# Aeo Geo Improvement Bot

> AEO / GEO Improvement Bot

- Skill: `saurabhshuklagrowisto/aeo-geo-improvement-bot` (Agent Skill, multi-file: 5 files)
- Install (CLI): `npx skillmds@latest add saurabhshuklagrowisto/aeo-geo-improvement-bot`
- Raw SKILL.md: https://api.skillmd.com/api/skills/saurabhshuklagrowisto/aeo-geo-improvement-bot/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: saurabhshuklagrowisto (https://skillmd.com/u/saurabhshuklagrowisto)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/saurabhshuklagrowisto/aeo-geo-improvement-bot

---


# AEO / GEO Improvement Bot

The visibility audit tells you *where* a brand is invisible in AI answers. This skill tells you *what to do about it* -- it is the second half of the loop: measure, diagnose, fix, re-measure.

## When to use

- Right after running [AEO / LLM Visibility Audit](../aeo-llm-visibility-audit) and getting back `gap_questions`.
- When the ask is to build an actual improvement plan, not another audit -- content briefs, schema fixes, citation targets, in priority order.
- To decide what a content/SEO team should work on first when there are more gaps than there is time.

## When NOT to use

- To measure share of voice in the first place -- that's the visibility audit skill.
- As a substitute for writing the content or shipping the schema -- this produces the plan and the specific instruction, not the final asset. Feed the `content` fixes to a content-writing skill and the `schema` fixes to a schema-audit skill.

## Method

1. For every gap question, read four signals: does an owned page answer it, does that page carry schema markup, how many independent (third-party) sources corroborate it, and how much demand it represents.
2. Classify the gap into exactly one root cause, cheapest-explanation-first: **no content** > **no schema** > **no citations** > **authority deficit**. Fixing the wrong layer wastes effort -- adding schema to a page that doesn't exist does nothing.
3. Prioritise: higher demand first; at equal demand, cheaper fixes (schema) surface before expensive ones (earning citations or authority) so the plan front-loads fast wins.
4. Attach a re-check window per fix type -- schema changes show up in re-crawls faster than earned citations do.

## Inputs

- `brand` -- the brand name
- `gap_questions` -- from the visibility audit's `gap_questions[]`
- `signals` -- per question: `has_owned_content`, `has_schema_markup`, `third_party_mentions`, `search_volume_proxy` (1-10). In production these come from a site crawl + a third-party source search; the demo ships a deterministic sample.

## Output (JSON)

`fixes[]` (each with `question`, `fix_type`, `action`, `recheck_after_days`, `priority`), `fix_type_mix`, and a one-line `summary` naming the top fix to start on.

## Run it

```bash
python scripts/geo_improve.py         # built-in sample
python scripts/geo_improve.py in.json # your own brand + gap_questions + signals
```

Zero dependencies, no API keys. To go live: swap the sampled `signals` for a real site crawl (does the page exist, does it carry JSON-LD) and a real third-party mention count; the classification and prioritisation logic stays identical. Chain it after `aeo_score.py`'s output and, on a schedule, after re-running the audit to confirm each fix moved the needle -- a scheduled n8n version of the full measure-diagnose-fix-recheck loop ships in `pipeline-automation/aeo-geo-improvement-loop.json`.