What 22,114 App Store reviews say about video apps

The numbers at a glance

The report splits teleprompter apps into two groups. All-in-one apps prompt, caption and edit in one place. Prompter-first apps put the script first and are often used with rigs and remotes. The counts below are for those groups, never for a single named app.

Across the collection, about 80 reviews complained that an app couldn’t be tried before paying. The full list of published counts, with groups and periods, is in the methodology summary.

What the collection covered

The collection used Apple’s public customer-review feeds for the US, UK, Canada and Australia, requesting the most recent reviews in each storefront and the most helpful in the US, and removed duplicates by app ID and review ID. The feed returns at most 500 recent reviews per app, storefront and ordering, so this is recent opinion, not every review ever written.

The theme counts come from the close reading and from keyword matching, so one review can count under more than one theme. The collection predates three newer apps, EyePrompt, Cue and HighlightMe, and doesn’t include them. The company behind the research makes Lightshot, which is why no per-app counts appear here.

Do store ratings match what people write?

Not in this sample. Twenty apps had 30 or more written reviews between January 2025 and September 2026, ten teleprompters and ten editors or creator tools. In all 20, the average of those written reviews was lower than the app’s store rating.

The teleprompters’ store ratings ran from 4.48 to 4.87, while their written reviews averaged between 1.55 and 4.57. The editors’ store ratings ran from 4.61 to 4.86, and their written reviews from 2.61 to 4.48. Part of the gap is who is counted: a store rating includes star taps from people who write nothing, and apps can ask for those taps, up to three times a year according to Apple’s ratings and reviews page.

How often is billing the problem?

Among the all-in-one apps, about 130 of 901 low-rated reviews described a billing or trial trap: a charge during a free trial, a charge after cancelling, a yearly charge when a weekly plan was chosen, or the belief that deleting the app cancels it. Their share of 1 to 3 star reviews rose from about 8% in 2022 and 2023 to about 18% in 2024 to 2026. Price itself drew about 100 objections in that group and about 20 among prompter-first apps.

Deleting an app never cancels an App Store subscription. Apple’s cancellation steps and its refund request process are separate actions, and cancelling doesn’t ask for a refund.

Our own response is to show every billing period on the pricing page, explain cancelling separately from deleting the app, and send refund questions to Apple. Lightshot has no trial, which removes the trial-charge surprise but leaves the roughly 80 “let me try it first” complaints as a real cost of that choice.

How often do recordings go missing?

About 70 reviews of all-in-one apps and about 48 of prompter-first apps reported a lost recording. Among prompter-first apps the count went from 2 in 2022 to 21 in 2025 and 14 in the first nine months of 2026. Losing a take costs the delivery as well as the time, which is why saving behavior belongs in a buying decision.

For Lightshot, take recovery describes the design: parts are written in short fragments as the take goes, and an interrupted take is offered back at the next launch. It hasn’t been through device testing yet, so there is no “never lose a take” promise here.

Where does voice scrolling go wrong?

About 51 reviews of prompter-first apps and about 22 of all-in-one apps described voice scrolling that stalled, lagged or jumped ahead, and complaints in the all-in-one group rose in 2026. At the same time, about 13 reviews of apps without it asked for a prompter that follows the voice, so the feature is wanted and also the one people trust least.

“Voice activated” can mean three things: sound starts the scroll, silence pauses it, or recognized words set the position. Apple’s SpeechAnalyzer session describes on-device speech recognition, which says where recognition can run, not how accurate an app is. Our voice-follow page describes the matching behavior and the conditions not yet measured.

What goes wrong with sound and captions?

About 50 reviews of all-in-one apps reported missing or garbled audio, and about 22 reported sound drifting out of sync once captions were added. A style count doesn’t answer either problem. W3C’s caption guidance explains that captions exist to carry the audio as text, so the words have to match what was said.

Lightshot’s captions start from what the iPhone heard, and the script fixes spelling and punctuation only where it agrees with the speech. Recognition can still mishear, so the exported captions still need a read-through.

What do people ask for?

