App Store Screenshot Keyword Research: 6-Step Workflow

App Store screenshot keyword research is a six-step workflow that runs in reverse of how most ASO guides describe it. Start by inventorying the title, subtitle, and keyword field you already have. Pull rank data on those existing keywords. Mine 5 competitor captions via OCR. Filter candidates against the reinforcement rule from ConsultMyApp's 64-phrase test (only 1 of 64 caption phrases ranked without metadata explanation [1]). Score the survivors for OCR confidence. Then map one keyword per slot, prioritized by your weakest rank positions.

TL;DR:

• Most ASO guides start with keyword discovery. That order is wrong for screenshots: ConsultMyApp tested 64 caption phrases across 8 leading apps, and 27 of the 36 that ranked were already in metadata, only 1 was unexplained [1].
• The workflow runs inventory first, discovery second. You're looking for keywords that reinforce existing metadata, not introduce new territory.
• Free workflow replacement for AppTweak's $79/month entry tier [3]: Apple's own Search Popularity scores (via Search Ads), App Store search-bar autocomplete, App Store Connect's App Analytics dashboard, the free ASO keyword researcher, and the free indexed-fields map cover what 95% of indie research needs.
• Score candidates on three dimensions, not one: relevance to your app, search volume (Apple Search Popularity 0 to 100 [6]), keyword difficulty [4].
• The caption text must clear OCR confidence: 16-point minimum, high contrast, top or bottom placement, standard sans-serif. Candidates that can't be rendered legibly don't earn the signal regardless of relevance.
• Re-run the full workflow quarterly. Re-run steps 5 and 6 (scoring and mapping) on every screenshot refresh.

This is the process companion to two existing posts. The screenshot SEO keyword strategy guide covers what keywords to put in captions and why captions matter at all. The Apple OCR mechanism deep-dive covers how Apple's text detection actually works and what the ConsultMyApp data really shows. This post covers how you do the research that produces the keywords those two posts assume you already have.

Table of Contents

• What is App Store screenshot keyword research?
• Why do most screenshot keyword research workflows fail?
• How do you research screenshot caption keywords in 6 steps?
• Which free tools can replace a $79/month AppTweak workflow?
• How often should you re-run the keyword research workflow?
• What are the most common mistakes in caption keyword research?
• Takeaways

What is App Store screenshot keyword research?

App Store screenshot keyword research is the process of identifying which specific words and phrases to place inside your screenshot captions so that Apple's text detection picks them up as ranking-signal text. It is a subset of general App Store keyword research, and it operates under a tighter constraint: the keywords have to render legibly at gallery-thumbnail size, in 3 to 6 word phrases, in caption zones (top or bottom of the frame), and they have to reinforce ranking signals you already earned through your title, subtitle, and keyword field [1].

That last constraint is what separates screenshot keyword research from general ASO keyword research. When you're picking title and subtitle keywords, you have 30 characters per field to introduce brand-new keyword bets. When you're picking screenshot caption keywords, ConsultMyApp's data shows that brand-new keyword bets almost never pay off: 36 of their 64 tested phrases didn't rank at all, and 27 of the 36 that ranked were already in metadata [1]. Only 1 was an unexplained anomaly.

So the working definition: screenshot keyword research is the process of picking the subset of your existing keyword universe that benefits most from caption-level reinforcement, then validating those candidates can actually be rendered legibly on the image. Discovery happens, but it happens inside the bounds your metadata already set.

Why do most screenshot keyword research workflows fail?

Most ASO content frames screenshot keyword research as a discovery problem. The pattern: open a keyword research tool, brainstorm 30 candidate terms, score them by volume and difficulty, pick 5 to 10, write captions around them. That's a workflow that produces good general ASO output. It produces poor screenshot caption output for three reasons.

Reason 1: It treats captions as a place to introduce new keywords. They are not. The data shows captions function as a reinforcement layer for metadata you already have [1]. A keyword research run that starts from a blank slate and surfaces 30 fresh candidates will spend most of its output budget on terms that will not rank from captions alone. Indie developers with limited screenshot slots (most apps ship 3 to 10) cannot afford to spend frames on keyword introduction bets that have a 35-out-of-36 chance of producing nothing.

Reason 2: It ignores rank-position weakness as a prioritization signal. A keyword you currently rank #1 for doesn't benefit from caption reinforcement; you've already won. A keyword you rank #15 for has measurable upside if the caption signal nudges you toward page 1. Most workflows don't pull current rank data before running discovery, so they can't sort candidates by where reinforcement matters most.

