Mobile Apps / App Store Optimization Beginner

30 ChatGPT prompts for mobile app founders to analyze App Store reviews

Your users are typing your next product roadmap into your App Store and Google Play listings every single day. They tell you what crashes, what feels slow, what they wish you’d build, and which tiny friction made them delete the app. Almost nobody reads those comments closely enough. I built this prompt pack so you can paste reviews into ChatGPT and walk away with a real plan: themes, churn signals, ASO rewrites, reply copy, and feature requests, all in the same afternoon.

These are the exact ChatGPT prompts for App Store review analysis I use with founders and indie teams. They are multi-line, structured, and tested against the way AppTweak, AppFollow, and Appfigures describe modern review data in 2026. You can run them on a 50-review export from App Store Connect, a 5,000-review dump from AppFollow, or a CSV from Sensor Tower. They work in ChatGPT-4o, ChatGPT-5, and the API, with minor tweaks.

Before we get to the prompts, let me show you why reviews are the cheapest research you have, and why 2026 is the right year to finally treat them like a system instead of a chore.

TL;DR - the cheat sheet

App store reviews are not feedback. They are a 2026-style focus group, free, and running 24/7. A drop from 4.2 to 4.0 stars can cost an app around 20% of its organic downloads, and Google says it blocked 160 million spam ratings and reviews on Google Play in 2025 alone (TechCrunch, Feb 19 2026). The 30 prompts below turn that firehose into decisions.

The 4 layers to mine: topics, emotions, churn, replies (more on this in a minute).
Use AppTweak-style semantic clusters to group reviews by intent, not by exact wording.
Use AppFollow’s reply-effect thinking to decide which review gets a manual answer vs. an AI-drafted one.
The 30 prompts are split into 6 categories: themes, sentiment, churn, feature requests, replies, ASO updates.
Run them on a rolling 30-day window every Friday. That is the entire workflow.

Why your reviews are the cheapest research you have

App Store review analysis is the practice of extracting product, support, and ASO insights from the free-text feedback users leave on the App Store and Google Play. Most founders treat reviews as a reputation chore. That is a mistake. Your reviews are a free, always-on, multi-language, high-intent research panel. They told you they love the new dashboard. They told you the paywall is confusing. They told you login breaks on Android 14. They told you, in plain English, what to build next.

A few hard numbers to anchor the value:

In 2025, the average conversion rate on the US App Store was 8.56%, and on Google Play it was 16.15%, while the average install rate from search or browse was 3.8% (AppTweak, May 18 2026). Reviews directly move that install rate.
A slide from 4.2 to 4.0 stars can mean a ~20% drop in downloads, and a coordinated review attack that pulls an app from 4.6 to 3.9 can cut installs in half until it recovers (AppFollow, Apr 23 2026).
Only 1 in 10 Google Play reviews gets a response, and only 2 in 10 on the App Store get one (AppFollow, Feb 22 2026). Reply, and you instantly beat 80% of your competition on a free trust signal.
99.9% of consumers read reviews at least occasionally when they decide what to buy, and that habit carries into app downloads (PowerReviews, cited in AppFollow).

If you only do one thing this month, do this: pull 200 recent reviews, paste them into Prompt #1 below, and read the output. You’ll know more about your app than most of your competitors know about theirs.

The 4-layer review analysis framework

I run every review-mining session through the same four layers. They work whether you have 50 reviews or 50,000, and they map cleanly to the prompt sections below.

Layer 1 - Topics (what they are saying). Group reviews into themes: login, crashes, pricing, onboarding, support, ads, feature X, etc. AppTweak calls this semantic clustering (AppTweak, Jan 19 2026). The point is to stop counting words and start counting themes.

Layer 2 - Emotions (how they feel). Separate frustration, delight, confusion, trust, and indifference. This is sentiment analysis by topic, not by review (AppFollow, Feb 22 2026). “Love the new UI but login is broken” is mixed, and you need to read both halves.

Layer 3 - Churn and competitive signals (what they will do). Look for “uninstalled,” “switched to,” “cancelling subscription,” “competitor X does this better.” These are your leading indicators for revenue loss, often weeks before the dashboard shows it.

Layer 4 - Replies and ASO updates (what you do). Use the topic + emotion + churn read to write review replies that lift the reply effect (Toca Boca moved its App Store rating from 3.9 to 4.1 stars and Google Play from 4.2 to 4.3 stars after automating replies with AppFollow). Then feed the same insights back into your metadata, screenshots, and short/long descriptions.

The prompts below are organized exactly in this order.

SECTION 1 - Theme & topic mining prompts (Prompts 1–5)

Theme mining is the act of clustering reviews into stable, named buckets so you can count problems, not just stars. This is the first move in any review analysis sprint. Once you have themes, every other layer (sentiment, churn, replies) gets easier.

Prompt 1 - The “give me my themes” prompt

Use it when: you have a fresh export of reviews (CSV or pasted text) and you have no idea where to start. This prompt gives you a clean taxonomy in one shot.

You are an expert App Store Optimization (ASO) and product analyst.

I will paste a block of user reviews from my iOS App Store and Google Play
listings below. Some reviews are short ("love it"), others are long. Reviews
may be in multiple languages; treat the dominant language as English and
preserve the original phrasing when quoting.

Your job is to extract a clean theme taxonomy from these reviews.

Follow these steps:
1. Read every review at least once.
2. Cluster reviews into 8-12 stable, named themes. Good theme names look
   like "Login & authentication," "Onboarding friction," "Subscription &
   paywall," "Crashes & stability," "Performance & speed," "Notifications,"
   "Customer support," "Feature request: dark mode," "Pricing complaints,"
   "Ads & monetization," "Bug: Android 14 only," and "Praise & delight."
3. For each theme, return:
   - theme name
   - short definition (one sentence)
   - approximate share of reviews that mention it (percentage)
   - 2-3 verbatim quotes that best represent the theme
   - whether the theme is mostly POSITIVE, NEGATIVE, or MIXED
4. Then list any "long-tail" themes that appeared fewer than 5 times but
   still feel important.
5. End with 3 "watchlist" themes: ones that feel like they could become
   a major problem in the next 30 days if we ignore them.

Do not summarize yet. Just output the theme map first. I will ask for
sentiment and action items in a follow-up prompt.

Here are the reviews:

<paste 100-500 reviews here>

Example output (truncated):

Theme	Share	Sentiment	Verbatim quote
Login & authentication	18%	Negative	”Can’t log in after the latest update, password reset loop is broken.”
Onboarding friction	12%	Negative	”Got lost in setup, didn’t know where to add my first budget.”
Subscription & paywall	9%	Mixed	”Worth it for me, but the trial is confusing.”
Crashes & stability	7%	Negative	”Crashes every time I open the shop tab on iOS 17.”

Pro tips:

Paste at least 200 reviews for a stable taxonomy. Below 100, you will get toy themes.
If your reviews cross languages, paste a 50/50 mix. ChatGPT handles this fine, but warn it in the prompt to preserve original phrasing.
Save the output as themes-v1.md. Re-run monthly. The taxonomy will drift as the product changes, and that drift is itself a signal.

