Some apps now come by prescription, with evidence that ranges from strong to thin

Key takeaways

• As of November 2025, FDA had authorized fewer than 20 digital mental health devices, and none of them used artificial intelligence.
• Online programs that teach cognitive behavioral therapy for insomnia have the strongest record, with benefits that lasted a year in two trials.
• The first prescription app for depression, Rejoyn, was cleared in 2024 even though its main trial analysis narrowly missed its own target.
• Medicare began paying for some prescription mental health apps in 2025 and added ADHD apps in 2026, but only alongside ongoing care from a clinician.
• Before you start one, ask what the trial compared it with, whether it changed daily life and what happens if the company closes.

In March 2024, the Food and Drug Administration cleared Rejoyn, a six-week smartphone program for adults with major depression⁷. News reports called it the first prescription digital treatment for the condition⁸.

To win clearance, its maker had to show the app was “substantially equivalent” to a device already on the market. The comparison device named in FDA’s records is reSET, a 2017 program for substance use disorder⁷.

The trial result is just as surprising. In the analysis the company picked in advance as its main test, Rejoyn users improved more than people using a look-alike sham app. But the gap fell just short of the study’s own bar for statistical significance⁷.

That doesn’t make Rejoyn useless. It shows that “FDA-cleared” on an app can mean very different things. Some prescription apps have strong, lasting evidence. Others rest on one small or mixed trial.

This guide explains what digital therapeutics are, what their trials found, how Medicare now pays and what to ask before you start one.

Digital therapeutics are software that has to back up a medical claim

A digital therapeutic is software, usually an app, that is meant to treat a disease or help treat it. FDA regulates this kind of software as a medical device. The mental health devices it has authorized are usually prescription-only and meant to treat a condition, add to its treatment or help diagnose it².

That sets them apart from the many wellness apps in app stores. FDA says most consumer mental health apps are wellness products it never reviews, and that the line between the two has confused users². Our wellness apps guide covers what research shows for that much larger group.

The regulated group is small. As of November 2025, FDA had authorized fewer than 20 digital mental health devices². It had authorized more than 1,200 devices that use artificial intelligence, but none for mental health².

Most of the treatment apps deliver a structured talk therapy. Cognitive behavioral therapy, or CBT, teaches skills for changing unhelpful thoughts and habits. The app takes over some of the teaching a therapist would do.

The first prescription app opened a door that later apps walked through

A new kind of device with no close match on the market can reach patients through a route called De Novo, if its risks can be managed with FDA’s general or special controls¹. When FDA grants one, it writes a new device category. Later products can then take a shorter path, called a 510(k), by showing they match the first one¹.

Pear Therapeutics’ reSET took the De Novo route in September 2017³. FDA created a category called computerized behavioral therapy devices for psychiatric disorders³. Somryst for insomnia in 2020 and Rejoyn for depression in 2024 both named reSET as the device they matched⁴⁷.

The trial behind reSET deserves a closer look. The study, run through the National Institute on Drug Abuse’s Clinical Trials Network, followed 507 people in outpatient addiction treatment for 12 weeks³. About half got their usual treatment. The rest got a reduced amount of usual treatment plus the program.

Among people whose main drug was not an opioid, 40.3% of program users stayed abstinent in the final four weeks, compared with 17.6% with usual care³. For people whose main drug was an opioid, the program did not help, so FDA left them off the label³.

Two details rarely make the headlines. First, the study tested a desktop version, not the phone app³.

Second, program users also earned rewards for clean drug tests and finished lessons, a method called contingency management³. The prizes were usually worth about $1, and the comparison group got none³. FDA noted that the trial couldn’t show how well the program works without those rewards³.

Our companion article on behavioral medicine explains why rewards work so well.

Insomnia programs have the strongest record

If any digital therapeutic has earned trust, it’s the online version of CBT for insomnia, or CBT-I. This approach teaches people to change the habits and thoughts that keep them awake, without sleeping pills.

Somryst, cleared in March 2020 for adults 22 and older, delivers CBT-I over nine weeks in six lessons⁴. FDA’s clearance rested on two trials of SHUTi, an earlier web-based version of the same program⁴. Both were randomized controlled trials, where people are assigned by chance to one group or another.

The larger trial, called GoodNight, enrolled 1,149 adults⁴. After nine weeks, 62.8% of program users had a meaningful drop in insomnia severity. That compared with 14.0% of people using a comparison program matched for time spent⁴.

A year later, the gap was smaller but still wide: 59.3% versus 25.2%⁴. At nine weeks, program users fell asleep about 23 minutes faster on average, while the control group barely changed⁴.

A second trial of 303 adults at the University of Virginia found similar results⁴. Neither trial had participants report side effects⁴. Big Health’s SleepioRx, another CBT-I program, was cleared in 2024⁹. Our wellness apps article covers the research on Sleepio and the company ties behind it.

For ADHD and depression, apps moved test scores more than daily life

In June 2020, EndeavorRx became the first device in a new FDA category: digital therapy for ADHD⁵. Children ages 8 to 12 play a video-game-style program on a tablet for about 25 minutes a day, five days a week⁵.

