant trend toward better 90-day function.
Pages: /reviews/viz-ai-review/
Sources: Viz.ai stroke workflow meta-analysis, Translational Stroke Research 2025
-
— We said a JMIR study found Ada performed closest to GP-level triage on routine cases, and cited independent studies showing Ada among the more accurate tools. The 2023 JMIR mHealth and uHealth study, which Ada did not fund, compared Ada with ChatGPT, WebMD and emergency physicians, not GPs: Ada gave the fewest unsafe triage recommendations (14%) but agreed with the physicians' triage less often (62%) than ChatGPT 4.0 (76%) or WebMD (70%). We also called Ada the most clinically validated symptom checker; we found no source that ranks symptom checkers that way.
Pages: /reviews/ada-health-review/
Sources: Fraser et al., JMIR mHealth and uHealth 2023 (PMC10582809)
-
— We called Woebot the most clinically validated digital mental health tool in existence and the most studied AI therapy chatbot, said its evidence base was unmatched, and said Wysa had comparable clinical validation. We found no source that ranks these tools that way. A PubMed search on 1 November 2026 found at least 7 randomized trials of Woebot, five of them with Woebot Health employees or its founder among the authors, at least 17 of the Headspace app and 2 of Wysa.
Pages: /reviews/woebot-review/; /guides/best-mental-health-apps/; /guides/best-free-mental-health-apps/; /reviews/wysa-review/
Sources: PubMed: 7 randomized trials of Woebot (search run 1 November 2026); PubMed: 17 randomized trials of the Headspace app (search run 1 November 2026)
-
— We said Headspace has the strongest clinical evidence of any meditation app and 70+ published studies, and that its Ebb AI companion had shown efficacy in randomized trials. The trial counts we can source compare Headspace only with Calm: 14 RCTs against 1 in a 2022 systematic review of the two apps, and at least 17 against at least 8 in a PubMed search on 1 November 2026. Half of the 14 Headspace trials in that review reported a conflict of interest involving the company. We could not confirm the 70+ count, and we found no randomized trial of Ebb.
Pages: /reviews/headspace-review/; /guides/best-mental-health-apps/; /guides/best-ai-for-anxiety/; /guides/best-free-mental-health-apps/; /reviews/insight-timer-review/; /guides/is-ai-therapy-safe/
Sources: Systematic review of Headspace and Calm RCTs, JMIR Mental Health 2022; PubMed: 17 randomized trials of the Headspace app (search run 1 November 2026)
-
— We placed Wysa in our Silver evidence tier, called it the strongest pure-AI option with the strongest evidence base among AI chatbots, and said its randomized trials demonstrated significant reductions in anxiety and depression. PubMed lists two randomized trials of Wysa, so under our methodology (Gold: multiple peer-reviewed RCTs) it belongs in Gold. In a 2024 trial of 68 adults with arthritis or diabetes, those using Wysa reported lower depression and anxiety scores after four weeks while the control group did not change; a 2026 NHS trial of 76 patients found no evidence that Wysa reduced depression. We found no source that ranks AI chatbots by evidence.
Pages: /guides/best-mental-health-apps/; /guides/best-ai-for-depression/; /guides/best-ai-for-anxiety/; /reviews/wysa-review/; /guides/is-ai-therapy-safe/
Sources: MacNeill et al., JMIR Formative Research 2024 (PMC11176869); Meinert et al., International Journal of Social Psychiatry 2026 (PMC13569919)
-
— We called Noom the most clinically validated weight-loss app on the market and said it has the strongest clinical evidence base of any weight-loss app. We found no source that ranks weight-loss apps that way. Noom says it has published more than 40 peer-reviewed scientific articles.
Pages: /reviews/noom-review/; /guides/best-ai-nutrition-apps/
Sources: Noom research page
-
— We said a 2021 randomized trial of Calm reported better sleep quality in adults with sleep disturbance. The trial compared Calm with a wait-list on fatigue, daytime sleepiness and pre-sleep arousal, which all improved; it measured sleep quality only in the Calm group.
Pages: /best-ai-mental-health-apps/
-
— We listed a clinically proven reduction in door-to-treatment time as a Viz.ai strength. The evidence is a 2025 meta-analysis of 12 studies comparing stroke care before and after Viz.ai: it found shorter door-to-groin-puncture and CT-to-treatment times for large vessel occlusion stroke, and no significant difference in patient outcomes. We now describe it that way.
Pages: /reviews/viz-ai-review/
Sources: Viz.ai stroke workflow meta-analysis, Translational Stroke Research 2025
-
— We described Tempus xT as an FDA 510(k) clearance. FDA records show xT CDx was approved through premarket approval (PMA P210011) on 28 April 2023.
