Vejledning til sideindhold, meta-tags og strukturerede data optimeret til både søgemaskiner og AI-svarmaskiner.
@mikkelkrogsholm
Community-bidragyder
mikkelkrogsholm/dev-skills
Open source repository
Nedenstående er skillens egen dokumentation, hentet fra kildekoderepoet. Ophavsret tilhører forfatteren.
Two distinct optimization targets:
Key insight: AI search engines do not rank pages — they cite sources. Being cited is the new ranking #1.
The SEO and GEO principles in this skill are stable, but the llms.txt standard is actively evolving.
Before generating or advising on llms.txt files, use WebFetch to read the current spec:
WebFetch("https://llmstxt.org")
Other reference sources (use WebFetch when you need specifics):
references/schema-patterns.md for common templates)Apply these whenever writing or reviewing public-facing content or markup, regardless of framework.
From Princeton research on what makes content get cited by AI engines:
| Method | Visibility boost | How to apply |
|---|---|---|
| Cite authoritative sources | +40% | Link to studies, official docs, standards |
| Include specific statistics | +37% | Use real numbers, not vague claims |
| Add expert quotes | +30% | Attribute statements to named sources |
| Authoritative tone | +25% | Write with confidence and precision |
| Readable and clear | +20% | Short paragraphs, plain language |
| Use technical terminology | +18% | Domain-specific terms signal expertise |
| Vocabulary diversity | +15% | Avoid repetition of the same words |
| Fluency optimization | +15-30% | Natural flow, no awkward phrasing |
| Keyword stuffing | -10% | Never — hurts both SEO and GEO |
Best combination: fluency + statistics = maximum boost.
Each AI engine has different citation preferences. See references/geo-principles.md for full detail.
ChatGPT: favours branded domains; content updated within 30 days cited 3.2x more Perplexity: allow PerplexityBot in robots.txt; prioritises PDF documents and FAQ schema Google AI Overview: E-E-A-T signals, structured data, topical authority through content clusters Microsoft Copilot: requires Bing indexing; page speed under 2 seconds; LinkedIn/GitHub mentions help Anthropic AI: uses Brave Search indexing (not Google); prioritises factual density and structural clarity
llms.txt is to AI engines what robots.txt is to crawlers — a file at your site root that tells AI systems how to understand your content. Always add it to any public-facing site.
Place at https://yourdomain.com/llms.txt. Format is Markdown:
# Site or Product Name
> One-paragraph summary of what the site is and who it is for.
Optional additional context (key facts, caveats, important notes).
## Docs
- [Page title](https://yourdomain.com/page.md): What this page covers
## API
- [API reference](https://yourdomain.com/api.md): Endpoint overview
## Optional
- [Changelog](https://yourdomain.com/changelog.md): Recent changes
Key rules:
.md versions of pages where possible — AI engines prefer clean Markdown over HTML## Optional section signals lower-priority content that can be skipped in short-context situationsAlways add JSON-LD schema. See references/schema-patterns.md for templates.
Priority schema types:
Apply when reviewing any public-facing page:
llms.txt present at site root with H1, summary blockquote, and key page links