Prompt 2 - Semantic cluster audit (AppTweak-style)

Use it when: you want to make sure your theme map matches how the app stores actually interpret search intent in 2026. AppTweak wrote in January 2026 that app store relevance is moving from exact-keyword matching to semantic interpretation (AppTweak, Jan 19 2026). Your reviews should match that shift.

You are an ASO and search-relevance expert trained on AppTweak's
semantic-clustering approach.

Below are user reviews from my app, grouped only by their raw 1-5 star
rating. I want you to re-cluster them using SEMANTIC intent, not exact
word matches.

For each cluster you find, provide:
- cluster label (a user-intent phrase, not a product feature)
- example user phrasing that should match this cluster even if the words
  are different ("can't log in" and "stuck on sign in" should land
  in the same cluster)
- representative reviews (3-5)
- the underlying user job-to-be-done ("I want to access my account
  quickly so I can check my balance")
- the ASO keywords and short-tail terms this cluster should feed

Important rules:
- A single review can belong to multiple clusters.
- Do not invent clusters that have fewer than 3 supporting reviews.
- Preserve original user language; do not paraphrase it into marketing
  copy.

Here are the reviews:

<paste reviews>

Why this prompt matters: in 2026, AppTweak argues that app stores are interpreting multiple intents per query (a search for “college” can return apps across many intents). Your review clusters should mirror the same multi-intent logic, so your ASO updates and your product fixes line up.

Pro tip: Run this prompt on competitor reviews too. Comparing your semantic clusters to a competitor’s reveals positioning gaps you can attack with new keywords.

Prompt 3 - The “first 30 days” new-version triage

Use it when: you just shipped an update, and the next 1,000 reviews will tell you whether you shipped a fix or a regression. This is the prompt I use inside the first 30 days of any release.

You are a release-quality analyst for a mobile app.

I'll paste reviews from the last 30 days. I just shipped version X.Y on
<date>. Many of these reviews were written after the release.

Please:
1. Sort reviews into "before release" vs "after release" using the date
   and any version mentions in the text. If you can't tell, mark the
   review "ambiguous."
2. For post-release reviews, find the top 5 NEW themes that did not
   appear (or appeared only rarely) in the pre-release reviews.
3. For each new theme, give:
   - the likely root cause (guess if you must, but say it's a guess)
   - a 1-sentence user impact summary
   - a severity score from 1 (annoyance) to 5 (users uninstalling)
   - a one-line suggested fix
4. Highlight any reviews that explicitly mention "uninstalled,"
   "switched to," or "refund." Treat these as P0.
5. End with a "ship / no-ship" verdict for the next hotfix. Should we
   rush a patch this week, or wait?

Reviews:

<paste reviews with dates>

Pro tips:

Pull only reviews with dates and version numbers for this prompt. A review without a date is useless for a release triage.
If your export does not include version numbers, do a quick regex pass in your spreadsheet first.
The “ship / no-ship” verdict is the single most useful output for a small team. It forces ChatGPT to commit to a recommendation.

Prompt 4 - Multi-language review clustering

Use it when: your app is live in 5+ countries and you want one unified theme map, not five separate ones. Most App Store Connect exports mix English, Spanish, German, Japanese, and more in the same CSV.

You are a multilingual app review analyst fluent in English, Spanish,
German, French, Portuguese, Japanese, Korean, and Simplified Chinese.

Below are reviews from our iOS and Google Play apps. They are mixed
across languages. Some are short, some are machine-translated, some
are written in regional slang.

Do the following:
1. Detect the language of each review.
2. Translate non-English reviews to English internally; quote the
   original phrase in parentheses for the top 10 most-mentioned issues.
3. Cluster all reviews into a single 8-12 theme taxonomy (in English)
   regardless of source language.
4. For each theme, show a "language mix" line:
   "Login issues: 45% EN, 20% ES, 15% DE, 10% JA, 10% other."
5. Flag any themes that are heavily concentrated in one language.
   That usually means a localization bug, a regional pricing issue,
   or a country-specific crash.

Here are the reviews:

<paste multi-language reviews>

Why this matters: AppFollow’s Semantic Tags work across 20 languages out of the box (AppFollow, May 29 2026). Your ChatGPT pipeline should aim for the same coverage, because the fastest-growing source of new reviews for most apps is now non-English.

Pro tip: If you localize your app, run this prompt in the language of the market you care about most, not just English. You’ll catch translation drift and weird literalisms that hurt conversion.

Prompt 5 - Trend detection across two time windows

Use it when: you want to know if a theme is growing, shrinking, or stable. The app stores reward app teams that can spot a slow-moving complaint before it becomes a 1-star avalanche.

You are a mobile app growth analyst.

Below are two blocks of reviews:
- BLOCK A: reviews from <60 days ago window, e.g. 60-90 days ago>
- BLOCK B: reviews from the most recent 30 days

Both blocks are from the same app, same stores, same product. The product
shipped 2 minor updates between the two windows.

Please:
1. Build a theme map for each block separately. Use the SAME theme
   labels across both blocks so they are comparable.
2. For each theme, compute:
   - share of reviews in block A
   - share of reviews in block B
   - delta (percentage point change)
   - direction (improving, stable, worsening)
3. Identify:
   - the 3 themes with the largest negative deltas
   - the 3 themes with the largest positive deltas
   - any NEW themes that appeared in block B but not in block A
   - any themes that DISAPPEARED between A and B (often a sign of a
     successful fix)
4. For the 3 largest negative deltas, write a short hypothesis for
   what changed in the product to cause it. Be specific.

BLOCK A (60-90 days ago):
<paste>

BLOCK B (last 30 days):
<paste>

Pro tips:

The themes that “disappear” between windows are some of the most useful outputs. They tell you a fix worked, and they earn you internal credit for the next ASO cycle.
Pair this prompt with the AppTweak conversion rate data (AppTweak, May 18 2026). A worsening theme paired with a falling CVR is a smoking gun.
Run it on a 30-day rolling cadence. That gives you a clean baseline that does not break on holidays or big marketing pushes.

SECTION 2 - Sentiment & emotion prompts (Prompts 6–10)

Sentiment analysis in app reviews is the practice of scoring not just whether a review is positive or negative, but what specific emotion the user is feeling and what product surface caused it. A 5-star review with a hidden complaint (“5 stars, just wish dark mode was real”) is more useful to you than a flat 1-star rant.

Prompt 6 - Mixed-sentiment reviewer detection

Use it when: you have a list of reviews and you want to find the ones that look positive on the surface but actually contain a real complaint. These are the highest-leverage reviews to reply to.

You are a customer-experience analyst for a mobile app.

