Ask what your users ask
People ask an assistant for an app the way they would ask a friend. The questions you care about live in the spec as intents, each with the prompts people type, in every country you target.
The questions are files
visibility init writes the intents from your listing, or the ones Appeeky already tracks. Your agent adds a prompt in a pull request; visibility sync sends it to Appeeky.
aliases lists the other names the assistants use for the app, so an answer that calls it something else still counts.
web:url: https://sleepsounds.app/keywords:us: [rain sounds, white noise app, sleep sounds for baby]aiVisibility:countries: [us]aliases: [Sleep Sounds]intents:us:- label: fall asleep to rain soundsprompts:- what app plays rain sounds to help me sleep- best free rain sounds app for iPhone
Asked on your keys, or by Appeeky
With an APPEEKY_API_KEY the questions are asked for you on a schedule. Without one, visibility scan --local asks them on your own model keys, shows what it will cost first, and writes the same files.
$ npx @appeeky/spec visibility scan --localUS 6 intents · 24 prompts × chatgpt, claude, gemini, perplexity · about $2.45Asks every assistant with web search on, the way the apps answer people.✓ appeeky/visibility/us.yaml winning 3 · weak 2 · missing 1✓ appeeky/visibility/history.csv one row per intent and assistant✓ appeeky/visibility/answers/us.jsonl every answer, word for word
What the assistants say,
and how Google sees the site
One run of each, played out: the assistants answering and the pages they cite, then the website checked, measured and ranked.
- 01
Ask
Your prompts, from the spec, asked of every assistant in each country you target.
- 02
Answer
Who each one names: you, at which place, or someone else instead.
- 03
Cite
The pages they point to: Reddit threads, review sites, your own website.
- 04
Fix
A score per question and the change most likely to get you named, written to the repo.
Your keys, your bill
Every source runs on an account you own and pays for directly. Nothing here needs an Appeeky account.
ChatGPT
OPENAI_API_KEY
gpt-4.1 with web search
Claude
ANTHROPIC_API_KEY
Claude Haiku with web search
Gemini
GEMINI_API_KEY
Gemini 2.5 Flash with Google Search
Perplexity
PERPLEXITY_API_KEY
Sonar, which always searches

Search Console
GSC_CREDENTIALS
a service account key, or your gcloud login

Google ranks
DATAFORSEO_LOGIN
with DATAFORSEO_PASSWORD, your DataForSEO account
Your website, the way a crawler reads it
People search Google before the store, and the assistants read web pages when they pick an app. Add web.url to the spec and the site is kept next to the listing.
Checked with no key
seo check reads the site like a crawler, as above. Each finding comes with the fix, and it exits 1 when something fails.
seo llms writes llms.txt from your listing: what the app is, where to get it and what it is good for.
# Sleep Sounds> Rain, fan and white noise sounds to fall asleep, for iPhone and Android.## Get it- App Store: https://apps.apple.com/app/id1488031122- Google Play: https://play.google.com/store/apps/details?id=com.sleepsounds.app## Good for- Falling asleep to rain sounds- White noise for a crying baby
Search Console in the repo
seo pull keeps clicks and impressions per day, the top queries and pages, and the queries close to page one: positions 4 to 15, where one better title moves you up.
$ npx @appeeky/spec seo pullSearch Console · sc-domain:sleepsounds.app · last 28 daysclicks 1,912 ↑18% impressions 84.3k ↑9%Close to the top (position 4–15):rain sounds for sleeping pos 6.2 3.1k impressionswhite noise for baby pos 8.9 1.8k impressions store targetWrote appeeky/seo/daily.csv, queries.csv, pages.csv, summary.md
Google ranks and the AI Overview
On your DataForSEO account, seo ranks reads Google for each keyword: the site, the store page, and whether the AI Overview cites you. seo keywords adds volume and difficulty, seo domain what the site ranks for and who ranks beside it.
$ npx @appeeky/spec seo ranks3 keywords in US · about $0.06 on DataForSEOrain sounds site 6 store page 2 AI Overview cites youwhite noise app site 14 store page —sleep sounds for baby site 9 store page 4Recorded in appeeky/seo/ranks.csv · changes in changes.md
Measured with every release
A description that says what the app is for helps store search and the assistants alike. So every push is judged on all of them.
Before and after, on the same days
verify reads ai_visibility (the share of answers that name the app), search_clicks, web_rank and ai_overview_citations around each push, next to store ranks and conversion.
What won goes into the playbook, and your agent reads it before the next change.
$ npx @appeeky/spec verify2026-11-03 · day 14 · WONai_visibility 42% → 58% named by 2 more assistantssearch_clicks/day 61 → 74web_rank 11.4 → 8.2ai_overview_citations 1 → 2playbook: + say what the app is for in the first line, it helps the assistants too
The commands
- visibility init
- Write the intents and prompts into the spec, from your listing or from Appeeky.
- visibility sync
- Send the spec's intents and prompts to Appeeky.
- visibility pull
- Write what each assistant answered, who it named and the pages it cited.
- visibility scan --local
- Ask the assistants now on your own keys, with the cost shown first.
- seo check
- Read the site like a crawler: search, the app, AI crawlers. No key.
- seo llms
- Write llms.txt from the listing.
- seo pull
- Search Console clicks, queries and pages into appeeky/seo/.
- seo ranks · keywords · domain
- Google itself, on your DataForSEO account.