Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,85 @@
+---
+license: cc-by-4.0
+pretty_name: "Search vs. Store: AI App Search Demand vs App Store Rank vs Web Traffic (US, 2026)"
+tags:
+  - ai-apps
+  - google-trends
+  - app-store
+  - similarweb
+  - search-demand
+configs:
+  - config_name: default
+    data_files:
+      - split: snapshot_2026_07_26
+        path: data/search-vs-store-ai-apps-2026-07-26.csv
+      - split: snapshot_2026_07_01
+        path: data/search-vs-store-ai-apps-2026-07-01.csv
+---
+
+# Search vs. Store — AI app search demand vs App Store rank vs web traffic (US, 2026)
+
+One row per major AI app joining **three independent popularity signals**. Two
+cross-sectional US snapshots (**2026-07-01** and **2026-07-26**), 40 curated AI apps each.
+Aggregate/derived only — no raw records, no PII. License: **CC BY 4.0**.
+
+- Study + methodology: <https://crawlora.net/blog/search-vs-store-2026?utm_source=huggingface&utm_medium=referral&utm_campaign=search-vs-store>
+- Dataset page: <https://crawlora.net/datasets/search-vs-store?utm_source=huggingface&utm_medium=referral&utm_campaign=search-vs-store>
+- Living index (updated panel + lead/lag test): <https://crawlora.net/search-vs-store?utm_source=huggingface&utm_medium=referral&utm_campaign=search-vs-store>
+
+## Why it's interesting
+
+Rank the 40 apps by web visits and you get roughly the **search** order (ChatGPT, Gemini,
+Canva, Claude, DeepSeek…), **not** the App Store order. Search demand and web traffic agree;
+**the App Store rank is the outlier**, scrambled by distribution deals and platform push.
+The dataset makes that divergence explicit in the `divergence` column.
+
+## Columns
+
+- **app** — display name.
+- **trends_term** — the exact query measured on Google Trends (hand-picked to disambiguate
+  brand names, e.g. "Claude AI", "Google Gemini").
+- **search_tier** — coarse Google search-demand tier: Dominant > High > Moderate > Low >
+  Minimal. Coarse on purpose: ChatGPT dwarfs the field ~30–1000x, so the sub-Minimal tail is
+  below Google Trends' resolution and is never individually ranked.
+- **trends_interest_chatgpt100** — trailing-12-month mean Google Trends interest,
+  anchor-normalized so ChatGPT = 100. Directional, **not** precise for the small tail.
+- **best_app_store_rank** — best current US chart position across iOS/Play, Top Free +
+  Grossing (null = off the top-100). **ios_rank** / **android_rank** — per-store best.
+- **chart_score** — 101 - best_app_store_rank (higher = better standing; null if off-chart).
+- **ratings_count** — combined iOS+Play ratings count (install proxy), where matched.
+- **sw_monthly_visits** — SimilarWeb estimated monthly web visits (absolute; blank = below
+  SimilarWeb's floor). **sw_global_rank** — SimilarWeb global site rank.
+- **divergence** — `downloads_not_search` | `search_not_downloads` | `aligned` (tier gap
+  between store standing and search demand).
+
+## Files
+
+| File | What |
+|---|---|
+| `data/search-vs-store-ai-apps-2026-07-26.csv` | Snapshot 2026-07-26, one row per app |
+| `data/search-vs-store-ai-apps-2026-07-26.json` | Same rows + metadata + the search x store tier matrix |
+| `data/search-vs-store-ai-apps-2026-07-01.csv` | Snapshot 2026-07-01, one row per app |
+| `data/search-vs-store-ai-apps-2026-07-01.json` | Same rows + metadata + tier matrix |
+
+The `.json` files are a metadata wrapper (`method`, `columns`, `tier_matrix`, `data[]`), so
+the CSVs are the flat, directly-loadable tables.
+
+## Caveats worth reading before you cite this
+
+- **Do not compute a cross-app search correlation from `trends_interest_chatgpt100`.** Google
+  Trends rescales 0-100 against the largest term in each request, and ChatGPT's volume
+  collapses the tail into noise. The tiers and the `divergence` label are robust; a
+  correlation coefficient over these values is an artifact. See the study for the full
+  writeup.
+- **`sw_monthly_visits` / `sw_global_rank` are joined from the nearest available panel date**
+  when the daily SimilarWeb panel had no row for the snapshot date (see `similarweb_as_of` in
+  the JSON metadata) — a few days off from the rest of the row.
+- App Store chart history behind these snapshots is only ~2-3 weeks deep upstream; these are
+  point-in-time snapshots, not a long rank history.
+
+## Citation
+
+> Crawlora (2026). *Search vs. Store: AI App Search Demand vs App Store Rank vs Web Traffic
+> (US, 2026)*. CC BY 4.0. https://crawlora.net/datasets/search-vs-store
+
+Collected with [Crawlora](https://crawlora.net?utm_source=huggingface&utm_medium=referral&utm_campaign=search-vs-store).