I will paste reviews below. For each one, classify it as:
- PURE_POSITIVE: praise with no complaints
- PURE_NEGATIVE: complaint with no praise
- MIXED: contains BOTH praise and a complaint
- NEUTRAL: factual observation, bug report, or feature request
- SPAM: clearly fake, off-topic, or promotional

For each MIXED review, do extra work:
- Quote the praise portion (1 sentence)
- Quote the complaint portion (1 sentence)
- Identify the product surface the complaint is about
- Suggest the best reply strategy (acknowledge, fix, redirect, escalate)
- Rate the churn risk from 1 (low) to 5 (high) based on tone

Sort the MIXED reviews by churn risk, highest first. These are the
reviews that matter most this week.

Reviews:

<paste reviews>

Pro tip: Most “5-star” reviews you reply to are wasted effort. Most MIXED reviews you reply to actually move the reply effect, which is the metric Toca Boca tracked when they lifted ratings by 0.2 stars in a single quarter using AppFollow’s automation.

Prompt 7 - Emotion taxonomy builder

Use it when: you want a richer read on your reviews than “positive / negative / neutral.” Real churn risk lives in emotions like frustration, distrust, and resignation, not in star counts.

You are a UX researcher specializing in emotional analysis of user
feedback.

Below are app reviews. I want you to tag each review with one primary
emotion and (optionally) one secondary emotion, using this taxonomy:

Primary emotions:
- DELIGHT: surprise, joy, excitement
- TRUST: confidence,安全感 (sense of safety), reassurance
- FRUSTRATION: annoyance, blocked, stuck
- CONFUSION: lost, uncertain, "I don't get it"
- DISAPPOINTMENT: unmet expectations
- DISTRUST: suspicion, fear of being scammed, privacy worry
- INDIFFERENCE: factual, low energy
- ANGER: rage, threat, demands
- RESIGNATION: "I'll just live with it," "whatever"
- GRATITUDE: thanks, appreciation

For each review, return:
- primary emotion
- secondary emotion (or "none")
- the phrase that triggered the primary tag (quote it)
- a 1-sentence interpretation: "The user is feeling X because Y."

Then output a summary table at the end:
- count of reviews per primary emotion
- the top 3 emotions linked to churn risk (FRUSTRATION + RESIGNATION
  + DISTRUST usually win)

Reviews:

<paste reviews>

Why this matters: AppFollow specifically calls out that resigned and frustrated users churn quietly, and you’ll miss them if you only look at star averages (AppFollow, Feb 22 2026). The emotion taxonomy makes that cohort visible.

Prompt 8 - Sentiment by app version (the post-release mood)

Use it when: you want to know if your v6.2 release made users happier or sadder than v6.1. This is the prompt that turns a “release is fine, I think” gut feel into a real chart you can show your team.

You are a release-impact analyst.

Below are reviews tagged with their app version. Each line includes the
version, star rating, and the review text.

Please:
1. Compute the average star rating per version.
2. Compute a sentiment score per version on a 1-10 scale (10 = pure
   delight, 1 = pure rage). Use both the star rating and the emotional
   language in the text to derive the score. Be conservative; do not
   inflate the score.
3. For each version, list:
   - the top 3 POSITIVE themes
   - the top 3 NEGATIVE themes
   - the single most "improved" theme vs the previous version
   - the single most "regressed" theme vs the previous version
4. Compute the "release mood delta": a one-line summary of whether
   the release made users feel better, the same, or worse.
5. Flag any version with a regression in onboarding, login, or
   payment. These three are conversion-killers.

Reviews:

<version, stars, text>

Pro tips:

If your export does not include versions, pull them from the Apple App Store RSS feed or use the App Store Connect API.
This is the prompt that convinces a stubborn PM to ship a hotfix. Show them the regression line.

Prompt 9 - Sentiment by country (your localization report card)

Use it when: you localize for 10+ markets and need to know which country is the canary. AppTweak’s conversion benchmarks vary by region and category (AppTweak, May 18 2026), and so does your review sentiment.

You are a localization and international growth analyst.

Below are reviews tagged with country code. Some are translated; some
are in the local language. Detect the language and the country when
the country is missing.

Please:
1. Compute the average star rating and sentiment score (1-10) per
   country. Sort by sentiment, lowest first.
2. For the bottom 5 countries, list:
   - the top 3 negative themes in that country
   - whether the issue looks like a translation bug, a pricing
     mismatch, a regional crash, or a payment-method problem
   - one concrete localization or product recommendation
3. For the top 3 countries, list:
   - the top 2 praise themes
   - one quote we can reuse in marketing (with the user's tone
     preserved)
4. Flag any country where the sentiment dropped more than 0.5
   points vs the global average. That is your canary market.

Reviews:

<country, stars, text>

Why this matters: Toca Boca’s review automation case study showed that even generic 1-star auto-replies, when done well, can carry a +1.71 reply effect (AppFollow, Feb 12 2026). Imagine the upside if you localize that reply workflow per market.

Prompt 10 - “Moment of joy” extraction

Use it when: you need raw material for screenshot copy, App Store preview videos, ad creatives, or your “what users love” press section. AppTweak argues that review language is one of the strongest signals for what to put on a screenshot (AppTweak, Apr 21 2026). This prompt is the engine that gets you that copy.

You are an app marketing copywriter and ASO specialist.

Below are app reviews with a 4 or 5 star rating. Your job is to extract
"MOMENTS OF JOY" - specific phrases where the user describes the
outcome they love.

For each moment of joy, return:
- the original review quote (1-2 sentences)
- the outcome the user got (one short phrase: "fell asleep faster,"
  "tracked my spending in 10 seconds," etc.)
- a screenshot-ready headline we could use (max 8 words, in the
  user's voice, NOT marketing-speak)
- a 1-sentence subtitle that supports the headline
- the underlying value prop it proves (so we don't repeat the same
  angle 10 times)

Then group the moments of joy into 5-7 "value clusters" so we can see
which outcomes we have the most proof for.

End with the 5 strongest moments of joy overall. These are the ones
we should test as the first screenshot headline.

Reviews:

<paste 4-5 star reviews>

Pro tip: Use the output of this prompt to test a Custom Product Page (CPP) in Apple Search Ads. AppTweak’s case data shows CPPs are now central to intent-led ASO (AppTweak, Jan 19 2026).

SECTION 3 - Churn-signal & competitor prompts (Prompts 11–15)

Churn-signal review analysis is the practice of finding the comments that predict a user leaving, downgrading, switching to a competitor, or requesting a refund - weeks before the cancel event shows up in your subscription dashboard. A 1-star rant is loud but usually late. A 4-star “it’s fine, but I’m also trying X” is the real warning.

Prompt 11 - Uninstall intent scoring

Use it when: you want to flag at-risk users from review text alone. Use this on a weekly rolling basis.

You are a customer success and churn analyst.

Below are app reviews. For each review, score the UNINSTALL INTENT
from 0 to 100, where:
- 0 = no signal, the user is happy and engaged
- 50 = the user is unhappy but staying (waiting for a fix)
- 100 = the user has already uninstalled, switched apps, requested
  a refund, or written a final goodbye

For each review, return:
- uninstall intent score (0-100)
- 1-sentence reason for the score (quote the trigger phrase)
- product surface driving the intent (login, paywall, etc.)
- recommended next action: (a) reply with empathy, (b) offer a fix
  in-app, (c) escalate to support, (d) do nothing

Sort the list by score, highest first. The top 10% are your "save
list" for the week. You should respond to them manually, not with
an auto-reply.

