Keyword Difficulty Analyzer

GEO difficulty

by OpenClaudia · MIT license · ★ 705 on GitHub

Score how hard a keyword is to rank for in the AI-search era — page-level URL Rating of real competitors (not just domain DR), Ahrefs keyword difficulty, and whether a given site already ranks or is cited in Google's AI Overview. Use…

---
name: geo-difficulty
description: "Score how hard a keyword is to rank for in the AI-search era — page-level URL Rating of real competitors (not just domain DR), Ahrefs keyword difficulty, and whether a given site already ranks or is cited in Google's AI Overview. Use when the user asks 'how hard is <keyword> to rank', 'geo difficulty for <keyword>', or 'keyword difficulty for <domain>'."
---

# GEO Difficulty

Score keyword ranking difficulty for the AI-search era. Unlike plain Ahrefs KD or domain DR, this:

- uses **page-level URL Rating (UR)** of the results that actually target the query — so a generic article on a huge domain (e.g. mckinsey.com) stops inflating the score;
- is **client-relative** — the same SERP is a wall for a no-name site but ~won for a site already ranking or cited in Google's AI Overview;
- folds in the **AI Overview** citation as a first-class signal.

**Disclosure:** contributed by Enception AI. Backed by the free public API behind https://www.enception.ai/tools/geo-difficulty (`POST https://www.enception.ai/api/tools/geo-difficulty`, no API key). Server-side data comes from Ahrefs and SerpAPI.

