GEO analysis skill

GEO Analysis

Run a full GEO/SEO analysis for a domain and get a client-ready PDF report: AI visibility (is the brand recommended by ChatGPT / Google AI), deep market research, and competitor landscape.

Thin wrapper over the Enception client-analysis API. Three operations: start, status, report (plus an optional brand-name correction).

Disclosure: contributed by Enception AI and calls Enception's paid external API. Docs: https://www.enception.ai/doc/api-client-analysis

Requirements

• ENCEPTION_ANALYSIS_API_KEY — sent as the x-api-key header. Self-serve: sign up at https://www.enception.ai, then generate a free-tier key under Account → API Key (2 analyses/month free; contact the team for higher limits). Rate limit: 10 requests/minute.

1. Start

POST https://www.enception.ai/api/external/client-analysis/start

Required: website_url, primary_region, core_goals. Sensible defaults: primary_region=us, core_goals=traffic, output_language=en, competitors=[]. The API normalizes URLs to a bare domain and rejects IPs. Returns { id, message } — save the id.

Pick output_language from an explicit request first, then the user's conversation language, then the domain TLD (.cn → zh; otherwise en).

curl -s -X POST "https://www.enception.ai/api/external/client-analysis/start" \
  -H "x-api-key: $ENCEPTION_ANALYSIS_API_KEY" -H "Content-Type: application/json" \
  -d '{"website_url":"https://example.com/","primary_region":"us","competitors":[],"core_goals":"traffic","output_language":"en"}'

Only override primary_region / core_goals / competitors if the user specifies. core_goals is free text (e.g. traffic, ranking). Competitors: [{"name":"X","domain":"x.com"}, ...] (max 5). For multiple domains, fire one POST each and collect the ids.

2. Status

POST https://www.enception.ai/api/external/client-analysis/status with body { "id": "<id>" } and the same x-api-key header. Returns { status: 'pending' | 'failed' | 'succeeded', progress_percent?, pdf_url? }.

A full run takes ~20–45 minutes — the bottleneck is the deep market-research step. Poll on a long interval (minutes, not seconds). succeeded ⇒ ready to fetch the report. If a run fails or stalls indefinitely, report that to the user rather than inventing results.

3. Report

GET https://www.enception.ai/api/external/client-analysis/report/<id> — public, no auth; shareable by id. Works once status is succeeded (also serves partial runs).

curl -s -o geo-analysis-example.pdf \
  "https://www.enception.ai/api/external/client-analysis/report/<id>"

Confirm it's a real PDF (file geo-analysis-example.pdf), then give the user the file and/or the public URL.

4. Fix "zero visibility" (brand-name correction)

POST https://www.enception.ai/api/external/client-analysis/update-brand with body { "id", "brand_name"?, "alternative_names": [...] } and the x-api-key header.

When a report shows zero/low AI visibility but the brand IS recommended by ChatGPT/Google, it's almost always because the brand is only known by its fused domain label (e.g. pudurobotics) while AI engines say "Pudu Robotics". Supply the real name + aliases and the route recomputes visibility from the already-stored answers (no re-query — fast), then re-fetch the report.

curl -s -X POST "https://www.enception.ai/api/external/client-analysis/update-brand" \
  -H "x-api-key: $ENCEPTION_ANALYSIS_API_KEY" -H "Content-Type: application/json" \
  -d '{"id":"<id>","brand_name":"Pudu Robotics","alternative_names":["Pudu","PuduBot"]}'

Returns { success, mention_rate, overall_score, average_position, analysis_regenerated }. New analyses resolve the brand name automatically; this is the manual correction path for existing reports.