Reviews:

<paste reviews>

Pro tip: Combine this with AppFollow’s reply-effect metric. Toca Boca’s automation tracked the reply effect on every automated message and pulled back automations that hurt the rating (AppFollow, Feb 12 2026). You should do the same: if your “save list” replies move the needle, double down. If they don’t, fix the underlying product issue.

Prompt 12 - Competitor mention miner

Use it when: you want a free competitive intel feed. Users will literally tell you which app they almost installed instead, and why they chose your competitor. Stop paying for that data on Sensor Tower when reviews give it to you raw.

You are a competitive intelligence analyst for mobile apps.

Below are user reviews. Find every review that mentions:
- a competitor app by name (e.g., "I switched from Notion to this")
- a category ("I tried 3 budgeting apps before this")
- a feature comparison ("unlike X, this one actually...")

For each mention, return:
- the full review quote
- the competitor name (or "unspecified")
- the comparison being made (price, feature, support, design, etc.)
- whether the user is staying with us, leaving us, or considering
  leaving
- the insight we should act on (one sentence)

Group the output by competitor. The competitor with the most
mentions is the one our positioning should be tuned against.

Reviews:

<paste reviews>

Why this matters: Sensor Tower and data.ai are great for downloads and revenue estimates (AppTweak, Feb 26 2026), but neither will tell you that 14% of your churned users said “switched to Y because Y’s onboarding is faster.” Only your reviews will.

Prompt 13 - Subscription and refund risk detector

Use it when: you’re a subscription app (which, in 2026, is most apps) and you want to catch refund requests and cancel threats early. AppFollow and AppTweak both flag subscription confusion as a top churn driver.

You are a subscription-economy analyst.

Below are reviews of a subscription-based mobile app. Identify any
review that signals:
- confusion about a charge ("I was charged and didn't know why")
- a refund request ("I want my money back")
- a billing dispute ("this is a scam, I never signed up")
- a downgrade intent ("switching to the free plan")
- a cancellation intent ("cancelling my subscription")
- praise for the value ("worth every penny")

For each signal, return:
- the review quote
- the signal category
- severity: 1 (mild confusion) to 5 (legal or chargeback risk)
- the likely root cause: paywall copy, renewal flow, free trial
  length, price change, missing cancel button, or unclear value
- the recommended fix (product, support, or comms)

End with the 3 highest-impact fixes for the next 30 days.

Reviews:

<paste reviews>

Pro tip: A subscription confusion spike is one of the few signals that can wipe out a quarter of LTV in a week. If Prompt #13 lights up after a price change or paywall update, treat it as P0.

Prompt 14 - Review-bomb and coordinated-attack detection

Use it when: you suddenly see a 0.5-star or larger drop in a 24-hour window, and you need to know if it’s a real product issue or a coordinated attack. Google Play’s 2025 safety report says the company blocked 160 million spam ratings and reviews in 2025, preventing an average 0.5-star drop for bombed apps (TechCrunch, Feb 19 2026). That is a real risk you need a workflow for.

You are an app store fraud and reputation analyst.

Below are reviews from the last 72 hours, with timestamps, star
ratings, account age (in days), and review text.

Please:
1. Compute the baseline 30-day review volume and rating distribution.
2. Compare the last 72 hours to the baseline. Flag any of:
   - volume spike >2x baseline
   - 1-star share spike >3x baseline
   - cluster of reviews with near-identical phrasing (>5 reviews
     sharing >50% of their words)
   - cluster of reviews from accounts <7 days old
   - mentions of external topics unrelated to the app (politics,
     celebrity, etc.)
3. Decide one of three verdicts:
   - LIKELY ORGANIC: matches a real product event (a release, a
     server issue, a price change)
   - LIKELY COORDINATED: signs of a bot campaign or review-bomb
   - INCONCLUSIVE: needs human investigation
4. For LIKELY COORDINATED verdicts, draft:
   - a short report to send to App Store / Google Play reporting
     the attack
   - 3 public-facing reply templates for legitimate users caught
     in the crossfire
   - a 7-day recovery plan to restore star rating
5. For LIKELY ORGANIC verdicts, draft a triage list with severity
   scores and suggested fixes.

Reviews:

<timestamp, stars, account age, text>

Why this matters: AppFollow’s fake review guide confirms that 60–80% of clearly fake reviews get removed when developers report them. If you do not have a workflow for this prompt, you are leaving rating points on the table every quarter.

Use it when: you have access to reviews that include any signal of when the user installed (some exports include “first review” or “Nth review” data). This is a more advanced prompt that pairs review tone with account age.

You are a cohort and retention analyst for a mobile app.

Below are reviews tagged with a "days since install" field (0-7,
8-30, 31-90, 91-365, 365+).

Please:
1. Compute average star rating and sentiment score per cohort.
2. For each cohort, list:
   - top 3 positive themes
   - top 3 negative themes
   - the "aha moment" theme (where the user first expresses delight)
   - the "drop-off" theme (where the user first expresses regret)
3. Identify the cohort with the WORST sentiment. Hypothesize why.
   Common causes are onboarding gaps, a recent UX change that broke
   muscle memory, or a price change that hit renewing users hardest.
4. Identify the cohort with the BEST sentiment. What is working
   there that we could ship to other cohorts?
5. Recommend 2 product experiments to lift sentiment in the
   worst cohort over the next 30 days.

Reviews:

<days since install, stars, text>

Pro tip: The most useful single chart this prompt produces is “drop-off theme” by cohort. If new users churn on onboarding and old users churn on price, you have two completely different retention problems. Most teams do not realize that until they run this prompt.

SECTION 4 - Feature-request extraction prompts (Prompts 16–20)

Feature-request extraction is the practice of turning messy, vague review text into a prioritized, deduplicated, RICE-scored roadmap. The hard part is not finding requests. The hard part is deduplicating “I want dark mode” from “Please add a dark theme for night shifts” from “Eyes hurt at night, need darker UI,” and ranking them with real signal.

Prompt 16 - Feature request deduplicator

Use it when: you have 1,000+ reviews and the same request shows up in 15 different phrasings. This prompt collapses them.

You are a product manager and feature-request analyst.

Below are app reviews. Some contain feature requests, some do not.

Step 1: Identify every review that contains a feature request, even
if the user phrased it as a complaint, a workaround, or a wish.

Step 2: Deduplicate requests into a clean list. For example, these
three reviews should collapse into one canonical request:
- "Add dark mode please"
- "my eyes hurt at night, can you make a darker UI?"
- "would love a black theme option"
Canonical request: "Dark mode / black theme for low-light use."