In the STARS-ADHD trial, 348 children used either EndeavorRx or a look-alike control program for four weeks⁵⁶. EndeavorRx players improved more on a computer test of attention⁵. But on seven rating scales filled out by parents and clinicians, the groups didn’t differ in a statistically meaningful way⁵.

The label warns that children may not show benefits in everyday behavior such as hyperactivity⁵. Side effects were mild: about 3% of players reported frustration and 2% reported headaches⁵⁶. The maker, Akili Interactive Labs, paid for the trial⁶.

Rejoyn, for depression, works differently. Adults who already take an antidepressant do 18 short memory exercises using pictures of emotional faces, plus 18 brief CBT lessons, over six weeks⁷.

Its trial assigned 386 adults to Rejoyn or a sham app with a shapes memory task⁷. On the main depression scale, Rejoyn users improved by about 9 points and sham users by about 7⁷. The main analysis needed a p-value, a common measure of chance, below 0.049. It came in at 0.0568⁷.

Other results leaned the same way. A broader analysis that counted every randomized participant showed a gap of about 2 points that did reach significance, and so did a patient questionnaire and a clinician’s overall rating⁷. An anxiety scale did not⁷. John Torous of Beth Israel Deaconess Medical Center in Boston, who wasn’t involved in developing Rejoyn, told CNN the research was still exploratory⁸.

DaylightRx, cleared in September 2024 for adults with generalized anxiety disorder, teaches CBT skills over 90 days⁹. Its trial of 256 people found a large drop in anxiety. But the comparison group was a waitlist, people who simply hadn’t started yet⁹. When an FDA advisory committee discussed testing AI mental health devices in 2025, it said waitlist comparisons may not be ideal for such studies¹⁰.

How the evidence stacks up, product by product

An unblinded trial is one where people know which group they are in, which can make results look better.

Product	Use	What the main trials found	Evidence level
Somryst and other CBT-I programs	Chronic insomnia in adults	Two trials, the larger with 1,149 people; benefits lasted 12 months4	Strong evidence
reSET	Substance use when the main drug isn’t an opioid	One unblinded trial of 507 people, with rewards added3	Promising but early
reSET	People whose main drug is an opioid	Did not help in the trial3	Evidence says no for this group
EndeavorRx	ADHD, ages 8 to 12	Attention test improved; parent and clinician ratings didn’t differ5	Promising but early
Rejoyn	Depression, with an antidepressant	Main analysis missed its target; a broader analysis and two other scales showed gains7	Promising but early
DaylightRx	Generalized anxiety	One trial against a waitlist9	Promising but early
AI chatbot “therapists”	Depression and other conditions	None authorized by FDA as of November 20252	Not enough evidence

A cleared app can still disappear, which is why Medicare’s move matters

FDA review checks the evidence. It doesn’t keep a company in business.

Pear Therapeutics, maker of reSET, its opioid version reSET-O and Somryst, filed for Chapter 11 bankruptcy in April 2023¹¹. It stopped taking new prescriptions and stopped filling refills¹¹. Its chief executive, Corey McCann, blamed insurers for denying payment¹¹.

The company had lost $30.7 million in one quarter of 2022 on $4.1 million in revenue¹². At auction, its products and patents sold for about $6 million in total¹². Nox Health paid $3.9 million for Somryst, and Harvest BIO bought the reSET programs along with other assets¹².

For patients, the lesson is practical. A prescription app is a service that can end. Ask how long you’ll have access, whether you can save your progress and what happens to your data if the company is sold. Our article on health data privacy covers those questions.

Payment was the problem Pear’s chief executive named. Medicare took a step toward paying for these apps in its 2025 physician fee schedule, the yearly rule that sets what it pays clinicians¹³.

Starting January 1, 2025, Medicare created three billing codes for “digital mental health treatment devices”¹³¹⁴. One covers supplying the device and teaching the patient to use it. The other two cover a clinician’s time managing that treatment each month, in 20-minute blocks¹⁴.

The conditions matter. The device must be cleared or authorized by FDA in the computerized behavioral therapy category that reSET created¹³¹⁵. It must be used alongside ongoing behavioral health care under a treatment plan, not handed off on its own¹³. The clinician who bills also bears the cost of supplying the device¹⁵.

A sleep medicine group asked Medicare to widen the codes, including for devices that don’t send data to clinicians automatically, and Medicare declined¹⁴.

For 2026, Medicare extended payment to devices in FDA’s separate ADHD category, under the same rules¹⁵.

No AI chatbot therapist has FDA authorization yet

Chatbots built on large language models, the technology behind tools such as ChatGPT, can hold a conversation about your mood. None is an FDA-authorized mental health treatment².

In November 2025, FDA’s Digital Health Advisory Committee met to discuss how such a device should be judged¹⁰. Members worked through a made-up example: a prescription chatbot therapist for adults with depression¹⁰.

They listed risks: invented answers, missed warning signs a human would catch, worsening symptoms including self-harm and unhealthy attachment to the bot¹⁰. They called for independent, high-quality trials of real length, a clinician’s evaluation before prescribing and a one-tap way to reach emergency help¹⁰. Labels, they said, should tell users plainly that they are dealing with software, not a human therapist¹⁰.