Reason 3: They score candidates on volume and difficulty but skip OCR confidence. A phrase with high search volume and low difficulty is useless if your screenshot can't render it legibly. Apple's Vision framework filters text below 5 percent of image height [2], drops low-confidence detections on stylized fonts, and weights top-and-bottom caption zones higher than middle-of-frame UI text. A candidate that passes ASO scoring but fails legibility scoring earns nothing.

The fix is sequencing: inventory before discovery, current rank data before candidate scoring, OCR confidence as the last filter. That's the order the 6-step workflow runs in.

How do you research screenshot caption keywords in 6 steps?

The workflow that aligns with the data. Each step takes 15 to 30 minutes the first time you run it. Quarterly refreshes (steps 4 to 6 only) take 20 minutes total.

Step 1: Inventory your title, subtitle, and keyword field

Pull the exact text from your three text metadata fields. Most indies have 80 to 150 characters of unique keyword phrases across the three. Break them into individual phrases: "Pomodoro Timer", "focus sessions", "deep work", "block distractions". Drop generic glue words ("the", "app", "for"). What's left is your reinforcement universe: the keyword set Apple already considers you eligible for. Every caption you write in step 6 will draw from this list.

The fields you care about live in App Store Connect under your app's "Product Page" tab. Copy each one (title, subtitle, keyword field) into a working doc and split on commas, line breaks, and natural phrase boundaries. To confirm which fields actually contribute to search indexing (not all of them do; the long description isn't indexed, for example), use the free indexed-fields map as a reference. The map lists every iOS and Google Play field with its character limit and whether Apple weights it for search.

Step 2: Pull rank data on your existing eligible keywords

For each phrase from step 1, check your current rank position. You're sorting them into four buckets:

• Rank 1-3 (saturated): Caption reinforcement is wasted here. You already win.
• Rank 4-10 (mid-tier): Reinforcement can push you to position 1-3, which is where install-rate gains compound.
• Rank 11-30 (weakest tier with eligibility): Biggest upside. You're already eligible, you just need to climb. This is where captions earn the most.
• Rank 50+ or unranked (eligibility gap): Apple has your keyword indexed but you're so far down that something else is broken (low velocity, weak reviews, bad ratings). Captions won't fix that.

For each phrase, the App Store search bar gives you the authoritative real-user view: type the phrase, see where your app actually lands in the results. Apple's App Analytics dashboard (free, inside App Store Connect) gives you keyword-level impressions and conversion data for your owned app, which is the ground-truth source for which queries are actually driving installs. The free ASO keyword researcher returns volume and difficulty estimates for the phrases you decide to expand into in step 3; it's a discovery tool, not a rank tracker.

Step 3: Mine 5 competitor captions via OCR

Pick the top 5 apps ranking for your single most important keyword (the one you most want to climb from rank 11-30 to top 5). Open each app's App Store page and screenshot the first three frames. Apple weights the top-and-bottom caption zones in the first three frames most heavily [2], so that's where competitor patterns matter.

Extract the visible text from each screenshot using any OCR utility. Free options that work: macOS Live Text (right-click on any image, "Show Live Text"), iOS Live Text in Photos, paste the screenshot into ChatGPT and ask for the visible text, or use the free caption readability checker which surfaces caption text as part of its scoring.

Build a quick comparison table: for each of the 5 competitors, list the caption text on frames 1, 2, 3. Look for 3-to-6-word phrases that repeat across 3 or more competitors. Those are the patterns Apple's index has rewarded for this keyword cluster, and they're your candidate library for step 4.

Ignore generic filler. "Easy to use", "the best", "trusted by millions" appear in everyone's screenshots and have no search volume. They're noise from the design phase, not signal from the keyword phase.

Step 4: Filter candidates against the reinforcement rule

For each candidate phrase from step 3, check whether it overlaps with your reinforcement universe (step 1 output). Three filter outcomes:

• Direct overlap (the candidate phrase contains a word from your title, subtitle, or keyword field): Keep it. This is the strongest signal type per the ConsultMyApp data [1].
• Semantic overlap (the candidate phrase is closely related to a word from your metadata but doesn't share exact tokens): Keep it conditionally. Apple's AI-based detection [2] handles semantic relationships better than legacy OCR did, so "training plan" can reinforce "workout routine" without exact token match. Mark these as second-tier candidates.
• No overlap (the candidate phrase is a brand-new keyword bet): Cut it. ConsultMyApp's 1-of-64 unexplained ranking [1] means caption introduction has roughly a 1.6 percent success rate. That's worse than coin-flip odds on a finite-frame budget.