Step 3: For each canonical request, return:
- a clean name (5-10 words)
- a 1-sentence user problem statement
- the number of reviews that support it
- the share of total reviews that mention it
- 3 verbatim quotes
- which user segment is most likely asking (new users, power users,
  paying users, churned users, etc.)
- whether the request is a quick win, a medium project, or a
  moonshot (your best guess)

Step 4: Sort by share, highest first.

Reviews:

<paste reviews>

Pro tip: If the top three requests are all “fix the existing thing, please,” you don’t have a feature problem, you have a stability problem. Push those to engineering, not to the roadmap.

Prompt 17 - RICE-scored feature shortlist

Use it when: you have the deduplicated list from Prompt #16 and need to defend a roadmap decision. RICE = Reach, Impact, Confidence, Effort.

You are a product strategy lead.

Below is a deduplicated list of feature requests extracted from user
reviews, with request volume and supporting quotes.

Score each request using the RICE framework:
- REACH: how many users per month would this affect? (rough number
  based on request volume and review volume)
- IMPACT: 0.25 (minimal) to 3.0 (massive). What is the user impact
  if we ship this?
- CONFIDENCE: 0.5 (low) to 1.0 (high). How sure are we about the
  above numbers?
- EFFORT: person-weeks to ship the smallest useful version

RICE score = (Reach x Impact x Confidence) / Effort

Return a ranked table with: request name, R, I, C, E, RICE score,
1-sentence rationale, and 1 risk.

Then recommend the top 5 to ship in the next quarter, the 3 to
explicitly reject (and why), and the 2 to table for later.

Request list:

<paste deduplicated list>

Pro tip: The most under-rated column here is Confidence. If a request has 200 mentions but they’re all power users who represent 2% of revenue, your confidence should be low. Resist the urge to ship only what loud users ask for.

Prompt 18 - JTBD jobs-to-be-done lens

Use it when: the feature request is loud, but you suspect the underlying job-to-be-done is different from what the user said. AppFollow and AppTweak both recommend pairing review language with user intent (see the AppTweak semantic relevance article).

You are a jobs-to-be-done (JTBD) coach for product teams.

Below are user reviews, all of which contain feature requests.

For each review, identify:
- the surface-level feature the user is asking for ("add a savings
  goal widget")
- the underlying JOB they are trying to get done ("I want to see
  progress toward a specific money milestone so I stay motivated")
- the emotional job ("I want to feel in control of my money")
- the social job ("I want to be able to tell my partner we're on
  track")

Then group the reviews by UNDERLYING JOB (not by feature). Many
features may map to the same job.

Recommend the smallest set of solutions that would address the
top 3 underlying jobs. A single feature can sometimes serve 3
jobs if you design it right.

Reviews:

<paste reviews>

Pro tip: This is the prompt that breaks a roadmapping deadlock. Stakeholders stop arguing about “should we ship a savings goal widget” and start arguing about “which underlying job do we want to own.” That reframe usually unlocks the meeting.

Prompt 19 - Willingness-to-pay detector

Use it when: you want to know which feature requests are actually worth monetizing vs. table stakes. AppTweak and Appfigures both note that revenue and review sentiment are linked (Appfigures, May 4 2026), so feature requests from paying users deserve a different weight.

You are a product monetization analyst.

Below are reviews tagged with the user's subscription tier (Free,
Pro, Premium, Trial, Unknown).

For each feature request, compute:
- the share of mentions from paying users (Pro + Premium)
- the share from free users
- the share from trial users
- the share with unknown tier

Then for each request, return:
- willingness-to-pay signal: STRONG (mostly paying), MODERATE
  (mixed), WEAK (mostly free/trial)
- monetization recommendation:
  (a) include in the base plan (high-volume, low-tier)
  (b) gate behind the Pro plan (high-volume, paying-heavy)
  (c) ignore (low volume, low-tier)
- one-sentence rationale

End with a "premium bait" list: the 3 features most likely to
convert free users to paid based on the review data alone.

Reviews:

<tier, request, text>

Pro tip: Use this prompt alongside RevenueCat or Superwall data. If a “premium bait” feature in reviews lines up with a real paywall A/B test lift, you have a winner. If it doesn’t, you have a feature that users say they want but do not pay for.

Prompt 20 - Request frequency vs. silence frequency

Use it when: the loudest users can dominate your roadmap. This prompt finds the silent majority - features that are missing but not complained about, because users assume they’re impossible or have churned already.

You are a product discovery analyst.

Below are reviews. For each review, classify as:
- REQUEST: user asks for a new feature
- COMPLAINT: user reports a bug or friction
- SILENCE-OPPORTUNITY: review neither requests nor complains but
  hints at a missing capability
  (Example: "I just export to CSV manually because I assume the
  app can't do bulk export." This is a SILENCE-OPPORTUNITY.)

Step 1: Extract all REQUEST and SILENCE-OPPORTUNITY mentions.

Step 2: Build a 2x2:
- HIGH DEMAND, HIGH FRUSTRATION: explicit, repeated requests
- HIGH DEMAND, LOW FRUSTRATION: silence opportunities (people want
  this but won't ask)
- LOW DEMAND, HIGH FRUSTRATION: niche bugs (probably ignore)
- LOW DEMAND, LOW FRUSTRATION: nice-to-haves (deprioritize)

Step 3: Recommend 3 SILENCE-OPPORTUNITY features to ship in the
next quarter. These are the easiest wins because users will
delight in getting something they didn't expect.

Reviews:

<paste reviews>

Pro tip: “Silence opportunities” are the best place to look for differentiation. AppTweak’s 2026 ASO trends note that AI-driven search rewards apps that clearly communicate use cases - silence opportunities are usually use cases users want but your metadata doesn’t mention.

SECTION 5 - Review reply & community prompts (Prompts 21–25)

Review reply analysis is the practice of using AI to draft, personalize, and prioritize replies that move the reply effect - the metric Toca Boca used to lift its App Store rating from 3.9 to 4.1 stars and Google Play from 4.2 to 4.3 stars (AppFollow, Feb 12 2026). The trick is not to reply to everything. The trick is to reply to the right things in the right tone.

Prompt 21 - Per-review reply drafter

Use it when: you want human-quality replies at AI speed. The output of this prompt is ready to send after a quick read.

You are a customer support lead who writes reply copy for App Store
and Google Play reviews. Your tone is:
- warm, never robotic
- short, max 3 sentences
- specific to the review (no generic "thank you for your feedback")
- always offers a next step (a fix, a link, an in-app action, an
  escalation path)
- never defensive
- never uses the words "valued customer," "we apologize for any
  inconvenience," or "as a company"

For each review below, draft 1 reply. Then draft a backup reply
that takes a different angle (e.g., the first is empathetic, the
second is action-focused). Mark which one you would actually send.

If the review is a 1-star crash report, include: "We've fixed this
in v6.2.1, please update and let us know."
If the review is a feature request, include: "Thanks - this is on
our roadmap, can we email you when it ships?"
If the review is a refund dispute, include: "Please email
support@<app>.com with your account email, we'll resolve today."

