Research Study

The AI App Discoverability Index 2026

How four AI platforms recommend mobile apps, and what the patterns reveal for app teams.

K
Kevser Imirogullari
· · 22 min read
195
queries tested
4
AI platforms
4,265
recommendations analyzed
16.2%
apps visible on all 4
Table of contents

Executive Summary

We sent 195 app discovery queries to four AI platforms (ChatGPT, Claude, Gemini, and Perplexity) and analyzed the 4,265 app recommendations they returned. The study covered 15 app categories with a structured query taxonomy designed to capture how different question types produce different results.

Key observations from this snapshot:

  • Price-qualified queries shift the entire recommendation set. Adding "free" to a query changes the share of free apps from 31.9% to 70.5%, a 38.7 percentage point swing. Only 2 of the top 20 apps overlap between price-qualified and unqualified queries.
  • Perplexity's citation sources are traceable. 62% of Perplexity's 1,285 citation URLs point to editorial content. A small number of publications account for a disproportionate share.
  • Cross-platform visibility is the exception. Only 199 of 1,230 unique apps (16.2%) appeared on all four platforms. 674 apps (54.8%) appeared on only one.
  • Some recommendations reference discontinued or renamed products. Mint, shut down by Intuit in March 2024, appeared 21 times across three platforms.

This is a single snapshot collected in March 2026, not a longitudinal analysis. Full methodology and limitations are detailed at the bottom of this page.

Price-Qualified Queries Produce a Different App Universe

This is the cleanest finding in the study.

When users add a price filter to their query ("best free budgeting app" instead of "best budgeting app"), the resulting recommendations shift dramatically.

Pricing model Price-qualified queries Other queries Shift
Free 70.5% 31.9% +38.7pp
Freemium 28.0% 56.7% -28.7pp
Paid 1.5% 11.4% -10.0pp

How "free" changes the recommendation mix

Standard queries

32%
57%
11%

Price-qualified queries (e.g., "best free budgeting app")

71%
28%
Free Freemium Paid

The price qualifier is not confounded with other query dimensions in our study design, and the effect is large and consistent across categories.

Among the top 20 recommended apps in price-qualified queries vs non-price queries, only 2 overlap (Duolingo and Khan Academy). The other 18 are completely different apps. Price-qualified queries do not filter the same list; they produce a different list.

Top 10: price-qualified vs standard queries

Standard queries

1. YNAB
2. Todoist
3. PocketGuard
4. Spotify
5. MyFitnessPal
6. WhatsApp
7. Headspace
8. Notion
9. Duolingo
10. Khan Academy

Price-qualified queries

1. Tubi
2. Pluto TV
3. Nike Training Club
4. HERE WeGo
5. MediBang Paint
6. Khan Academy
7. Crackle
8. Insight Timer
9. Google Maps
10. OsmAnd

Highlighted apps appear in both lists. Only 2 of the full top 20 overlap.

We identified 15 apps where 40% or more of their total mentions came from price-qualified queries. These include Crackle, Kanopy, Smiling Mind, Pluto TV, OsmAnd, and HERE WeGo. All are free or freemium. They are largely absent from non-price queries but appear consistently when price is specified.

What this suggests for app teams: If your app is free, price-qualified queries represent a distinct discovery channel where competition is different from the general recommendation pool. If your app is paid, it will be largely absent from this channel. Making your pricing explicitly clear in your web presence may influence which recommendation pool you land in.

Perplexity's Citation Sources Are Traceable and Concentrated

Perplexity is the only platform in our study that returns citation URLs with its responses. We analyzed 1,285 citation URLs across 195 Perplexity responses.

