People are uneasy about AI notetakers right now. An AP story on meeting-recording discomfort ran in dozens of outlets last week. So we read 663 recent App Store reviews across 8 apps that record, transcribe, or coach conversations. Three things stood out. First, the anger is mostly about billing, not AI: 38% of recent 1–2 star reviews mention subscriptions, trials, or surprise charges. Second, permission barely comes up: 2 mentions in 200 notetaker reviews, both from the recorder’s side. Third, not one reviewer says a notetaker made them better at talking. The tools remember your conversations. They don’t change them.
Here’s the method, so you can judge the numbers. On July 14, 2026 we pulled the most-recent public US App Store reviews for eight apps. The set: Otter and Plaud (notetakers), Vocal Image and CloserCoach (AI speaking/sales coaches), Siro and Rilla (team tools that coach from real sales calls), and Cloze and Contacts Journal (client-notes apps). We counted themes with keyword matching, then read every match by hand. Full counts are below. Review windows differ by app. Otter’s last 100 written reviews span five weeks; Cloze’s span years. So treat the percentages as "what recent writers said," not lifetime scores.
The recording itself works. That’s not the problem
If you only read the news, you’d expect the reviews to be full of dread. They aren’t. Otter’s last 100 written reviews are 76% five-star. Plaud’s are 72%. People describe real relief:
"As someone with pretty severe ADHD, one of the hardest parts of life has always been holding onto important conversations, ideas, and breakthrough moments long enough to actually do something with them." from a five-star Plaud review titled "Life Changing" (July 2026)
"This app is saving my writing career." the title of a five-star Otter review from a disabled writer (June 2026)
The capture layer (record, transcribe, summarize) is a solved problem people happily pay for. The discomfort in that AP story lives somewhere else: in the room, between people. Which brings us to what reviewers don’t say.
The #1 complaint isn’t AI. It’s billing.
Across all eight apps, we found 170 recent one- and two-star reviews. 65 of them, or 38%, mention subscriptions, free trials, charges, or refunds. Not accuracy. Not privacy. Money.
It’s worst in the AI coaching category, where 54% of low-star reviews (52 of 97) are about billing. The language is not mild:
"Congratulations! You successfully unlocked a new level of predatory subscription!" one-star Vocal Image review (July 2026)
"SCAM ALERT!!! DON’T FALL FOR THE FREE TRIAL! … IT CHARGES YOU IMMEDIATELY" one-star CloserCoach review (July 2026)
"Charged me for a yearly subscription and wouldn’t refund" the title of a one-star Otter review (June 2026)
This also explains something strange you’ll see on the store. The front-of-store rating and the recent written reviews can tell two different stories. Vocal Image shows 4.5 stars across 10,048 ratings. Yet 62 of its 100 most-recent written reviews are one-star. CloserCoach shows 4.7; 44% of its recent written reviews are one-star. Part of that gap is normal selection bias: angry people write, happy people tap a star. But read the reviews and the pattern is specific: people who feel tricked by a trial write, in detail, every time.
| App | What it is | Lifetime ★ | Recent written reviews (window) | 1★ share of recent |
|---|---|---|---|---|
| Otter | notetaker | — | 100 (Jun 5 – Jul 12, 2026) | 6% |
| Plaud | notetaker + device | 4.9 (20,373) | 100 (Mar 24 – Jul 11, 2026) | 10% |
| Vocal Image | AI speaking coach | 4.5 (10,048) | 100 (Dec 24, 2025 – Jul 11, 2026) | 62% |
| CloserCoach | AI sales roleplay | 4.7 (689) | 34 (Mar 2025 – Jul 2026) | 44% |
| Siro | real-call coaching (teams) | 4.8 (1,304) | 76 (2021 – Jul 2026) | 7% |
| Rilla | real-call coaching (teams) | 4.5 (779) | 53 (2023 – Jun 2026) | 17% |
| Cloze | relationship CRM | 4.6 (2,729) | 100 (2018 – Mar 2026) | 21% |
| Contacts Journal | client notes, one-time purchase | 4.7 (2,332) | 100 (Sep 2025 – Jul 2026) | 2% |
Counts as of July 14, 2026. Lifetime ratings shown where banked in our dataset; percentages are within each app’s recent written-review window.
One more data point from the table. The only app with almost no billing anger, Contacts Journal, is also the only one sold as a one-time purchase. Its recent reviews are 84% five-star, and zero of its five low-star reviews mention billing. The pattern writes itself: the anger tracks the billing model, not the AI. We later sorted this same category by what each app actually does — see what an AI communication coach actually is for the rehearsal-vs-notes-vs-real-call breakdown.
Almost nobody talks about permission. That’s the blind spot
Here’s the number that surprised us most. In 200 recent reviews of the two big notetakers, the topic of permission to record comes up twice. Both times it’s the person doing the recording, and both are five-star reviews:
"I primarily use Otter during interviews, always with the candidate’s permission." five-star Otter review (July 2026)
"I’ve had to get some legal backup to convince places like my HOA board to allow me to record and transcribe…" five-star Otter review (June 2026)
Zero reviews raise consent as a concern. Outside the reviews, it’s a different story. Otter faces federal class actions that allege it recorded people without everyone’s consent. And last week’s news cycle was about people feeling ambushed by bots in meetings. The people buying recorders aren’t writing about the people being recorded. The reviews are the receipts.
