The best apps for ADHD adults who take forever to reply

Replying breaks at five separate points, so there is no single best app for an ADHD adult who takes forever to reply. The message has to survive in memory until you have a moment. You have to open it and work out what it is asking. You have to compose something. You have to decide the composed thing will land the way you mean it. And you have to start. Different people get stuck at different points along that chain, and an app built for one point does nothing for the others. A reminder app will not write the reply, and a rewriting tool will not bring back a message that dropped out of your inbox. The apps below that fit each step are Gmail snooze or Apple Mail Remind Me for a lost message, Gmail’s AI Overview, WhatsApp transcripts or Subtext for a long message or voice note, Goblin Tools or Subtext for drafting and tone checks, and Focusmate or a phone timer for starting.

So this list is sorted by the step that stops you, with the apps that address each step. Subtext is the app I co-founded, and it appears under three of the five steps, drafting, reading and tone, so I have a stake in how those three sections come out. I describe its neighbours in their own words, from their own pages, the way I did for venting apps. Prices are what I saw on the product’s own site or store listing on 4 and 5 October 2026, and they move.

If you want the research on why the reply is hard before the shopping list, I wrote that up separately in why I can’t reply to messages I want to answer. This piece answers the next question, which app to use.

The research on avoiding a reply, forgetting it and holding it in mind

Avoiding a message can be a way of managing a feeling. Sirois and Pychyl argue that procrastination has a great deal to do with repairing your mood in the short term¹. The message provokes something, you look away, the feeling eases, and you pay for it later. That paper is an argument summarising earlier research on procrastination in general, not messages. The evidence specific to ADHD is small, and my write-up of it lays it out with its caveats. Oguchi and colleagues surveyed 490 adults from a Japanese online panel in 2019 and found a d = 0.83 gap on procrastination between groups with higher and lower ADHD symptoms. They screened by self-report, with no diagnostic assessment². In 2014, Niermann and Scheres reported that among 54 undergraduates at Radboud University, inattention went with general procrastination after hyperactivity and impulsivity were set aside statistically, but not with academic procrastination³.

Two more findings fit specific steps of a reply. Altgassen, Scheres and Edel compared 29 adults with ADHD to 24 controls on prospective memory, which is the memory for intentions you have not acted on yet. The ADHD group recalled and carried out fewer of their own intentions from real life, and that gap accounted for part of the link between ADHD symptoms and procrastination⁴. “I’ll reply tonight” is a prospective memory task, and this is the finding behind the reminders section. Alderson and colleagues pooled 38 studies of working memory in adults with ADHD and found moderate differences from controls in both phonological and visuospatial working memory⁵; Pievsky and McGrath’s review puts that beside the other cognitive domains⁶. Holding a message with four questions while composing an answer to it is a working memory task, and that is the finding behind the summarising section.

What the research says about composing and starting a reply

Composing and starting are the two steps without a tidy study of ADHD and replying, and I have said where the evidence runs out. The closest is a 2013 laboratory task in Germany, in which 45 clinically diagnosed adults with ADHD and 45 matched controls did a complex prospective memory task. The largest difference was in planning (d = 1.60), and remembering to start a task unprompted did not differ significantly, though groups of 45 cannot show there is no gap⁷. One finding is not specific to ADHD. Sarrafzadeh and colleagues, mostly at Microsoft, read Microsoft’s own Outlook logs from about 40,000 enterprise users in May 2018. They found that people defer the emails that need a reply or careful reading, and that deferral rose with workload⁸. There was no ADHD sample, and an email counted as deferred when it was read in one session and replied to or forwarded in a later one.

Reminders and snooze for lost messages

The usual fix is to make the message come back on its own, because remembering to find it is a prospective memory task, the one Altgassen’s group found impaired⁴. I found one randomised test of reminders in adults with ADHD. In an online programme that 109 Norwegian adults did alone in 2020, it randomised each module, for people who had not finished within two days, to one extra SMS reminder or none. It did not improve completion or logins overall and sped up the first login in a few modules⁹; it was not about replying. Marking a message unread or flagging it is a common way to defer email, and a weak guarantee. In Sarrafzadeh and colleagues’ Outlook logs, assuming that every message marked unread or flagged would later be replied to or forwarded was right about one time in seven⁸. The logs counted only a later reply, reply-all or forward, so a message dealt with by reading, archiving or a call counts as a miss. Enterprise users in 2018, no ADHD sample.