Reviews:

<paste reviews>

Pro tip: Save the backup reply you don’t pick. Over time, that backup library becomes a reply template bank that fits more edge cases than any vendor gives you out of the box.

Prompt 22 - Tone of voice calibrator

Use it when: your team is split on whether replies sound too corporate, too casual, too apologetic, or too curt. Use this prompt to lock in a tone guide and stress-test it.

You are a brand voice consultant.

Below are 5 sample reviews and 3 sample replies written by my team.

Please:
1. Define the current tone of our replies across 5 dimensions:
   warm / cold, formal / casual, apologetic / assertive, specific /
   generic, action-oriented / passive.
2. Identify the 2 dimensions where the team is most inconsistent.
3. Write a one-paragraph "voice guide" for future replies (max
   120 words).
4. Rewrite each of the 3 sample replies in the corrected tone.
5. Generate 5 "before / after" reply pairs: a generic reply
   transformed into the corrected tone. Use placeholder review
   text (e.g., "1-star review about login bug on Android 14").

Reviews and replies:

<paste>

Pro tip: Toca Boca’s team kept iterating on their reply templates every quarter because player sentiment shifts and what sounded fine in Q1 felt cold in Q3 (AppFollow, Feb 12 2026). Run this prompt quarterly. Treat tone like a product surface.

Prompt 23 - Public-relations reply for a viral complaint

Use it when: a single 1-star review hits Reddit, X, or a Discord and suddenly your app is trending for the wrong reason. You need a reply that satisfies the original user, the public reading the thread, and your own support team.

You are a crisis communications lead for a mobile app.

A user posted the review below. It is now visible publicly, has
been shared on social media, and is being screenshotted by
competitors and press.

Draft 3 versions of a public reply:
1. SHORT: 1-2 sentences, for the App Store / Google Play public
   reply box.
2. MEDIUM: a paragraph for our public social media response
   (X, Threads, LinkedIn).
3. LONG: a full statement for our blog / status page, with a
   timeline commitment.

Rules:
- Do not lie, do not minimize, do not promise a date you cannot
  meet.
- Acknowledge the user's experience first, then explain, then
  commit to a next step.
- Include a clear, single CTA: where the user should go for
  support.
- Include an internal-only "what we will actually do this week"
  list at the end (4-6 bullet points).

The review:

<paste review>

Pro tip: After you send the public reply, drop the user’s review ID and the prompt output into a “crisis playbook” doc. Next time, you will not have to write it from scratch.

Prompt 24 - Community manager reply for positive reviews

Use it when: your team is so focused on 1-stars that you forget the 5-star reviews are the ones that close the next sale. AppTweak’s ASO tips for 2026 highlight that positive review language is the strongest source of screenshot copy. Replying to positive reviews is also your chance to recruit a power user.

You are a community manager.

Below are 4 and 5 star reviews. For each one, draft a reply that:
- thanks the user by their first name (use the display name)
- quotes back 1 phrase from their review so they feel heard
- invites them to do ONE of: (a) join your beta program, (b)
  follow you on social, (c) refer a friend, (d) share a tip
  in your community Discord / subreddit
- mentions a hidden feature or shortcut they might enjoy (we
  want to reward power users)

Keep replies under 3 sentences. Do not use emojis. Do not
sound like a bot.

Reviews:

<paste reviews>

Pro tip: A 5-star user who gets a thoughtful reply becomes a 5-star user who tells 3 friends. That is the cheapest UA you will ever buy.

Prompt 25 - Internal Slack digest generator

Use it when: your team is too busy to read reviews, and you need a 90-second weekly digest they will actually open. This is the prompt I send to a private Slack channel every Friday.

You are a chief of staff for a mobile app product team.

I will paste the week's reviews and the App Store / Google Play
ratings. Draft a Slack-friendly weekly digest with these
sections, each one short enough to read in 10 seconds:

1. HEADLINE: a single sentence. Format: "Rating moved from X to Y.
   Top theme this week: Z."

2. WHAT IMPROVED: 2 bullets. Use the phrase "fewer X reviews" or
   "more Y reviews" with rough numbers.

3. WHAT BROKE: 2 bullets. Same format. Be specific.

4. CHURN WATCH: 1-2 reviews that look like P0 churn risk. Quote
   one phrase from each. Suggest a one-line action.

5. ASO WATCH: 1-2 reviews that contain language we should test
   in our next screenshot or Apple Search Ads creative.

6. SHIP / NO-SHIP: should we ship a hotfix this week? (Yes / No
   / Wait). One-line reason.

7. NEXT WEEK: 1-2 things to investigate.

Tone: blunt, no fluff, no "Great week, team!" opener. The team is
busy and allergic to corporate speak.

Reviews and ratings:

<paste>

Pro tip: Pin this digest in your team channel every Friday. After 3 months, the trend lines will tell your story better than any dashboard.

SECTION 6 - ASO update prompts (Prompts 26–30)

ASO update review analysis is the practice of turning review language into new title, subtitle, keyword field, description, and screenshot copy. AppTweak’s January 2026 piece argues that app store relevance is shifting to semantic interpretation, where keywords, creatives, reviews, and store pages work as connected signals (AppTweak, Jan 19 2026). Your reviews are the single richest source of semantic signals you have.

Prompt 26 - Title and subtitle rewriter from review language

Use it when: your title and subtitle are stale, your conversion is soft, and you want to test copy drawn from how real users describe your app.

You are an ASO copywriter trained on AppTweak's 2026 ASO playbook.

Below are 4-5 star reviews from my iOS app. Your job is to extract
a new TITLE and SUBTITLE for the App Store.

Rules:
- Title: max 30 characters
- Subtitle: max 30 characters
- Use a real phrase from the reviews if possible (lightly cleaned)
- Lead with the user OUTCOME, not the feature
- Include the strongest primary keyword in the title
- Do not use ALL CAPS, emojis, or marketing hype words ("best," "ultimate")

Produce 5 options for each (title + subtitle pair). For each option,
explain in one sentence:
- which review phrases it draws from
- which primary keyword it targets
- which user intent it satisfies

Then recommend 1 winner and 1 backup.

Reviews:

<paste reviews>

Pro tip: Test the winner in an App Store Connect Product Page Optimization (PPO) experiment first, not a hard cut. AppTweak’s 2026 ASO trends say PPO tests are the safest way to validate creative changes.

Prompt 27 - Screenshot copy from review “moments of joy”

Use it when: your first screenshot headline is the biggest conversion lever on the page, and you want copy that comes from users, not from your marketing team.

You are a conversion copywriter for App Store screenshots.

Below are 4-5 star reviews and a list of "moments of joy" extracted
from them.

For each of my 5 screenshot slots, write:
- a HEADLINE (max 6 words, in the user's voice)
- a SUBLINE (max 12 words, explains the outcome)
- a TINY visual cue to pair with it (1 phrase: "phone in hand, dark UI")

The five slots should be:
1. HERO: the single most powerful moment of joy
2. PROOF: a number or stat the user mentioned
3. FEATURE: the feature that most often gets praise
4. TRUST: an objecting-buster ("no ads," "no signup," "free forever")
5. CTA: the next step ("Try it free," "Set up in 2 minutes")

Do not use the same word in more than 2 headlines.