Source type Count % of citations
Editorial content (listicles, reviews, roundups) 802 62.4%
YouTube 183 14.2%
App store listings (Google Play, App Store) 122 9.5%
Other (app websites, niche sites, forums) 178 13.9%

Top 8 Editorial Domains by Citation Count

Domain Citations Categories covered
zapier.com 31 6 (productivity, business, news, etc.)
nerdwallet.com 17 3 (primarily finance)
techradar.com 16 8
garagegymreviews.com 13 3 (primarily fitness)
thedigitalprojectmanager.com 9 2 (productivity, business)
fortune.com 8 3
computerworld.com 7 2
creativebloq.com 7 2 (art, design)

YouTube is the second-largest citation source at 14.2%. YouTube citations are most concentrated in Entertainment (36% of that category's citations) and Creativity/Art (29%).

We also found 13 instances where an app's own website was cited by Perplexity in a response that recommended that same app. Mealime had the most self-citations (4), followed by Canva and Goodbudget (3 each). This suggests that well-structured content on your own domain can directly feed your Perplexity recommendation presence.

Important scoping note: This analysis describes Perplexity's retrieval sources specifically. Perplexity uses real-time web search to supplement its model. ChatGPT, Claude, and Gemini do not expose citation data through their APIs, and their recommendation sources may differ. These findings should not be generalized to "what feeds AI recommendations" broadly.

What this suggests for app teams: Being featured in editorial listicles on sites like Zapier, NerdWallet, or TechRadar may influence Perplexity's recommendations. Whether similar editorial presence influences other platforms' training data is plausible but unconfirmed by this study.

Cross-Platform Visibility Is the Exception, Not the Rule

Of 1,230 unique normalized apps in our dataset, only 199 (16.2%) appeared on all four platforms. 674 apps (54.8%) appeared on only one platform.

Platforms Apps %
All 4 199 16.2%
3 148 12.0%
2 209 17.0%
1 only 674 54.8%

Apps by number of AI platforms they appeared on

674
54.8%
1 platform
209
17.0%
2 platforms
148
12.0%
3 platforms
199
16.2%
All 4

1,230 unique apps. Over half appeared on only one platform.

88 apps appeared on a single platform with 2 or more mentions, suggesting they are genuinely favored by one model but absent from the others.

Single-platform exclusives (2+ mentions, one platform only)

Perplexity
36
Gemini
22
ChatGPT
16
Claude
14

88 apps total that appeared on only one platform with 2+ mentions.

The 54.8% single-platform figure is sensitive to normalization. Some single-platform apps may be the same app under different names that our normalization did not catch. Additionally, our 13 queries per category do not exhaust all possible queries. An app that did not appear in our query set may still appear on a platform for queries we did not test.

What this suggests for app teams: Testing your app's visibility on one platform does not reliably predict its visibility on others. If cross-platform presence matters to you, test each platform independently.

Query Type Shifts Which Apps Get Recommended

When we compared recommendations across different intent and specificity types, we observed consistent shifts in which apps appeared.

Intent and Framing Effect

270 apps (22.0% of all unique apps) appeared only in use-case queries with problem-solving or exploratory framing, and never in direct/comparison queries with recommendation framing. This suggests that different query approaches surface different parts of the AI's knowledge.

Because framing and intent are confounded in our query design (see Methodology), we cannot determine whether this difference is driven by the intent type, the framing style, or both.

Specificity Effect

Broad queries surfaced 198 unique apps, of which 34.3% were niche (3 or fewer total mentions). More specific queries (medium and narrow) surfaced 64-66% niche apps. This suggests that specific queries create more room for lesser-known apps, while broad queries tend to favor well-known options.

The effect is a step function (broad vs non-broad) rather than a smooth gradient. The difference between medium (64.4%) and narrow (65.6%) specificity was minimal.

How query specificity affects which apps surface

Share of lesser-known apps (3 or fewer total mentions) in results

Broad
34.3%
Medium
64.4%
Narrow
65.6%

The jump is from broad to non-broad. Medium and narrow produce similar results.

What this suggests for app teams: The way users phrase their questions matters. A broad "best fitness app" query likely returns established brands. A specific "best free yoga app for beginners at home" query may surface different apps. Building web content that addresses specific use cases and problems, rather than competing for broad "best app" positioning, may improve your chances of appearing in the relevant recommendation set.