Gmail snooze is already in Gmail. Google says “When you snooze an email, the email is removed from your inbox temporarily” and it comes back “to the top of your inbox when you want it to”¹⁰. It snoozes nothing outside Gmail, and it is only as good as the time you pick.

Apple Mail Remind Me is the same idea for Apple’s Mail app. Apple’s iCloud.com guide says you “can set a time and date to receive a reminder and the email will be moved to the top of your inbox”¹¹. On iPhone, Apple’s guide says to tap Reply, then Remind Me, on a Primary message¹². I could not open the Mac steps. Email only.

Task and nag apps for messages outside email

Email snooze does nothing for a message that arrived in WhatsApp or a text.

Todoist is for the messages that live outside email. You add “reply to Anna” as a task with a reminder, and the task list becomes the inbox you actually check. The free Beginner plan includes task reminders and five personal projects, which is plenty for a reply list¹³. Pro, which adds custom task reminders and more projects, is $7 a month or $60 a year on Todoist’s site and was $6.99 a month or $59.99 a year in the US App Store listing¹³. Todoist does not read your messages, so writing the task down is still your job.

Due is the reminder that keeps coming back. Its own pitch is “Due remembers everything you need so you don’t have to,” and its Auto Snooze “repeatedly notifies you of overdue reminders until you mark them complete, reschedule them, or turn off their auto snooze”¹⁴. On iOS it repeats 5 times by default, up to 10, and carries on past that only after you open Due or act on a notification. It was $7.99 in the US App Store, with a $4.99 Annual Upgrade Pass as an in-app purchase, for iPhone, iPad and Apple Watch; the Mac version is sold separately. Apple only, and it cannot tell a task you have a plan for from one you do not.

Drafting and tone tools for the blank page

Subtext reads a draft, a rough note or an instruction such as “say no politely”. On a draft it marks the words at fault, names the issue and says why, then offers up to three versions, each with a safe-to-send score for how likely it is to land as you meant. You can use it in 17+ languages, and the input can be a long, unedited draft you do not have to tidy first, which Subtext says on its page for neurodivergent adults. Your text goes to Subtext’s servers and to the providers its privacy policy names, and nothing you submit is used to train an AI model¹⁵. It runs on iPhone, Android and the web. A few analyses are free to start, after which Subtext is a paid app. I could not find an evaluation of Subtext with ADHD users by anyone outside the company, so this is what the app shows and what its prompts and privacy policy say it does. It will not make you open the message, but it can write the first version for you, which is the step some people avoid.

Goblin Tools Formalizer for drafting on a budget

Goblin Tools Formalizer is the one I would send most ADHD readers to first if money is the constraint. Goblin Tools describes itself as “a collection of small, simple tools, for when things feel too big or complicated,” and the web version is “kept free by our sponsors”¹⁶. The Formalizer takes your text and rewrites it in a chosen style while keeping the content: options include More professional, More polite, Less snarky, More to the point, Less emotional and Easier to read, with a chilli scale for how far to push it. The iOS app was $1.99 up front, with a Pro subscription at $3.99 a month or $39.99 a year for history, sync and sharing, and there is an Android app. It gives no reason for what it changed, and each press produces a new answer. On privacy, Goblin Tools’ policy says that without a Pro account nothing you type is stored anywhere but your own browser, and that requests to the tools built on OpenAI’s technology go to OpenAI’s servers and are not stored on Goblin Tools’ servers¹⁷. It points to OpenAI’s policy for the rest and does not say whether OpenAI trains on what it receives, so apply the test in the piece on writing assistants that train on your messages before you paste a private message.