Score the survivors on the three standard ASO dimensions [4]: relevance to your app's core feature (1-10, scored manually by you), search volume from Apple Search Popularity [6] (which Apple Ads publishes on a 0-100 scale), and keyword difficulty. The free ASO keyword researcher returns volume and difficulty estimates for any candidate phrase. The output is a ranked candidate list, usually 5 to 15 phrases.

Step 5: Score survivors for OCR confidence

A high-relevance, high-volume, low-difficulty phrase that can't be rendered legibly is worth zero. For each candidate from step 4, draft a 3-to-6-word caption sentence that uses the phrase naturally, then test whether the rendered caption will clear Apple's OCR confidence floor.

Five hard rules from the Vision framework's documented behavior [2]:

• 16 points minimum, ideally 18 to 24 points (caption text below 14 points loses confidence even on clean backgrounds)
• Standard sans-serif font (San Francisco, Helvetica, Arial, Inter; no decorative or display faces for captions)
• Text-to-background contrast of WCAG AA (4.5:1 ratio for body text) or better
• Top third or bottom third of the frame (middle competes with UI text and gets weighted lower)
• No drop shadows, no glow effects, no layered text effects that blur character edges

The free caption readability checker scores your draft caption against character count, word count, and rendered font-size-at-thumbnail-scale per device, which are the legibility dimensions that drive OCR confidence at gallery thumbnail size (the same scale Apple's index processes). Candidates that score badly on font-size-at-thumbnail get cut first. The other rules above (contrast, font choice, placement) need a design-eye review the checker doesn't automate; use them as your manual checklist on each candidate.

Step 6: Map one keyword per slot, prioritized by weakest rank

The final mapping. For each remaining screenshot slot (usually 3 to 10 frames), pick one candidate phrase to feature. The priority order:

• Screenshot 1 gets your highest-relevance candidate that you currently rank weakest for (rank 11-30 from step 2). Frame 1 is the highest-impact slot for both conversion and OCR signal weight [2].
• Screenshots 2 and 3 get your next two strongest candidates by the same ranking-weakness criterion. These two also get user attention at near the same rate as frame 1.
• Screenshots 4 through 6 get filler reinforcement of your top-3 candidates (slight rephrasing of the same phrases, not new keywords). Repetition in moderation strengthens the signal [1].
• Screenshot 7+ is optional. One slot in here is a reasonable place to test a single brand-new keyword introduction (the 1-in-64 bet) since the cost of one failed frame is low.

You now have a frame-by-frame caption brief. The free screenshot copy generator produces benefit-focused caption text from a keyword list, which is the step that turns the brief into actual on-screenshot copy.

Which free tools can replace a $79/month AppTweak workflow?

AppTweak's entry tier ASO Intelligence plan runs about $79 per month as of 2026 [3]. MobileAction's equivalent runs higher, and Sensor Tower's published pricing requires annual contracts [3]. For an indie developer running screenshot keyword research on 1 to 3 apps, that monthly spend is hard to justify when free tools cover the workflow steps above.

The free-tool replacement set for the 6-step workflow:

• Step 1 (inventory): Manual copy from App Store Connect or your public listing. The free indexed-fields map is the reference for which fields contribute to search indexing (so you don't waste inventory time on fields that don't count).
• Step 2 (rank data): Manual App Store search for phrase-by-phrase rank position. App Store Connect's App Analytics dashboard for impressions and conversion data on your owned app's queries.
• Step 3 (competitor OCR): macOS Live Text, iOS Photos Live Text, or paste-into-ChatGPT for caption text extraction. No paid tool needed.
• Step 4 (filter and score): The free ASO keyword researcher returns volume and difficulty estimates for any candidate phrase; Apple's Search Popularity 0-100 scale is the same volume signal AppTweak licenses from Apple Ads [6].
• Step 5 (OCR confidence): The free caption readability checker scores character count, word count, and font-size-at-thumbnail-scale per device. The contrast, font, and placement rules are a manual checklist you apply on top.
• Step 6 (mapping): A spreadsheet or the free screenshot copy generator for turning the keyword brief into draft caption text.

AppDrift's published indie playbook reaches the same conclusion: a free-tier toolkit covers the 95 percent case for solo developers [5]. The paid tools earn their cost when you're running ASO at agency scale (50+ apps, 30+ countries, weekly competitive intelligence). For caption research on a single indie app, the free stack is enough.

For the broader catalog of free indie-tier ASO tools that exist outside the screenshot research workflow, the best ASO tools for indies in 2026 guide covers the full landscape.

How often should you re-run the keyword research workflow?