Some Recommendations Reference Discontinued or Renamed Products

We identified 5 stale or outdated entity references in the dataset, totaling 67 mentions (1.6% of all mentions). We categorize these by subtype to avoid conflating different phenomena.

Stale Reference Mentions Platforms Subtype Current Status
Mint 21 ChatGPT, Claude, Gemini Discontinued Shut down by Intuit March 2024
Personal Capital 17 All 4 Renamed Product still exists as Empower Personal Dashboard (rebranded 2023)
Mealime 15 Claude, Gemini, Perplexity Uncertain May have reduced operations; status unconfirmed
Autodesk SketchBook 13 ChatGPT, Gemini, Perplexity Divested App still exists but was divested by Autodesk in 2021; now maintained by Sketchbook Inc.
Google Trips 1 ChatGPT Discontinued Shut down by Google August 2019

Mint is the most notable case. Intuit shut down Mint over two years before our data collection, yet three platforms still recommended it 21 times across our queries. When it was mentioned, it averaged a position of 1.7, making it one of the highest-positioned apps in the Finance category. AI models are not just mentioning it; they are actively recommending it as a top choice.

The Personal Capital case is different: the product itself still exists (now branded as Empower Personal Dashboard), but some platforms continue to reference it under the legacy name.

These represent a small share of total recommendations (1.6%), but they illustrate that AI recommendations can reflect historical training data and may lag behind product changes by months or years.

Recommendation Frequency Follows a Power Law

The top 100 apps (8.1% of all unique apps) accounted for 38.1% of all mentions. 586 apps (47.6%) were mentioned exactly once. 100 apps were mentioned 10 or more times.

This distribution is consistent with how language models work: apps that appear more frequently in training corpora are more likely to be recommended. It is also consistent with app store dynamics, where a small number of apps capture most visibility.

We present this as structural context, not as a novel finding. It sets the landscape for the more specific observations above.

Exploratory Signals

The following observations are based on smaller samples, exploratory analyses, or correlations with known confounds. They are included for completeness and as potential directions for future research, but should not be treated as established findings.

Web Presence Signals (n=99 audited apps)

We audited 99 of the top 100 most-recommended apps for web presence signals. All correlations below are exploratory and likely confounded with overall brand strength and digital investment.

llms.txt: 27 of 99 audited apps have an llms.txt file. These apps averaged 17.9 mentions vs 14.6 for apps without (+22.9%). They also had a better average position (2.6 vs 2.9). This correlation likely reflects broader digital sophistication rather than a direct causal effect of llms.txt adoption.

SoftwareApplication or WebApplication schema: 9 of 99 audited apps use app-specific schema markup. These apps had a notably better average position (2.3 vs 2.8), the largest position difference of any signal we checked. However, n=9 is too small for reliable inference.

AI crawler blocking: 12 of 99 audited apps block one or more AI crawlers via robots.txt. These apps averaged 10.8% fewer mentions than non-blockers (14.0 vs 15.7). We cannot determine whether blocking reduces visibility or whether this reflects other characteristics of the apps that block.

Multi-Category Apps

75 apps appeared in recommendations across 2 or more categories. These averaged 8.4 mentions and 3.0 platforms, compared to 3.0 mentions and 1.8 platforms for single-category apps. This is a correlation, not a causal observation. Popular apps are more likely to be relevant across categories.

Category Reference Tables

Select a category to see the top 10 AI-recommended apps, with platform coverage and average recommendation position.