Pages: /guides/fda-cleared-ai-health-tools/; /reviews/tempus-review/
-
— A Tempus feature line described an April 2026 FDA De Novo for real-time arrhythmia detection. We found no such FDA record. The cleared product is Tempus ECG-AF, a 510(k) (K233549, 21 June 2024) that looks for signs of atrial fibrillation or flutter within the next 12 months; FDA's summary says its results should not be used for patient monitoring.
-
— We said Viz.ai held more than 50 FDA 510(k) clearances, the broadest portfolio in imaging AI. FDA's device database lists 10 510(k) clearances and 2 De Novo grants with Viz.ai as the applicant, and 35 510(k) clearances for Aidoc (openFDA data as of 19 October 2026). We removed the claims that Viz.ai has the broadest portfolio and leads on clearances.
Pages: /guides/fda-cleared-ai-health-tools/; /guides/ai-healthcare-2026/; /reviews/viz-ai-review/; /best-ai-medical-scribes/
Sources: openFDA 510(k) and De Novo records for Viz.ai; openFDA 510(k) count for Aidoc
-
— We gave the FDA-qualified name of PathAI's drug development tool as AIM-MASH. The FDA qualified it on 8 December 2025 as the AI-Based Histologic Measurement of NASH (AIM-NASH); PathAI now sells it as AIM-MASH AI Assist.
Pages: /guides/fda-cleared-ai-health-tools/; /reviews/pathai-review/; /best-ai-medical-scribes/
Sources: FDA: FDA Qualifies First AI Drug Development Tool (8 December 2025)
-
— We said Freed had no published studies. One study is indexed in PubMed: a 2025 paper in South Dakota Medicine that compared Freed's notes with first-year medical students' notes in simulated clinical exams. Freed still has no clinical-outcome studies.
Pages: /compare/freed-vs-deepscribe/; /compare/nuance-dax-vs-suki-vs-freed/; /reviews/freed-review/
-
— We compared Ada's results with a 69.5% average for other apps that the cited 2020 BMJ Open study does not contain. From the study's own figures, the other seven apps averaged 33.4% top-3 accuracy (Ada 70.5%) and 73.4% condition-suggestion coverage (Ada 99%). We also called the study independent: Ada Health funded it, and all of its authors but one are or were employees of, contractors for, or equity holders in Ada Health.
-
— We described Noom's 75% weight-maintenance result as a randomized trial. It comes from a 2023 observational survey of 840 people who had already lost at least 10% of their weight with Noom.
Pages: /reviews/noom-review/
Sources: May et al., Obesity Science & Practice 2023 (PMC10551118)
-
— We gave Calm's single RCT as a current count. It comes from a 2022 systematic review whose searches ended in May 2021 (14 RCTs of Headspace, 1 of Calm). At least one more Calm RCT, a 2022 workplace trial, has been published since.
Pages: /best-ai-health-apps/; /guides/is-ai-therapy-safe/; /guides/best-mental-health-apps/; /reviews/headspace-review/; /reviews/calm-review/
Sources: O'Daffer et al., JMIR Mental Health 2022 (PMC9533203); Huberty et al., JMIR mHealth and uHealth 2022 (PMC9557765)
-
— We listed pulmonary embolism, aortic emergencies, intracranial hemorrhage, cervical spine fractures and pneumothorax among the conditions the FDA cleared for Aidoc's CARE model in January 2026. FDA's 510(k) summary (K252970) lists 11 body-CT findings, such as appendicitis, diverticulitis, bowel obstruction and liver, spleen and kidney injury; Aidoc counts 14 by adding three previously cleared indications.
Pages: /reviews/aidoc-review/
Sources: FDA 510(k) summary K252970; Aidoc release, January 2026
-
— We described Lark Health's programmes as FDA-cleared. We found no FDA 510(k) or premarket approval record for Lark (openFDA, 31 October 2026), so we removed the claim.
Pages: /guides/best-ai-nutrition-apps/
Sources: FDA 510(k) database (search the applicant name Lark); FDA premarket approval database
-
— We cited independent studies showing Ada among the more accurate symptom checkers. The study behind that line, a 2020 BMJ Open vignette study, was funded by Ada Health, and all of its authors but one are or were employees of, contractors for, or equity holders in the company.
Pages: /compare/ada-vs-k-health/
-
— Our scribes hub answered 'Will an AI scribe pay for itself?' with: for most solo physicians, yes; KLAS data and published research show typical time savings of around 50% on documentation, which translates to 1–2 additional patient encounters per day, and a $300/month scribe pays back in roughly 3–6 patient visits per month. We changed that answer on 31 October 2026 but did not log the change at the time. The answer now says it depends on your schedule, cites the perceived time savings reported in a 2024 Stanford study of DAX Copilot, says we found no study showing a fixed number of extra visits per day, and gives the payback as roughly 3–4 extra patient visits a month: at $80–$120 of wRVU revenue per visit, $300 takes 3 extra visits at $120 each and 4 at $80.
Pages: /best-ai-medical-scribes/