Pulse
Concrete pains from app reviews — and how strong they are
- AI assistants
The bot forgets what we talked about
Pain 8 out of 10, strong. 285 reviews, in 41 of 58 apps. - AI assistants
The content filter is too strict
Pain 7 out of 10, strong. 176 reviews, in 29 of 58 apps. - AI assistants
The bot keeps repeating itself
Pain 6 out of 10, noticeable. 114 reviews, in 33 of 58 apps. - AI assistants
The AI gives false information
Pain 5 out of 10, noticeable. 91 reviews, in 24 of 58 apps. - AI assistants
It doesn’t do what I ask
Pain 5 out of 10, noticeable. 91 reviews, in 29 of 58 apps. - AI assistants
The AI’s knowledge is out of date
Pain 4 out of 10, noticeable. 63 reviews, in 21 of 58 apps. - AI assistants
Answers are politically slanted
Pain 4 out of 10, noticeable. 44 reviews, in 21 of 58 apps. - AI assistants
The bot talks for my character
Pain 2 out of 10, mild. 31 reviews, in 9 of 58 apps. - AI assistants
The bot suddenly switches languages
Pain 2 out of 10, mild. 27 reviews, in 10 of 58 apps.
How we count
The sample is up to 500 recent App Store reviews for each of about 100 popular apps per category. We look for each need across all of the category’s reviews and check a random sample of matches by hand: at least 80% are correct.
“Pain” is a rough score from 1 to 10: half comes from the share of the category’s reviews that mention it, 30% from how many such reviews there are in total, 20% from the share of apps where it comes up. It is not search demand and not market size.
In words: 1–3 mild · 4–6 noticeable · 7–8 strong · 9–10 acute.