Business / Professional Tools 124 unique apps | 14 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1Trello24ChatGPT, Claude, Gemini, Perplexity2.210
2Notion23ChatGPT, Claude, Gemini, Perplexity3.26
3Asana18ChatGPT, Claude, Gemini, Perplexity2.34
4Evernote15ChatGPT, Gemini, Perplexity3.32
5FreshBooks13ChatGPT, Claude, Gemini, Perplexity3.10
6ClickUp12ChatGPT, Claude, Gemini, Perplexity4.10
7Zoho Invoice11ChatGPT, Claude, Gemini, Perplexity2.33
8Toggl Track11ChatGPT, Claude, Gemini, Perplexity1.66
9Zoom10ChatGPT, Claude, Perplexity2.54
10Wave9ChatGPT, Claude, Perplexity1.65
Creativity / Art & Design 88 unique apps | 12 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1CapCut22ChatGPT, Claude, Gemini, Perplexity1.911
2Snapseed20ChatGPT, Claude, Gemini, Perplexity2.36
3Canva17ChatGPT, Claude, Gemini, Perplexity2.19
4Procreate16ChatGPT, Claude, Gemini, Perplexity1.214
5Adobe Fresco16ChatGPT, Claude, Gemini, Perplexity2.91
6InShot16ChatGPT, Claude, Gemini, Perplexity3.03
7MediBang Paint14ChatGPT, Claude, Gemini, Perplexity3.61
8VN Video Editor14ChatGPT, Claude, Gemini, Perplexity3.12
9ibisPaint X13ChatGPT, Claude, Gemini, Perplexity4.00
10Autodesk SketchBook *13ChatGPT, Gemini, Perplexity2.35
* Autodesk SketchBook was divested from Autodesk in 2021 and is now maintained by Sketchbook Inc.
Education / Learning 93 unique apps | 14 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1Khan Academy28ChatGPT, Claude, Gemini, Perplexity2.216
2Duolingo27ChatGPT, Claude, Gemini, Perplexity2.013
3Coursera23ChatGPT, Claude, Gemini, Perplexity2.37
4Procreate16ChatGPT, Claude, Gemini, Perplexity1.214
5Udemy13ChatGPT, Claude, Gemini, Perplexity4.20
6Skillshare11ChatGPT, Claude, Gemini, Perplexity4.21
7Quizlet11ChatGPT, Claude, Gemini, Perplexity3.50
8Memrise10ChatGPT, Claude, Gemini, Perplexity4.60
9edX10ChatGPT, Claude, Gemini, Perplexity3.20
10Zoom10ChatGPT, Claude, Perplexity2.54
Entertainment / Streaming 80 unique apps | 15 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1Spotify31ChatGPT, Claude, Gemini, Perplexity1.822
2Tubi18ChatGPT, Claude, Gemini, Perplexity2.19
3Pluto TV14ChatGPT, Claude, Gemini, Perplexity2.82
4Netflix11ChatGPT, Claude, Gemini, Perplexity1.210
5Amazon Prime Video11ChatGPT, Claude, Gemini, Perplexity2.80
6Disney+10ChatGPT, Claude, Gemini, Perplexity2.90
7Hulu8ChatGPT, Claude, Gemini3.90
8Pocket Casts8ChatGPT, Claude, Gemini, Perplexity3.01
9Freevee7Claude, Gemini, Perplexity4.30
10Crackle7ChatGPT, Claude, Gemini2.91
Finance / Budgeting 74 unique apps | 6 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1YNAB34ChatGPT, Claude, Gemini, Perplexity2.012
2PocketGuard34ChatGPT, Claude, Gemini, Perplexity3.72
3Goodbudget24ChatGPT, Claude, Gemini, Perplexity3.61
4Monarch Money21Claude, Gemini, Perplexity2.67
5EveryDollar21ChatGPT, Claude, Gemini, Perplexity3.82
6Mint *21ChatGPT, Claude, Gemini1.712
7Empower Personal Dashboard *17ChatGPT, Claude, Gemini, Perplexity3.42
8Wave9ChatGPT, Claude, Perplexity1.65
9Yahoo Finance7ChatGPT, Claude, Gemini, Perplexity3.02