The full 6-step workflow takes 90 to 150 minutes the first time you run it on an app. After that, three refresh cadences make sense:

Quarterly: re-run steps 4 to 6. Your reinforcement universe (step 1) is stable unless you change title, subtitle, or keyword field. Your rank positions (step 2) drift, but quarterly is frequent enough to catch meaningful shifts. The candidate scoring and OCR mapping benefit from a fresh look every 90 days because the SERP and competitor caption text both shift.

On every metadata change: re-run the full workflow. If you ship a new title or subtitle, your reinforcement universe changes, and the entire candidate set from old captions may now be misaligned. Don't ship metadata changes and keep old screenshots; the App Store ranking factors guide covers how metadata alignment compounds across fields.

On every PPO test result: re-run steps 5 and 6. If a PPO test surfaces that a specific frame underperforms, the candidate keyword on that frame is a likely cause. Re-score and re-map only that frame. Keep the rest of the gallery stable so you isolate the variable.

A common indie mistake is treating screenshot keyword research as a one-time event. The competitive landscape shifts every quarter; your rank positions shift every month; Apple's algorithm tuning shifts roughly once per year (the June 2025 update [2] being the most recent material one). A quarterly cadence keeps you ahead of those shifts without over-rotating.

What are the most common mistakes in caption keyword research?

Four patterns show up repeatedly in indie research workflows. Each has a fix.

Mistake 1: Starting with discovery, not inventory. As covered in "Why do most workflows fail" above: discovery without inventory burns frames on keyword introduction bets that the ConsultMyApp data shows almost never pay off [1]. Fix: run step 1 before any keyword tool gets opened.

Mistake 2: Treating volume as the only ranking signal. Search volume is necessary but not sufficient. A high-volume keyword you can't rank for is worth less than a mid-volume keyword you can dominate. The three-dimension scoring (relevance, volume, difficulty) from step 4 [4] catches this. Fix: never sort candidates by volume alone.

Mistake 3: Skipping OCR confidence scoring. A candidate that passes ASO scoring but fails legibility scoring earns nothing. This shows up most often when designers fall back to decorative fonts, low-contrast color pairs, or middle-of-frame placement for aesthetic reasons. Fix: step 5 is a gate, not a guideline. Candidates that fail two rules don't ship.

Mistake 4: Keyword stuffing the caption. Apple's text detection works on natural-language phrases, not comma-separated lists [1]. "Run, Jog, Sprint, Marathon" reads as spam to the model and to users. "Track Your Marathon Training" reads as a phrase and ranks for both "marathon training" and adjacent keywords via the AI-based detection's semantic handling [2]. Fix: write captions as sentences a human would say, not as keyword lists.

For the design-side of these mistakes (font choice, contrast, placement) at depth, the Apple OCR mechanism guide covers the technical floor each rule sits on top of.

Takeaways

The 6-step workflow distilled:

• Screenshot caption keywords reinforce existing metadata. ConsultMyApp's 64-phrase test across 8 leading apps found only 1 unexplained ranking [1]. Caption-only keyword introduction is a 1.6 percent success bet.
• The workflow runs inventory first, discovery second. Step 1 pulls your title, subtitle, and keyword field. Step 2 pulls rank data on those existing phrases. Discovery (step 3 onward) operates inside that universe, not outside it.
• Prioritize candidates by current rank weakness. Caption reinforcement of a #1 ranking is wasted. Reinforcement of a #15 ranking with relevance fit has measurable upside.
• OCR confidence is the last filter, not a guideline. Candidates that can't be rendered at 16 points, high contrast, top or bottom placement, in standard sans-serif fonts earn no signal regardless of relevance or volume.
• Free tools cover 95% of indie research. Apple Search Popularity (via Search Ads), App Store search-bar autocomplete, App Store Connect's App Analytics, and the free ASO keyword researcher replace the $79/month AppTweak entry tier [3] for caption work.
• Re-run quarterly. Refresh steps 4 to 6 every 90 days. Re-run the full workflow whenever title or subtitle changes. Re-run steps 5 and 6 after any PPO test result.

For the strategy side of caption keywords (what to write, not how to research them), the screenshot SEO keyword strategy guide is the upstream pillar. For the technical mechanism that determines which caption text Apple's index actually sees, the Apple OCR deep-dive is the sibling reference. For the full ranking-factor hierarchy that screenshot signals fit into, the App Store ranking factors 2026 guide covers the bigger picture.

The whole point of running this workflow is so the captions on your screenshots earn their share of the conversion and ranking signal Apple gives screenshots that meet the bar. We built the free caption readability checker and free ASO keyword researcher so the research and validation steps don't require a $79/month tool. Once the keywords are picked, the screenshot builder generates the finished captions on every frame at every device size, so the keyword research is the only decision left to make.