Grammarly for drafting, and what the research on drafting tools shows

Grammarly is the mainstream option, and it is already in a lot of inboxes. The free plan includes “See your writing tone” alongside spelling and grammar; the Pro plan, which lets you “Adjust your writing tone with ease,” was shown to me at €12 a month with a free trial of seven days¹⁸, a euro price as shown in Europe, so it will not match the dollar figures in the big tech comparison. It runs inside Gmail, Outlook, Slack and the browsers, which is its real advantage for someone who will not open another app. It is built for correctness first and tone second, and it works mainly on text you have already written. Its training setting is on by default for individual accounts, so check it before you use it on private messages.

The research on this step is thin. I found no trial of drafting tools with ADHD adults. Composing is the kind of effortful task people defer⁸, and removing it may make replying easier, though no study I read tests whether it does. For the comparison with Grammarly on personal messages, see Subtext vs Grammarly for personal messages.

Summaries and transcripts for reading the message first

Gmail and Subtext can shorten the reading step, and WhatsApp’s transcripts make a voice note skimmable. That matters because a long message asks you to hold its contents in mind while you work out an answer, and working memory is one of the places adults with ADHD differ from controls on average⁵. A voice note is worse, because you cannot skim it and tend to replay it. Once the reading is shorter, composing can start.

Gmail’s AI Overview summarises a thread in place. Google says it will “Generate an AI Overview at the top of the thread that synthesizes key points and replies in the conversation,” triggered by clicking “Summarize this email”¹⁹. Google says it is available to all Gmail users once smart features are turned on, in some languages. Google also warns that Gemini features “may suggest inaccurate or inappropriate information”, so open the thread before you act on the overview¹⁹. That page does not say how Google handles the thread text, so I cannot tell you, and I would check before using it on private threads. It only covers Gmail, and turning smart features on means accepting the rest of that bundle.

WhatsApp voice message transcripts are built into WhatsApp and run on the phone. In WhatsApp’s words, “Transcripts are generated on your device so that no one else, not even WhatsApp, can hear or read your personal messages”²⁰. Turn them on under Settings, Chats, Voice message transcripts, then long press a voice note and tap transcribe. A transcript is still the whole message, so a voice note seven minutes long becomes a long wall of text. It transcribes and does not summarise. Transcripts launched in a few select languages, and WhatsApp keeps a list of the supported ones²⁰.

How Subtext summarises a message you received

Subtext summarises a received message as well as checking your own. Paste a screenshot of the thread or drop in the voice note, and it gives you one line on what the other person wants from you, then two to five short bullets with the action items, open questions, deadlines and decisions, written in the language the message arrived in. For a long thread, a summary can take over some of the work of holding four questions in mind, which is the working memory load described above. When you ask for a reply, it drafts one that picks up where they left off, and if the message asks you to choose something you have not answered, it asks you first. It does not label the sender’s tone, and no tool can tell you what the other person privately feels; it only has their words to go on, and I have written about how wide that gap is. The workflow for a message you received is on the page about reading the message you received before you reply, and turning your own voice note into a written message is the opposite direction, when you are the one speaking.

Tone checks for the fear of how it lands

Subtext is an AI writing assistant and tone checker for text messages, and tone is the step this section is about. The fear that a reply will read as blunt or cold is the fear behind a lot of unsent drafts. The perfectionism research I went through for my piece on why replies stall says what stalls people is concern about being judged, not high standards. A tone check that points to a specific sentence gives you something to fix. Subtext runs that check on your own draft, not on what the other person wrote. On an iPhone, Apple’s Writing Tools rewrite a draft on request and do not report how it reads.

Goblin Tools Judge is the free alternative, and its framing is good. “The Judge tries to figure out what the emotional content of the text is. Does it sound frustrated, rude, happy, or inquisitive? How does it come across to a neutral party?”¹⁶. You can add follow-up messages to see how a conversation’s tone develops. The page I read describes a verdict and does not mention marking which words produced it, so fixing the draft is a second step with the Formalizer.

Grammarly’s tone detector and the limits of every tone check

Grammarly’s tone detector “instantly identifies the tone of your writing by analyzing your word choice, phrasing, punctuation, and even capitalization,” and it is on the free plan¹⁸. On the free plan it shows you a label while you type inside the apps you already use, such as “confident” or “worried”, and Pro adds a way to adjust the tone.