10Adobe Scan6ChatGPT, Claude, Gemini, Perplexity1.55
* Mint was discontinued by Intuit in March 2024. Empower Personal Dashboard was formerly Personal Capital (renamed 2023).
Fitness / Workout 85 unique apps | 11 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1MyFitnessPal26ChatGPT, Claude, Gemini, Perplexity2.211
2Nike Training Club25ChatGPT, Claude, Gemini, Perplexity2.211
3Strava20ChatGPT, Claude, Gemini, Perplexity3.07
4Nike Run Club15ChatGPT, Claude, Gemini, Perplexity2.42
5FitOn14ChatGPT, Claude, Gemini, Perplexity2.92
6Fitbit12ChatGPT, Claude, Gemini3.40
7Peloton11Claude, Gemini3.51
8Lose It!10ChatGPT, Claude, Gemini, Perplexity2.71
9JEFIT9Claude, Perplexity2.93
10Hevy9Claude, Gemini, Perplexity4.01
Food & Drink / Cooking 87 unique apps | 15 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1MyFitnessPal26ChatGPT, Claude, Gemini, Perplexity2.211
2Paprika Recipe Manager16ChatGPT, Claude, Gemini, Perplexity3.92
3Mealime *15Claude, Gemini, Perplexity1.97
4Tasty15ChatGPT, Claude, Gemini, Perplexity2.45
5Allrecipes Dinner Spinner13ChatGPT, Claude, Gemini2.83
6Eat This Much12ChatGPT, Claude, Gemini, Perplexity3.22
7Lose It!10ChatGPT, Claude, Gemini, Perplexity2.71
8Yummly10ChatGPT, Claude, Gemini, Perplexity3.11
9Instacart9ChatGPT, Claude, Gemini, Perplexity3.14
10Cronometer8ChatGPT, Claude, Gemini, Perplexity2.62
* Mealime's operational status is uncertain.
Health / Wellness 120 unique apps | 10 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1MyFitnessPal26ChatGPT, Claude, Gemini, Perplexity2.211
2Headspace25ChatGPT, Claude, Gemini, Perplexity2.26
3Calm22ChatGPT, Claude, Gemini, Perplexity1.99
4Strava20ChatGPT, Claude, Gemini, Perplexity3.07
5Insight Timer17ChatGPT, Claude, Gemini, Perplexity2.76
6Fitbit12ChatGPT, Claude, Gemini3.40
7Peloton11Claude, Gemini3.51
8Lose It!10ChatGPT, Claude, Gemini, Perplexity2.71
9JEFIT9Claude, Perplexity2.93
10Apple Health8ChatGPT, Claude, Gemini2.54
Music / Audio 107 unique apps | 15 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1Spotify31ChatGPT, Claude, Gemini, Perplexity1.822
2Apple Music15ChatGPT, Claude, Gemini, Perplexity2.71
3YouTube Music15ChatGPT, Claude, Gemini, Perplexity3.01
4GarageBand12ChatGPT, Claude, Gemini, Perplexity1.58
5FL Studio Mobile11ChatGPT, Claude, Gemini, Perplexity2.11
6Pandora9ChatGPT, Gemini, Perplexity3.80
7SoundCloud8ChatGPT, Claude, Gemini, Perplexity3.90
8BandLab7Claude, Gemini, Perplexity3.01
9Yousician7ChatGPT, Claude, Gemini, Perplexity2.03
10Amazon Music7ChatGPT, Gemini, Perplexity4.30
News / Reading 71 unique apps | 15 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1Notion23ChatGPT, Claude, Gemini, Perplexity3.26
2Google News23ChatGPT, Claude, Gemini, Perplexity1.913
3Flipboard18ChatGPT, Claude, Gemini, Perplexity3.11
4SmartNews17ChatGPT, Claude, Gemini, Perplexity3.51
5Feedly16ChatGPT, Claude, Gemini, Perplexity2.96
6Reuters15ChatGPT, Claude, Gemini, Perplexity3.21
7Ground News15ChatGPT, Claude, Gemini, Perplexity3.32
8Evernote15ChatGPT, Gemini, Perplexity3.32
9BBC News14ChatGPT, Claude, Gemini, Perplexity2.91