Moments of joy and reviews:

<paste>

Pro tip: AppTweak and AppFollow both report that the first screenshot is now a decision tool, not an introduction (AppTweak, Apr 21 2026). Use the highest-volume “moment of joy” as your hero, not your most polished feature.

Prompt 28 - Long description rewriter for semantic relevance

Use it when: your long description (Google Play) and your “promotional text” (App Store) are stale, and you want copy that is dense with the user language the app stores are now interpreting semantically.

You are an ASO copywriter specializing in semantic relevance for
2026 app store search.

Below are my app's 4-5 star reviews AND my competitor's 4-5 star
reviews. Your job is to write a new long description (up to 4,000
characters) for Google Play and a new "promotional text" (up to
170 characters) for the App Store.

The new copy must:
- use natural language that mirrors how users describe the app
- include 5-7 high-intent phrases drawn directly from reviews
- avoid keyword stuffing
- have a clear structure: hook (1 sentence) > benefits (3-5
  bullets) > proof (1-2 lines) > CTA (1 line)
- be skimmable on mobile

Then list the 5-7 phrases you used and the review count that
backs each one. This is your semantic relevance audit.

My reviews:
<paste>

Competitor reviews:
<paste>

Pro tip: AppTweak’s June 2025 algorithm change showed that app store search is now surfacing a broader mix of intent-related apps (AppTweak, Jan 19 2026). Your long description is one of the only places you can spell out the multiple intents you serve. Use it.

Prompt 29 - Apple Search Ads keyword and Creative Set builder

Use it when: you want your paid keyword list to feel as human as your organic metadata, and you want Creative Sets that look like screenshots, not ads.

You are an Apple Search Ads expert.

Below are app reviews, with stars and text.

Step 1: Build a keyword list of 30 Apple Search Ads terms. Mix:
- 10 high-intent head terms (2-3 words)
- 10 long-tail problem terms ("how to track X without Y")
- 10 competitor or alternative terms

For each term, mark the match type (Exact, Broad, or Search Match)
and a suggested starting CPT based on category benchmarks.

Step 2: Pick the top 5 keywords and write a Creative Set for each:
- a headline (max 30 chars)
- a description (max 90 chars)
- a CTA suggestion

All copy must pull from real review language. Cite the review
count for each phrase used.

Step 3: Flag any keywords that look like they have low commercial
intent ("free," "alternative," "vs"). Recommend a negative
keyword list.

Reviews:

<paste>

Pro tip: AppTweak wrote that Apple Search Ads is now expanding into more ad placements (AppTweak, Mar 2 2026). Your Creative Set copy from this prompt will get more surface area than ever in 2026 - don’t waste it on generic copy.

Prompt 30 - Custom Product Page generator per intent cluster

Use it when: you want to use CPPs and Custom Store Listings to match the multi-intent world AppTweak describes, and your current CPPs are just slightly-different versions of the same screenshot.

You are a CPP (Custom Product Page) and CSL (Custom Store Listing)
specialist for 2026 iOS and Google Play.

Below are my app's review clusters (output of Prompt #2). I have
3-5 intent clusters.

For EACH intent cluster, build a full CPP / CSL package:
- a unique value-prop headline (max 30 chars, pulled from cluster
  language)
- a unique subheadline (max 60 chars)
- a screenshot copy set (3 frames)
- a recommended ad group or campaign pairing
- a recommended keyword theme to target this CPP with
- a 1-sentence success metric to measure

End with a rollout plan:
- which CPP to launch first
- which to test in PPO before scaling
- which to retire if the cluster is shrinking

Intent clusters:

<paste>

Pro tip: CPPs are the only place in the App Store where you can show different first screenshots to different intent audiences (AppTweak, Jan 19 2026). If you’re not running 3+ CPPs in 2026, you’re leaking conversion to competitors who are.

Comparison table - prompt categories vs. review layer vs. output

Prompt range	Review layer covered	Primary output	Best run on	Cadence
1–5 (Theme & topic mining)	Topics	Theme taxonomy, semantic clusters, trend map	200–1,000 reviews	Weekly
6–10 (Sentiment & emotion)	Emotions	Emotion tags, mixed-sentiment list, moment-of-joy copy	200–500 reviews	Weekly
11–15 (Churn & competitors)	Churn signals	Uninstall intent score, competitor mentions, refund risks, attack detection	Last 30–90 days	Weekly
16–20 (Feature requests)	Topics + intent	Deduplicated request list, RICE scores, JTBD jobs, premium bait	Last 90 days	Monthly
21–25 (Replies & community)	Replies	Per-review replies, tone guide, crisis reply, Slack digest	This week’s reviews	Weekly / per-review
26–30 (ASO updates)	Topics + emotions → metadata	Title, subtitle, screenshot copy, long description, Apple Search Ads, CPPs	Last 90 days of 4–5 star reviews	Monthly

How to read this table: the left column is what you run, the middle column is which layer of the 4-layer framework it serves, the right column tells you when. If you only have time to do one row, do the bottom two. They are where review language directly becomes revenue.

A 30-day “review sprint” workflow

Most founders read reviews when they remember to. That’s a recipe for slow, reactive decisions. Here’s the 30-day loop I run with my own apps and with clients. Total time: 4–5 hours a month.

Week 1 - Triage and themes (90 min)

Pull the last 30 days of reviews from App Store Connect, Google Play, AppFollow, or AppTweak. Aim for at least 300 reviews.
Run Prompt #1 (theme taxonomy) and Prompt #5 (trend vs. 60 days ago). Save both outputs.
Run Prompt #3 if you shipped an update in the last 30 days.

Week 2 - Sentiment and churn (90 min)

Run Prompt #6 (mixed-sentiment) on the last 7 days of reviews. Reply to the top 10 manually.
Run Prompt #11 (uninstall intent) on the same 7 days. Save your “save list” and reply within 24 hours.
Run Prompt #14 (review-bomb check) on the last 72 hours.

Week 3 - Feature requests and competitors (90 min)

Run Prompt #16 (deduplicate) on 90 days of reviews.
Run Prompt #17 (RICE) on the deduped list.
Run Prompt #12 (competitor mentions). Update your positioning doc.

Week 4 - ASO updates and copy (120 min)

Run Prompt #26 (title/subtitle) and Prompt #27 (screenshots) on 4–5 star reviews.
Run Prompt #30 (CPPs) if you have a CPP program in Apple Search Ads.
Run Prompt #25 (Slack digest) to wrap the month. Pin it in your team channel.

That’s it. Repeat monthly. By month 3, you’ll have a clear before/after, and your roadmap will start to look like a review of itself.

Common mistakes to avoid

I see the same five mistakes in almost every founder’s first attempt. Save yourself the time.

Mistake 1 - Pasting the review text without dates and versions. Without dates, you cannot detect trends. Without versions, you cannot triage releases. Always re-export with both. If your tool does not give you versions, scrape them from the Apple RSS feed or use the App Store Connect API.