A caution that applies to all three, including mine. No tone tool can tell you how one specific person on a specific day will read your words. They can catch some of the readings a neutral stranger would have, which may be part of what people are worrying about when they reread a text of two lines for the ninth time. If the rereading itself is the loop, the piece on why a short reply takes an hour is about that, and the longer piece on AuDHD and communication covers what the research does and does not support. The wider field of tools is compared in best AI tone checkers.

Body doubling for starting

Body doubling is a documented practice, and the experiments I know of are small and none is about replying. Eagle, Baltaxe-Admony and Ringland surveyed 220 people about a practice they describe as used primarily by neurodivergent people, and report the community’s own definition as “using the presence of others to start, stay focused on, or accomplish a task”²¹. The survey covers use and motivation but no outcomes, so it cannot say whether it works. In Ara and colleagues’ 2025 preprint, 12 adults recruited as having ADHD placed more bricks per minute in a virtual reality task with a human body double, or an “AI” one operated by a human, than alone, with no significant gain in accuracy. They also cite two earlier studies that found no significant effect or no behavioural difference, a 2024 master’s thesis with ADHD participants and one with neurotypical adults²². Schuenke and colleagues’ 2025 EEG study of 26 people with and without ADHD found no statistically significant difference between working with and without a body double²³.

Focusmate is a body doubling service. It calls itself “Virtual coworking for getting anything done” and describes body doubling as “working on any task with another person present, without them participating in your task”²⁴. You book a session of 25, 50 or 75 minutes, join a video call with another member, say what you will do, and do it. The free plan gives three sessions a week with no card; Plus, with unlimited sessions, was $8 a month billed annually or $12 billed monthly. A session of 25 minutes called “reply to the six messages” is the most direct use here. A stranger on video can add pressure.

I left out Flown, the other service people name, because I could not load its pricing.

Timers and a rule for when to start

Timers are a cheaper version of body doubling. Set any phone timer for ten minutes and attach the rule “I only have to open the message, I do not have to answer it.” A rule like that is called an implementation intention, and it is the only tool on this page I found a meta-analysis for. That meta-analysis covers other goals, in samples without ADHD, and none of it is about messages. Gollwitzer and Sheeran pooled 94 tests in 2006, with 8,461 participants, mostly university students. They found an effect between medium and large on goal attainment, d = .65, and the effect held for starting a goal too, at d = .61²⁵. The format is “if it is 7pm, then I open Anna’s message.” It is free. What that meta-analysis does not tell you is whether a timer set to ten minutes does the same job, or whether the rule works on a message you dread, because it was not about either. The cue still has to be something you notice, and a timer is easy to notice.

Which app to try at each step, with prices

What none of these apps do

No app here decides for you or sends for you, Subtext included, and the ones that come closest to doing the replying are the ones I would trust least. Most tools cover one step. A snooze brings the message back, a summariser reads it, a drafting tool writes the reply, and none of them sends it for you, which is how it should be.

The one thing I would add to any combination is a holding message, which no app sells. “Saw this, proper reply at the weekend” takes ten seconds, and the reasoning in my research on why replies stall is that it probably works even though nobody has tested it on correspondence. Telling them you are short of time changes how they explain your silence. The person waiting is unlikely to read your delay as overload; the attribution research there says they read it as indifference, and a short message gives them another explanation.

If I had to pick one tool per step for an ADHD adult on a budget, I’d take Gmail snooze or Apple Mail Remind Me for lost messages, Goblin Tools for drafting and tone, WhatsApp’s own transcripts to make voice notes skimmable, and a phone timer with a rule attached for starting. Then, if the drafting step is still where you stall, try Subtext.

Subtext helps when you know you should reply but have nothing you are willing to send. Paste the message you received and Subtext tells you what they want. Paste your draft and it shows you how it could land and offers up to three versions to choose from. Try Subtext in your browserGet the appTry Subtext in your browser

Sources

Numbered in order of first appearance. Pages were checked on 4 and 5 October 2026, and I reopened three of them on 10 October 2026: the Gmail AI Overview help page, Goblin Tools and Subtext’s privacy policy. Preprints are marked as such, and where I could read only an abstract, the entry says so.