Requests to pause, resume, remove a mistake and carry on came up repeatedly, in reviewers’ own words such as “quick retake” and “delete last clip”. The report gives no single count for them. It does count these:

Several of these shaped Lightshot: sentence redo and part controls answer the retake requests, and mirrored text on a prompter-only iPhone is in every plan. That is evidence of a design problem worth solving, not evidence that no other app has solved it; the iPhone teleprompter guide and the app comparisons name the apps that have.

What did a second collection in October 2026 count?

On October 9, 2026 we collected the same 30 apps again from the same feeds and storefronts: 21,827 written reviews, 8,582 of them for 18 teleprompter apps, 12,138 for 10 video editors and 1,107 for Riverside and Loom, dated from November 2014 to October 8, 2026. Of those, 7,408 were rated 1 or 2 stars. A script matched fixed keyword patterns against the title and text of every 1 or 2 star review, so these are counts of reviews that mention a theme, not a close reading.

The patterns miss in both directions. A 1 star review that says “cancel” in passing counts under billing, one review can count under several themes, and a complaint put in other words isn’t counted at all. The patterns, and every count below with its denominators, are in the methodology summary. They measure something different from the September report’s close reading of 1 to 3 star reviews, so the two sets of numbers aren’t compared one for one.

Which themes come up most?

Billing, subscriptions and price came up most often: in 2,173 of the 7,408 low-rated reviews, which is 10.0% of every review in the snapshot. Crashes and bugs came next, in 1,569. Privacy and eye contact each turned up fewer than 30 times, too rarely to read anything into.

“The prompter or script” counts low-rated reviews that mention a prompter or a script, whatever the complaint, so it is a topic rather than a problem type.

How do teleprompters, editors and recording tools differ?

From January 2025 on, billing, subscriptions and price came up in 38% of teleprompter apps’ low-rated reviews and 27% of the video editors’. The editors’ complaints spread across more themes: export and video quality (14%), AI features (13%) and credits and usage limits (5%) all came up far more often than in teleprompter reviews. For Riverside and Loom, crashes and bugs (21%) and speed, processing and uploads (20%) came up at least as often as billing (17%).

Each cell gives the number of 1–2 star reviews that mention the theme, then its share of that group’s 1–2 star reviews. The same counts for all dates are in the methodology summary.

Did the themes change from month to month?

Not in a way these counts can show. Most months hold a few dozen low-rated reviews per group, so five reviews either way can move a share by ten points. Apple’s feed returns at most 500 reviews per app, storefront and ordering, and for 6 of the 117 app-storefront feeds those 500 didn’t reach back to October 1, 2025: five feeds of video editors and one of a teleprompter app. The monthly tables leave those six feeds out of every month, 3,340 reviews from October 2025 on, so each month counts the same apps and storefronts. The tables above include them.

Among teleprompter apps, billing, subscriptions and price led the six themes shown in 8 of the 12 months, tied with crashes and bugs in 2, and trailed it in October 2025 and September 2026.

Among the video editors, crashes and bugs led from November 2025 to January 2026, and billing, subscriptions and price led in every month from May to September 2026. The same feeds drew 37 to 55 low-rated reviews a month from October 2025 to April 2026 and 75 to 129 a month from May to September 2026; the counts can’t say why.

Riverside and Loom drew between 1 and 14 low-rated reviews a month, too few for monthly shares to mean much.

What this research can’t show

The feed limits which reviews are available by recency, storefront and ordering, and the people who write reviews choose themselves. A keyword such as “subscription” can appear in praise as well as complaint, one review can match several themes, and an app update can fix what an older review describes. For those reasons there is no league table of failure rates, and theme counts aren’t turned into shares of all customers.

The September report doesn’t give theme counts by month, per-theme counts for the video editors beyond the size of their low-rated sample (2,704 reviews rated 1 or 2 stars), or theme shares of the whole 22,114, and its review file wasn’t kept. Those numbers therefore come from the October snapshot above and aren’t attributed to the September collection. The FTC’s rule on consumer reviews is the background for using review evidence carefully: no reviewer is quoted or named here, and no competitor’s reviewer is turned into an endorsement.