That’s the real "AI notetaker problem" in 2026. Not transcription quality. The social layer: whether the other person knows, and how it feels. It’s a legal layer too. About a dozen US states require everyone on a call to consent to recording. (General information, not legal advice; state laws vary.) We’ve published a plain-English consent guide with word-for-word scripts as the follow-up to this study, and a closer look at what this means specifically for client calls, including a fresh 2026 survey on how often people actually get asked.
Zero of 200 notetaker reviews say the tool made them better at talking
We ran a deliberately wide net over the Otter and Plaud reviews: any mention of improving, learning, practicing, speaking, confidence, feedback, delivery, pace, or tone. Twenty-eight reviews matched. We read all of them. Every single one is about transcript quality, speaker identification, the app improving, or capturing thoughts more easily. Not one says the reviewer got better at speaking or communicating.
That’s not a knock on notetakers. It’s just not the job they do. A transcript is a record of how the conversation went. It doesn’t change how the next one goes. If your conversations have a problem, a perfect transcript preserves the problem in high fidelity. You ramble? It keeps every ramble. You forget what matters to people? It won’t remind you before the next call.
Where does improvement language show up? In the two apps built for reviewing your own real calls:
"The app really helps me to improve my speech patterns and gain confidence when I talk." five-star Siro review (2023)
"As a former athlete being able to watch my game film in sales is an invaluable tool." five-star Rilla review (2023)
Athletes review game film; that’s what these reviewers are describing, for conversations. The catch: Siro and Rilla are team tools, typically rolled out by managers. When recording is imposed, the reviews turn dark. One one-star Rilla review reads, in full: "I used to enjoy my job. I have nothing else to say." Another describes being made to use the app to follow a company script, and rigging it. A third complains it recorded TV conversations for weeks. Same recorder, opposite experience. Chosen, it’s game film. Imposed, it’s surveillance.
What to check before you pay for any AI recording or coaching app
Five checks, straight from the failure patterns in these 663 reviews:
- Screenshot the trial terms on the paywall. The single most common complaint is "charged immediately" or "charged without warning." If the price and renewal date aren’t plainly shown, that’s your answer.
- Find the cancel path before you subscribe. On iPhone: Settings → your name → Subscriptions. If the app pushes its own billing outside the App Store, check its cancellation reviews first.
- Decide your consent habit up front. A bot in the participant list or a recorder on the table isn’t automatic consent. One sentence works: "I’d like to record this so I’m not scribbling while you talk — okay with you?"
- Ask what the tool changes. A notetaker gives you a record. If what you actually want is to get better (fewer rambles, sharper questions, real follow-ups), you need something that helps you review and adjust, not just archive.
- Prefer honest pricing signals. Real price shown before the trial, no countdown timers, working cancel flow. In this dataset, the anger followed the billing model everywhere it went.
Sources
- Our data: 663 public US App Store reviews (most-recent windows) + app metadata for 8 apps, pulled 2026-07-14 via the Apple App Store data API; counts recomputed 2026-07-15. Classification: keyword matching with manual reading of every match; exact patterns and per-app counts in our data sheet.
- AP (Cathy Bussewitz), "AI notetakers promise easy meeting recaps, but some professionals question their use," 2026-07-09 — syndicated via US News, The Washington Post, ABC News and others: usnews.com/news/technology/articles/2026-07-09/
- Fortune, "AI notetakers are creating HR nightmares," 2026-02-09 — fortune.com/2026/02/09/ai-notetakers-are-creating-hr-nightmares/
- NPR on the Otter.ai class action, 2025-08-15 — npr.org/2025/08/15/g-s1-83087/ · Jackson Lewis, Workplace Privacy Report (consolidated actions; allegations) — workplaceprivacyreport.com
Published July 15, 2026. Last verified July 15, 2026. Review counts reflect each app’s most-recent written-review window on the pull date.
This article is part of our research library. We publish what we learn building James, whether or not it sells an app.
Frequently asked questions
Are AI notetakers worth it?
For capture, yes. Recent reviews of the leading apps run 72–76% five-star. Users with ADHD or heavy meeting loads describe real relief. Check the billing terms carefully: 38% of low-star reviews across the category mention charges or trials. And don’t expect a notetaker to improve how you communicate: in the 200 recent reviews we analyzed, not one said that.
What do people complain about most in AI notetaker reviews?
Billing: surprise trial conversions, charges that are hard to refund, and cancel flows that don’t work. Accuracy complaints exist but are far rarer in recent reviews.
Do AI notetakers ask for consent?
Mostly no. A meeting bot appears in the participant list and a recorder just sits there. Audible announcements aren’t guaranteed, and vendors put the legal responsibility on you. Roughly a dozen US states require everyone’s consent to record (general information, not legal advice). Ask out loud; it takes one sentence.
Is there an AI tool that helps you speak better, not just take notes?
That’s a different category: apps that let you review your own real conversations (the “game film” approach) or coach your delivery. In our data, improvement language only appears in reviews of real-call review tools, and the experience depends heavily on whether recording was chosen or imposed.