## Usage

```bash

Overview

Score keyword ranking difficulty for the AI-search era. Unlike plain Ahrefs KD or domain DR, this:

What you get

absolutedifficulty 0–100 for a brand-new site (0.35·KD + 0.40·median competitor page-UR + 0.25·targeting%).
client_verdict / client_diffonly with --client. WON (top-3 or AI-Overview cited), FOOTHOLD (ranks 4–10), NOT RANKING (with the gap to close, credited for the client's DR).
KDAhrefs keyword difficulty; - for zero-volume long-tail.
URmedian page-level URL Rating of the real competitors.
tgt%share of the top-10 that actually target the query.
AIOcited (client in the AI Overview), Nsrc (AIO shown, N sources), none, or n/a (AIO not available for this query).

Manual install (npx, ZIP)

Claude Codenpx skills add OpenClaudia/openclaudia-skills --skill geo-difficulty -g -a claude-code

Claude Code (degit)npx degit OpenClaudia/openclaudia-skills/skills/geo-difficulty#main ~/.claude/skills/geo-difficulty

Claude.ai / desktopZip the folder and upload it under Settings → Capabilities → Skills.

Codex, Cursor, other agentsPoint them at SKILL.md. The other files are plain text.

Connects with Agent Alley MCP · agentalley.io/mcp

Files included

SKILL.md2.7 KB

Read SKILL.md here

OpenClaudia/main1 file shown

SKILL.md2.7 KB

Browse every file on GitHub↗

SKILL.md

Show the full text49 lines

geo-difficulty/SKILL.md49 lines · 2.7 KB

PreviewCode

Outline

RawView on GitHub

GEO Difficulty

Score keyword ranking difficulty for the AI-search era. Unlike plain Ahrefs KD or domain DR, this:

uses page-level URL Rating (UR) of the results that actually target the query — so a generic article on a huge domain (e.g. mckinsey.com) stops inflating the score;
is client-relative — the same SERP is a wall for a no-name site but ~won for a site already ranking or cited in Google's AI Overview;
folds in the AI Overview citation as a first-class signal.

Disclosure: contributed by Enception AI. Backed by the free public API behind https://www.enception.ai/tools/geo-difficulty (POST https://www.enception.ai/api/tools/geo-difficulty, no API key). Server-side data comes from Ahrefs and SerpAPI.

Usage

bash scripts/score.sh "ai form real estate" "ai slide legal" --client example.com

Pass one or more keywords (quote each). Up to 10 per call.
--client <domain> is optional; add it to get the client-relative verdict (WON / FOOTHOLD / NOT RANKING).
GEO_DIFFICULTY_URL env var overrides the endpoint (rarely needed).

Or call the API directly:

curl -s -X POST https://www.enception.ai/api/tools/geo-difficulty \
  -H 'content-type: application/json' \
  -d '{"keywords":["ai presentation maker"],"client":"example.com"}' --max-time 150

Output columns

absolute — difficulty 0–100 for a brand-new site (0.35·KD + 0.40·median competitor page-UR + 0.25·targeting%).
client_verdict / client_diff — only with --client. WON (top-3 or AI-Overview cited), FOOTHOLD (ranks 4–10), NOT RANKING (with the gap to close, credited for the client's DR).
KD — Ahrefs keyword difficulty; - for zero-volume long-tail.
UR — median page-level URL Rating of the real competitors.
tgt% — share of the top-10 that actually target the query.
AIO — cited (client in the AI Overview), Nsrc (AIO shown, N sources), none, or n/a (AIO not available for this query).
competitors — the targeting results with their UR.

Notes

The AI Overview leg depends on Google actually showing an AIO for the query (intermittent); it degrades gracefully to none/n/a.
Numbers shift day to day as SERPs and UR change — treat as a current snapshot, not a fixed grade.
Each keyword takes a few seconds server-side; keep --max-time generous for batches.

1	---
2	name: geo-difficulty
3	description: "Score how hard a keyword is to rank for in the AI-search era — page-level URL Rating of real competitors (not just domain DR), Ahrefs keyword difficulty, and whether a given site already ranks or is cited in Google's AI Overview. Use when the user asks 'how hard is <keyword> to rank', 'geo difficulty for <keyword>', or 'keyword difficulty for <domain>'."
4	---
5
6	# GEO Difficulty
7
8	Score keyword ranking difficulty for the AI-search era. Unlike plain Ahrefs KD or domain DR, this:
9
10	- uses page-level URL Rating (UR) of the results that actually target the query — so a generic article on a huge domain (e.g. mckinsey.com) stops inflating the score;
11	- is client-relative — the same SERP is a wall for a no-name site but ~won for a site already ranking or cited in Google's AI Overview;
12	- folds in the AI Overview citation as a first-class signal.
13
14	Disclosure: contributed by Enception AI. Backed by the free public API behind https://www.enception.ai/tools/geo-difficulty (`POST https://www.enception.ai/api/tools/geo-difficulty`, no API key). Server-side data comes from Ahrefs and SerpAPI.
15
16	## Usage
17
18	```bash
19	bash scripts/score.sh "ai form real estate" "ai slide legal" --client example.com
20	```
21
22	- Pass one or more keywords (quote each). Up to 10 per call.
23	- `--client <domain>` is optional; add it to get the client-relative verdict (`WON` / `FOOTHOLD` / `NOT RANKING`).
24	- `GEO_DIFFICULTY_URL` env var overrides the endpoint (rarely needed).
25
26	Or call the API directly:
27
28	```bash
29	curl -s -X POST https://www.enception.ai/api/tools/geo-difficulty \
30	-H 'content-type: application/json' \
31	-d '{"keywords":["ai presentation maker"],"client":"example.com"}' --max-time 150
32	```
33
34	## Output columns
35
36	- absolute — difficulty 0–100 for a brand-new site (`0.35·KD + 0.40·median competitor page-UR + 0.25·targeting%`).
37	- client_verdict / client_diff — only with `--client`. WON (top-3 or AI-Overview cited), FOOTHOLD (ranks 4–10), NOT RANKING (with the gap to close, credited for the client's DR).
38	- KD — Ahrefs keyword difficulty; `-` for zero-volume long-tail.
39	- UR — median page-level URL Rating of the real competitors.
40	- tgt% — share of the top-10 that actually target the query.
41	- AIO — `cited` (client in the AI Overview), `Nsrc` (AIO shown, N sources), `none`, or `n/a` (AIO not available for this query).
42	- competitors — the targeting results with their UR.
43
44	## Notes
45
46	- The AI Overview leg depends on Google actually showing an AIO for the query (intermittent); it degrades gracefully to `none`/`n/a`.
47	- Numbers shift day to day as SERPs and UR change — treat as a current snapshot, not a fixed grade.
48	- Each keyword takes a few seconds server-side; keep `--max-time` generous for batches.
49