Mistake 2 - Asking ChatGPT to give you the “top 3 issues.” It will lie. It will give you the top 3 issues it thinks you want to hear, not the real top 3 issues from your data. Use Prompt #1 to extract the full taxonomy, then sort by share. The truth is rarely the first thing the model says.

Mistake 3 - Replying to every review with AI copy. Reply effect is a per-message metric. Toca Boca’s team pulled back automations that hurt ratings (AppFollow, Feb 12 2026). Run auto-replies in approval mode for the first 60 days. Track reply effect weekly. Cut the replies that hurt, scale the ones that help.

Mistake 4 - Treating reviews as a marketing problem. They are not. They are a product problem wearing a marketing costume. If your top 3 themes are crashes, login, and paywall confusion, no amount of ASO will save you. Fix the product, then refresh the metadata. That order is the difference between App Store top 10 and an app that quietly dies at 3.6 stars.

Mistake 5 - Ignoring silent opportunities. Most feature requests are loud, repetitive, and often already on your competitor’s roadmap. The real win lives in the “I just export manually because I assume the app can’t” reviews. Run Prompt #20 monthly. It will change the way you prioritize.

Final word - make the firehose your strategy

Your reviews are the most under-used asset in your app. They are free, always-on, multilingual, and brutally honest. Most of your competitors are not reading them. Most of the ones who are reading them are not acting on them. That is your moat for 2026.

The 30 prompts above are not a one-time project. They are a system. Run them on a cadence, store the outputs, compare them month over month, and your roadmap, your ASO, and your reply quality will all start to compound. Three months in, your app will feel different in the reviews. Six months in, your rating will move. Twelve months in, you will have the kind of compounding review-driven growth that money cannot buy.

If you want to go further, the natural next step is to pipe this workflow into AppFollow, AppTweak, or Appfigures for the data layer, and use ChatGPT for the strategy layer. Most teams I work with end up using ChatGPT for the prompts in this article and AppFollow for the always-on monitoring and reply automation. That combo is what Toca Boca used to lift its App Store rating from 3.9 to 4.1 and add 26,700 stars to its overall rating in a single half (AppFollow, Feb 12 2026). It is not magic. It is a system. Build yours today.

Reader disclosure & educational-purpose notice

This page is published by SuperFreshAI for general informational and educational purposes only. By reading it, you agree to the points below.

Editorial independence. All reviews, guides, and recommendations are written by our editorial team based on hands-on use. Some links on this site are affiliate links, and some articles are produced as partner content — both are always clearly labeled. Our editorial conclusions are never shaped by partners or affiliates.
Not professional advice. Nothing on this page constitutes legal, financial, medical, tax, or other professional advice. AI tools, pricing, and capabilities change quickly — always verify current information with the tool's official documentation before making a decision.
Educational purpose only. The content here is intended to help you learn about AI tools and workflows. It is not a guarantee of results, performance, fitness for a particular purpose, or suitability for your specific situation. Your results may vary.
No warranties. The site and its content are provided on an "as is" and "as available" basis. We make no warranties, express or implied, about accuracy, completeness, reliability, or availability. See our Terms and Privacy for the full legal terms.
Your responsibility. You are responsible for how you use the information on this page, including any decisions you make based on it. Always do your own research and consult a qualified professional when appropriate.
Affiliate & partner disclosure. When you click certain outbound links, we may earn a commission at no extra cost to you. When a piece of content is produced as partner content, it is labeled at the top of the page. See our Editorial Policy for the full standards we follow.

By continuing to read, you acknowledge that you have read and understood this notice.

30 ChatGPT prompts for mobile app founders to analyze App Store reviews

TL;DR - the cheat sheet

Why your reviews are the cheapest research you have

The 4-layer review analysis framework

SECTION 1 - Theme & topic mining prompts (Prompts 1–5)

Prompt 1 - The “give me my themes” prompt

Prompt 2 - Semantic cluster audit (AppTweak-style)

Prompt 3 - The “first 30 days” new-version triage

Prompt 4 - Multi-language review clustering

Prompt 5 - Trend detection across two time windows

SECTION 2 - Sentiment & emotion prompts (Prompts 6–10)

Prompt 6 - Mixed-sentiment reviewer detection

Prompt 7 - Emotion taxonomy builder

Prompt 8 - Sentiment by app version (the post-release mood)

Prompt 9 - Sentiment by country (your localization report card)

Prompt 10 - “Moment of joy” extraction

SECTION 3 - Churn-signal & competitor prompts (Prompts 11–15)

Prompt 11 - Uninstall intent scoring

Prompt 12 - Competitor mention miner

Prompt 13 - Subscription and refund risk detector

Prompt 14 - Review-bomb and coordinated-attack detection

Prompt 15 - Cohort churn detector (by signup month)

SECTION 4 - Feature-request extraction prompts (Prompts 16–20)

Prompt 16 - Feature request deduplicator

Prompt 17 - RICE-scored feature shortlist

Prompt 18 - JTBD jobs-to-be-done lens

Prompt 19 - Willingness-to-pay detector

Prompt 20 - Request frequency vs. silence frequency

SECTION 5 - Review reply & community prompts (Prompts 21–25)

Prompt 21 - Per-review reply drafter

Prompt 22 - Tone of voice calibrator

Prompt 23 - Public-relations reply for a viral complaint

Prompt 24 - Community manager reply for positive reviews

Prompt 25 - Internal Slack digest generator

SECTION 6 - ASO update prompts (Prompts 26–30)

Prompt 26 - Title and subtitle rewriter from review language

Prompt 27 - Screenshot copy from review “moments of joy”

Prompt 28 - Long description rewriter for semantic relevance

Prompt 29 - Apple Search Ads keyword and Creative Set builder

Prompt 30 - Custom Product Page generator per intent cluster

Comparison table - prompt categories vs. review layer vs. output

People Also Ask - 10 questions founders actually type

1. What are the best ChatGPT prompts for App Store review analysis?

2. Can ChatGPT really analyze App Store reviews accurately?

3. How many reviews do I need to analyze for ChatGPT to find useful patterns?

4. How often should I run review analysis with ChatGPT?

5. What is the difference between AppFollow and using ChatGPT directly?

6. Do I need to use ChatGPT Plus or is the free version fine?

7. How do I get reviews out of App Store Connect for ChatGPT?

8. Can ChatGPT write App Store replies that don’t sound like AI?

9. What should I do if my reviews are mostly negative?

10. How does Google detect and remove fake reviews in 2026?

A 30-day “review sprint” workflow

Common mistakes to avoid

Final word - make the firehose your strategy

SuperFresh AI

43 ChatGPT prompts for non-native English speakers to polish interview answers

41 ChatGPT prompts for SaaS founders in San Francisco to map local partnership opportunities

How to Detect AI-Generated Content

What Is the Best AI Tool for Writing?

AI Newsletter Writing Guide

Get practical AI insights in your inbox