1. Sirois, F. M., and Pychyl, T. (2013). Procrastination and the priority of short-term mood regulation: Consequences for future self. Social and Personality Psychology Compass, 7(2), 115 to 127. Argument paper summarising earlier research, per its abstract (the publisher page would not load).
2. Oguchi, M., Takahashi, T., Nitta, Y., and Kumano, H. (2021). The moderating effect of attention-deficit hyperactivity disorder symptoms on the relationship between procrastination and internalizing symptoms in the general adult population. Frontiers in Psychology, 12. Online panel of adults in Japan, February to March 2019; 490 in the Methods and tables (the abstract says 470), 248 in the higher-symptom group and 242 in the lower. Screening cutoff, cross-sectional, self-report, no diagnostic assessment; the procrastination finding was incidental to the study’s hypothesis. Funded by the Ibuka fund and a JSPS KAKENHI grant; the authors declared no conflicts of interest.
3. Niermann, H. C. M., and Scheres, A. (2014). The relation between procrastination and symptoms of attention-deficit hyperactivity disorder (ADHD) in undergraduate students. International Journal of Methods in Psychiatric Research, 23(4), 411 to 421. 54 Radboud University students picked for high, medium or low self-reported symptoms, no diagnoses; cross-sectional, self-report. Inattention and the Pure Procrastination Scale, r = .51, partial r = .43 controlling for hyperactivity and impulsivity; academic procrastination and observed delay on a maths task were not significantly related. Supported by the Radboud honours programme and the Stichting Nijmeegs Universiteitsfonds; no competing-interest statement found.
4. Altgassen, M., Scheres, A., and Edel, M.-A. (2019). Prospective memory (partially) mediates the link between ADHD symptoms and procrastination. ADHD Attention Deficit and Hyperactivity Disorders, 11, 59 to 71. 29 adults with ADHD, 24 controls.
5. Alderson, R. M., Kasper, L. J., Hudec, K. L., and Patros, C. H. G. (2013). Attention-deficit/hyperactivity disorder (ADHD) and working memory in adults: A meta-analytic review. Neuropsychology, 27(3), 287 to 302. 38 studies.
6. Pievsky, M. A., and McGrath, R. E. (2018). The neurocognitive profile of attention-deficit/hyperactivity disorder: A review of meta-analyses. Archives of Clinical Neuropsychology, 33(2), 143 to 157.
7. Fuermaier, A. B. M., et al. (2013). Complex prospective memory in adults with attention deficit hyperactivity disorder. PLOS ONE, 6 March 2013. Laboratory task; 45 adults with ADHD diagnosed by clinicians and not taking stimulants, 45 matched controls, Germany. Planning d = 1.60; self-initiation 40% of the ADHD group against 56% of controls succeeding, not significant, and not a messaging task. The authors reported no funding and no competing interests.
8. Sarrafzadeh, B., Hassan Awadallah, A., Lin, C. H., Lee, C.-J., Shokouhi, M., and Dumais, S. T. (2019). Characterizing and predicting email deferral behavior. Proceedings of the Twelfth ACM International Conference on Web Search and Data Mining (WSDM 2019). Interviews with 15 Microsoft employees and logs of about 40,000 enterprise Outlook web users, 6 to 19 May 2018; five of the six authors at Microsoft; no ADHD sample. Deferral means a reply, reply-all or forward in a later session than the first read; deferral probability rose with workload, from about 1.5% with no unhandled messages to about 11% with more than 20 unhandled (their Figure 1, read off the chart), and, using calendar meetings as a second proxy, from about 3% to 4% as meetings went from 1 to 5; a rule predicting a reply or forward for every message marked unread or flagged had precision 0.14. The ACM page would not load, so I read the authors’ preprint at . https://doi.org/10.1145/3289600.3291028