If a chatbot app claims to treat depression today, it hasn’t passed that kind of review. Our article on AI chatbots for health explains what they do well and where they fall short.

Skills, not pills, with a clinician still in the loop

For readers who prefer to limit medication, these programs have real appeal. Most teach skills, the same CBT methods therapists use, and trials have reported few side effects⁴⁶.

But most are built as add-ons. The labels for reSET, EndeavorRx and Rejoyn say they are not standalone treatments and don’t replace prescribed medicine³⁵⁷. Rejoyn is only for people already taking an antidepressant, and it doesn’t track your symptoms or alert your prescriber if you get worse⁷.

How to tell them apart, and what to ask

Question	Wellness app	Digital therapeutic
Who checks the claims?	Usually not FDA2	FDA, before it’s sold1
How you get it	App store	Prescription from a clinician2
What it can promise	Healthy habits, not disease treatment2	Treatment of a named condition2
Evidence required	None by law	Clinical data showing the therapy works as designed3
Who pays	Usually you	Sometimes Medicare, with a clinician involved1315

If a clinician suggests one, bring these questions:

1. “What did the main trial compare it with: a sham app, usual care or a waitlist?”
2. “Did it improve how people felt and functioned, or only a test score?”
3. “Is this added to my current treatment, or does it replace part of it?”
4. “Who sees my progress, and what happens if my symptoms get worse?”
5. “How long can I use it, what will it cost me and what if the company closes?”
6. “Is the same therapy available in person or by phone if the app doesn’t suit me?”

If you’re thinking about suicide or self-harm, call or text 988, or call 911. These apps aren’t crisis services, and some don’t alert anyone if you get worse⁷. You can report a problem with an FDA-cleared app through FDA’s MedWatch program.

Questions readers ask

The bottom line

Digital therapeutics are a real step forward: software that has to show FDA evidence before it can claim to treat anything. Online CBT for insomnia has the best record. Apps for ADHD, depression, anxiety and substance use have thinner or more mixed evidence, and no AI chatbot has cleared the bar.

Treat a prescription app like any other treatment. Ask what the trial compared it with, whether it changed daily life, how it fits with the care you already have and what happens if the company goes away.

Related reading

• Wellness apps help a little if you keep using them, and most people don’t
• “FDA registered” isn’t “FDA approved”: a buyer’s map to wellness devices
• AI chatbots pass medical exams, yet people using them don’t diagnose better
• Health apps share more than you think, and you can lock yours down
• Behavior science tested its own tricks at scale, and rewards and repetition held up

Sources

1. U.S. Food and Drug Administration. “De Novo Classification Request.” FDA, 2025. fda.gov
2. U.S. Food and Drug Administration. “Generative Artificial Intelligence-Enabled Digital Mental Health Medical Devices: Executive Summary for the Digital Health Advisory Committee.” FDA, November 2025. fda.gov
3. U.S. Food and Drug Administration. “De Novo Classification Request for reSET (DEN160018): Decision Summary.” FDA, 2017. accessdata.fda.gov
4. U.S. Food and Drug Administration. “510(k) Summary, Somryst (K191716).” FDA, 2020. accessdata.fda.gov
5. U.S. Food and Drug Administration. “De Novo Classification Request for EndeavorRx (DEN200026): Decision Summary.” FDA, 2020. accessdata.fda.gov
6. Kollins SH, DeLoss DJ, Cañadas E, et al. “A novel digital intervention for actively reducing severity of paediatric ADHD (STARS-ADHD): a randomised controlled trial.” The Lancet Digital Health, 2020. doi.org
7. U.S. Food and Drug Administration. “510(k) Summary, Rejoyn (K231209).” FDA, 2024. accessdata.fda.gov
8. Cheng M. “FDA clears first digital treatment for depression, but experts caution that research is still early.” CNN, via ABC7 New York, April 2024. abc7ny.com
9. Kuntz L. “First Digital Treatment for Generalized Anxiety Disorder Cleared by the FDA.” Psychiatric Times, September 2024. psychiatrictimes.com
10. U.S. Food and Drug Administration. “Digital Health Advisory Committee Meeting, November 6, 2025: 24-Hour Summary.” FDA, November 2025. fda.gov
11. Comstock J. “Pear Tx files for bankruptcy, will seek to sell assets.” Pharmaphorum, April 2023. pharmaphorum.com
12. Hagen J. “Pear Therapeutics assets sold for over $6M at auction after bankruptcy.” MobiHealthNews, May 2023. mobihealthnews.com
13. Centers for Medicare & Medicaid Services. “Calendar Year (CY) 2025 Medicare Physician Fee Schedule Final Rule.” CMS fact sheet, November 2024. cms.gov
14. American Academy of Sleep Medicine. “AASM analysis of the 2025 Medicare physician fee schedule final rule.” AASM, November 2024. aasm.org
15. Centers for Medicare & Medicaid Services. “Medicare Physician Fee Schedule Final Rule Summary: CY 2026 (MLN Matters MM14315).” CMS, December 2025. cms.gov