10Apple News13ChatGPT, Claude, Gemini, Perplexity3.03
Photo & Video 92 unique apps | 14 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1CapCut22ChatGPT, Claude, Gemini, Perplexity1.911
2Snapseed20ChatGPT, Claude, Gemini, Perplexity2.36
3Canva17ChatGPT, Claude, Gemini, Perplexity2.19
4InShot16ChatGPT, Claude, Gemini, Perplexity3.03
5VN Video Editor14ChatGPT, Claude, Gemini, Perplexity3.12
6Adobe Lightroom14ChatGPT, Claude, Gemini, Perplexity1.49
7VSCO13ChatGPT, Claude, Gemini, Perplexity3.70
8Adobe Premiere Rush12ChatGPT, Claude, Gemini, Perplexity3.61
9iMovie12ChatGPT, Claude, Gemini, Perplexity3.03
10PhotoDirector11ChatGPT, Claude, Gemini, Perplexity4.51
Productivity / Task Management 58 unique apps | 14 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1Todoist33ChatGPT, Claude, Gemini, Perplexity2.613
2Trello24ChatGPT, Claude, Gemini, Perplexity2.210
3Notion23ChatGPT, Claude, Gemini, Perplexity3.26
4Microsoft To Do22ChatGPT, Claude, Gemini, Perplexity2.93
5TickTick19ChatGPT, Claude, Gemini, Perplexity3.31
6Asana18ChatGPT, Claude, Gemini, Perplexity2.34
7Evernote15ChatGPT, Gemini, Perplexity3.32
8ClickUp12ChatGPT, Claude, Gemini, Perplexity4.10
9Toggl Track11ChatGPT, Claude, Gemini, Perplexity1.66
10Any.do11ChatGPT, Claude, Gemini, Perplexity4.60
Shopping / Deals 65 unique apps | 15 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1Rakuten23ChatGPT, Claude, Gemini, Perplexity2.59
2Ibotta18ChatGPT, Claude, Gemini, Perplexity2.27
3Honey15ChatGPT, Claude, Gemini, Perplexity2.16
4eBay13ChatGPT, Claude, Gemini, Perplexity2.93
5Fetch Rewards12ChatGPT, Claude, Gemini, Perplexity3.20
6RetailMeNot11ChatGPT, Claude, Gemini, Perplexity3.02
7Poshmark10ChatGPT, Claude, Gemini, Perplexity2.81
8Mercari9ChatGPT, Claude, Gemini, Perplexity4.40
9Depop9ChatGPT, Claude, Gemini, Perplexity2.44
10Walmart8ChatGPT, Claude, Gemini, Perplexity3.60
Social / Communication 81 unique apps | 15 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1WhatsApp29ChatGPT, Claude, Gemini, Perplexity2.215
2Signal18ChatGPT, Claude, Gemini, Perplexity3.08
3Telegram17ChatGPT, Claude, Gemini, Perplexity2.61
4Discord14ChatGPT, Claude, Gemini, Perplexity3.63
5Meetup12ChatGPT, Claude, Gemini, Perplexity1.84
6Google Meet11ChatGPT, Claude, Gemini, Perplexity2.91
7Bumble11ChatGPT, Claude, Gemini, Perplexity1.66
8Zoom10ChatGPT, Claude, Perplexity2.54
9Microsoft Teams9ChatGPT, Claude, Perplexity4.20
10Nextdoor9ChatGPT, Claude, Gemini, Perplexity4.30
Travel / Navigation 88 unique apps | 20 on all 4 platforms
# App Mentions Platforms Avg Pos #1 Count
1Google Maps24ChatGPT, Claude, Gemini, Perplexity1.814
2Skyscanner13ChatGPT, Claude, Gemini, Perplexity2.74
3MAPS.ME13ChatGPT, Claude, Gemini, Perplexity2.63
4HERE WeGo12ChatGPT, Claude, Gemini, Perplexity3.40
5Hopper9ChatGPT, Claude, Gemini, Perplexity3.60
6OsmAnd9ChatGPT, Claude, Gemini, Perplexity4.10
7Dosh8ChatGPT, Claude, Gemini, Perplexity4.20
8KAYAK8ChatGPT, Claude, Gemini, Perplexity3.10
9Waze8ChatGPT, Claude, Perplexity2.50
10Rome2rio8ChatGPT, Claude, Gemini, Perplexity4.20