9. Nordby, E., et al. (2022). The effect of SMS reminders on adherence in a self-guided internet-delivered intervention for adults with ADHD. Frontiers in Digital Health, 16 May 2022. Multiple randomised trial; 109 adults in Norway with a self-reported ADHD diagnosis confirmed at interview, recruited in May 2020; extra SMS reminder against the usual new-content SMS, randomised per module among people who had not finished within two days. The extra reminder sped up login in some modules, with no effect on completion, number of logins or practice of coping strategies. Funded through the INTROMAT project of the Research Council of Norway; the first author’s doctoral funding came from Western Norway Regional Health Authorities (Helse Vest); no competing interests declared.
10. Google. Snooze emails until later. Gmail Help.
11. Apple. Receive a reminder to reply to email in Mail on iCloud.com. iCloud User Guide.
12. Apple. Check your email in Mail on iPhone, section on using Remind Me to come back to a Primary message later. iPhone User Guide.
13. Todoist. Pricing. https://www.todoist.com/pricing. Pro prices in the App Store from the US listing. https://apps.apple.com/us/app/todoist-to-do-list-planner/id572688855. Pro prices on the web from Todoist Help, Todoist Pro pricing update in 2025. https://www.todoist.com/help/account-and-billing/plans/todoist-pro-pricing-update-in-2025-bxBvHZuJZ
14. Due Apps. Due: Superfast Reminder App. https://www.dueapp.com/. Price and Auto Snooze repeat limits from the US App Store listing. https://apps.apple.com/us/app/due-reminders-timers/id390017969
15. Subtext, Terms, privacy and your account. Privacy policy last updated 26 July 2026. Checked 10 October 2026.
16. Goblin Tools. https://goblin.tools/, the Formalizer at https://goblin.tools/Formalizer and the Judge at https://goblin.tools/Judge. App price from the US App Store listing. https://apps.apple.com/us/app/goblin-tools/id6449003064
17. Goblin Tools. Privacy Policy. . Opened 10 October 2026; the page shows no date, only the version v3.14.1.
18. Grammarly. Plans. https://www.grammarly.com/plans. Tone detector. https://www.grammarly.com/tone
19. Google. Catch up on email threads with AI Overview conversation summaries. Gmail Help.
20. WhatsApp (21 November 2024). Introducing voice message transcripts. . The settings path and the long-press step to transcribe are quoted from this blog post. The WhatsApp Help Center page it links to, How to turn voice message transcripts on or off, would not show its content without JavaScript, so the list of supported languages is not quoted from it.
21. Eagle, T., Baltaxe-Admony, L. B., and Ringland, K. E. (2024). “It was something I naturally found worked and heard about later”: An investigation of body doubling with neurodivergent participants. ACM Transactions on Accessible Computing, 17(3), 1 to 30. Online survey of 220 people; descriptive, self-report, no control condition. I could not open the full text, so I quote only the abstract’s figures and definition.
22. Ara, et al. (2025). You are not alone: designing body doubling for ADHD in virtual reality. arXiv preprint 2509.12153, version 1, 15 September 2025, no venue stated. 12 adults recruited as having ADHD by survey and a self-report scale, within-subjects, order counterbalanced; bricks placed per minute about 8.5 alone, 10.8 with a human double and 11.1 with the “AI” double, which was operated by a human; accuracy did not differ significantly. No funding or competing-interest statement found.
23. Schuenke, R., Dickenson, and Moore (2025). Reading between the lines: exploring body doubling in ADHD using EEG. Proceedings of ACM SIGACCESS ASSETS ’25, 1 to 6. 26 participants with and without ADHD, reading comprehension tests with and without a body double; I could read only the abstract; Crossref lists no funder.
24. Focusmate. https://www.focusmate.com/ and pricing at https://www.focusmate.com/pricing
25. Gollwitzer, P. M., and Sheeran, P. (2006). Implementation intentions and goal achievement: A meta-analysis of effects and processes. Advances in Experimental Social Psychology, 38, 69 to 119. 94 independent tests, 8,461 participants, d = .65 (95% confidence interval .60 to .70); 79 tests used university students; goals such as health and study; no ADHD sample; initiation of goal striving d = .61 across 52 tests; 49% of tests unpublished, with no difference from published ones. Read in an open-access copy of the full text.