---
name: optimize-for-ai
description: Optimize content to get cited by AI search and answer engines (ChatGPT, Perplexity, Google AI Overviews, Gemini) instead of ranked in a list. Use when asked about AI SEO, GEO, AEO, answer engine optimization, appearing in AI answers or AI Overviews, zero-click search, or getting cited by ChatGPT or Perplexity. Checks crawler access first, scores each page on a five-dimension citation model, and produces a prioritised plan to earn citations. Ships a scanner that measures the mechanical signals and the crawler veto.
license: See LICENSE.md
---

# Optimize for AI

Gets content cited as a source inside AI-generated answers, which is a different game
from ranking in a list of links. Answer engines lift a sentence, a stat, or a table
straight into the answer and credit the page it came from, so the work is making the
page easy to extract from, easy to quote, and easy to trust, from a brand the engine
can identify.

One rule sits above the rest: if the AI crawlers cannot read the page, none of it
matters. So crawler access is checked first, and a block vetoes the whole score.

## When to use

Trigger for "AI SEO / GEO / AEO / answer engine optimization", "how do I show up in
AI answers / AI Overviews", "get cited by ChatGPT or Perplexity", or "zero-click
search". This is about being cited by answer engines. For traditional ranked search,
hand off to the **seo-audit** skill; the two work together, they do not replace each
other.

## Inputs

- **the site or page(s)** to optimise, and the topics they should be cited for
  (required).
- **competitors** currently cited for those topics, if known (for the gap analysis).
- **brand** name and the profiles it should be linked to (for entity work).

## Process

**1. Diagnose (`references/diagnosis.md`).** Check robots.txt for AI-crawler blocks
first (the veto). Then measure the current state: search your topics on the main
engines, record who is cited instead of you, and ask the engines what your brand is
to see if they get it right. The gaps become the work list.

**2. Score (`references/citation-scoring.md`).** Rate each target page on the five
dimensions, worst first: extractability, quotability, authority, freshness, entity
clarity. A crawler block sets the visibility score to zero regardless.

**3. Fix the content (`references/content-structures.md`).** Restructure for the
engine: the answer in the first 150 words, comparison tables and key-findings
callouts up top, "X is Y" definitions, and a named source and date on every statistic.
Use the format that gets cited for the topic.

**4. Build the entity (`references/entity-and-authority.md`).** Ship Organization,
Article, FAQ, HowTo, and Person schema; add sameAs to authoritative profiles; keep
the brand name identical everywhere; and work the three-tier entity plan so the
engines recognise and describe the brand correctly.

**5. Clear the technical bar (`references/engines-and-technical.md`).** Serve the
answer in the initial HTML (server-render or static, no login wall, no
JavaScript-only body), use semantic HTML, and consider a concise llms.txt. Note how
each engine weighs freshness, authority, and structure.

**6. Run the scanner and re-measure.** On each target page:
```
node scripts/ai-scan.mjs https://yoursite.com/page
```
Fix every VETO and the weakest dimensions it surfaces, then re-run the engine queries
from step 1 over time to see the citations move.

## Principles

- **Cited, not ranked.** Optimise to be lifted into an answer, not to sit at position
  three in a list.
- **The veto first.** A blocked crawler makes every other effort worthless; it is the
  first thing checked and the first thing fixed.
- **Extractable, quotable, trusted, from a known entity.** Those four are the whole
  game; a page that fails one of them does not get cited.
- **Raise the weakest dimension.** One failing dimension caps the page. Fix that
  before polishing a strong one.
- **Honest baseline.** The engines are non-deterministic; measure over a few clean
  queries and state what is measured versus assumed.

## Output

An AI-visibility strategy: the current presence (crawler access, who is cited, entity
accuracy), a prioritised action table, a per-page content plan, and the entity
checklist, with the scanner output folded in. See `references/report-format.md`.

## What this skill will not do

- It will not optimise for traditional ranked search positions; that is the
  **seo-audit** skill.
- It will not fabricate statistics, sources, credentials, or entity facts to look more
  citable; false claims lower citation odds and break trust.
- It will not promise a citation; the engines are non-deterministic. It raises
  citation-readiness and measures the mechanical signals honestly.

## Reference map

- `references/diagnosis.md` — crawler veto, current AI presence, citation gap analysis.
- `references/citation-scoring.md` — the five-dimension AI Citation Score.
- `references/content-structures.md` — formats and structures that get cited.
- `references/entity-and-authority.md` — entity building, authority signals, schema.
- `references/engines-and-technical.md` — engine behaviours, technical bar, llms.txt.
- `references/report-format.md` — how to present the strategy.
- `scripts/ai-scan.mjs` — the mechanical signal + crawler-veto scanner.