Methodology

Query Design

195 queries were designed across 15 app categories (13 queries per category). Each query was classified along four dimensions:

Dimension Values What It Captures
Intent direct, use_case, comparison How the user frames the request
Specificity broad, medium, narrow How targeted the query is
Qualifier none, feature, audience, price, platform, time What filter is applied
Framing recommendation, problem_solving, exploratory The conversational style

Design constraint: Framing and intent are not independently varied. Recommendation framing co-occurs only with direct and comparison intent; problem_solving and exploratory framing co-occur only with use_case intent. This means framing effects cannot be cleanly separated from intent effects in this study.

Platforms and Models

Platform Model Access
ChatGPT gpt-4o OpenAI API
Claude claude-sonnet-4-6 Anthropic API
Gemini gemini-2.0-flash Google Gemini API
Perplexity sonar Perplexity API

All four platforms received a system prompt instructing the model to recommend 4-6 apps per query, formatted as structured JSON with app name, description, and pricing label (free, freemium, or paid). The Perplexity script additionally appended "recommend 4-6 specific mobile apps with app name and one sentence why" to the user query text.

The number of apps per response (predominantly 5-6) reflects this prompt constraint. It should not be interpreted as a natural property of AI app recommendations.

Data Processing

Stage Count
Raw response files780 (195 queries x 4 platforms)
Successfully parsed responses778
Failed parses2 (Perplexity BIZ-01 and SHP-09)
Total app mentions analyzed4,265
Unique raw app name strings1,647
Canonical name map entries148
Unique normalized app names1,230
Names merged by normalization417

App names were normalized in two stages: first by matching against a 148-entry canonical name map, then by automatic subtitle stripping. A systematic duplicate scan was conducted across all 1,230 normalized names to identify remaining variants (e.g., "WhatsApp" vs "WhatsApp Messenger"). The top 100 apps were manually verified for false merges.

Limitations

  • Snapshot, not longitudinal. Data collected March 19-20, 2026. AI models update continuously.
  • Constrained output format. Responses were constrained to 4-6 apps in JSON format. Natural conversational responses may differ.
  • English only. No regional, language, or cultural variation was tested.
  • Single-turn queries. No follow-up questions, conversation history, or personalization.
  • Normalization sensitivity. The 1,230 figure reflects best-effort normalization but should be treated as approximate.
  • Confounded dimensions. Framing and intent are not independently varied.
  • Web presence audit coverage. Results are correlational and exploratory, based on automated checks of 99 apps.

Data Availability

The following materials are available for download:

Full raw dataset (780 individual response files) and collection scripts are available on request: hello@growthbykev.com.

Citation

Imirogullari, K. (2026). The AI App Discoverability Index 2026: How Four AI Platforms Recommend Mobile Apps. growthbykev.com.

What To Do With These Findings

This study was conducted independently. No AI company sponsored, reviewed, or influenced the research or findings.

AI Discoverability Research Mobile Apps LLM
Share:
K

Written by Kevser Imirogullari

Independent mobile marketing consultant helping apps by connecting acquisition, store, and monetization insights they missed.

Explore free tools →

Get notified when we publish the next edition

Join 500+ app marketers getting weekly tips on ASO, Apple Search Ads, and mobile growth.

No spam. Unsubscribe anytime.

Newsletter

Weekly mobile growth insights

What I'm seeing inside real app growth work, before it becomes common advice.

Subscribe

or get in touch