Imagine a SaaS site: it has 18 pages, content changes rarely, and all support documentation lives in one place. Now imagine an e-commerce site on the same day: 8,000 products, constant price changes, and hundreds of category pages. Choosing the same AI documentation file for both structures often creates unnecessary maintenance work or incomplete guidance.
Short answer: If your site is small and stable, llms-full.txt may make more sense; for medium and large sites, llms.txt is often a healthier starting point. Because the goal is not to dump as much as possible into an AI system, but to present the right pages, with the right priority, in an up-to-date format.
The key difference between llms.txt and llms-full.txt
There is no single official standard in this space that works the same in every tool, so you should think about the file’s function rather than its name. In practice, llms.txt is used more like an AI-friendly site map that includes selected pages and short guidance. llms-full.txt, on the other hand, is preferred by teams that want to provide more context, a broader page list, or more extensive documentation.
The critical point here is this: AI agents and AI search systems often work with limited context. Giving them 200 URLs may not produce better results than giving them 12 well-chosen URLs. Especially if content changes quickly, a broader file goes stale more often. That is why the right choice depends on content volume, update frequency, and maintenance capacity.
How to choose correctly for 3 site types
| Site type | More sensible file | Why? |
|---|---|---|
| Small corporate site, portfolio, single-product SaaS | llms-full.txt | The number of URLs is low; pages are stable; providing broad context in a single file is easier to maintain. |
| Mid-sized blog, content site, agency site | llms.txt | You need to choose only the most valuable pages; otherwise the file grows quickly and may go out of date. |
| Large e-commerce, marketplace, multilingual setup | llms.txt + section-based guidance | A complete list for thousands of URLs is often inefficient; category, guide, and policy pages should be prioritized. |
1) Small and stable sites: llms-full.txt
If you have between 5 and 30 critical pages and they change less than once a month, llms-full.txt can be a sensible choice. Here, the goal is to explain nearly the entire site to the AI system in a clean way. For example, if you have service pages, about, contact, pricing, and a few documentation pages, presenting the full picture in one file remains straightforward.
2) Mid-sized content sites: llms.txt
In a blog with 50, 100, or 300 pages, not every URL has the same value. On sites like this, llms.txt acts like a site map for AI: main category pages, high-traffic articles, product comparisons, and conversion pages are brought to the front. This approach both reduces update costs and gives the AI system a clearer roadmap.
3) Large and complex sites: a layered structure with llms.txt
In e-commerce, marketplaces, and multilingual setups, llms-full.txt is usually not the first choice. Product counts, variants, stock changes, and filtered URLs can quickly bloat the file. A healthier approach is to keep only strategic pages in llms.txt and, if needed, produce additional documentation by category. That way, readability is preserved and maintenance burden stays under control.
Technical criteria to review before deciding
Make the file choice based on measurable criteria, not intuition. The questions below quickly show which file is more sensible:
- How many critical URLs are there? On sites under 30 pages, llms-full.txt may be more manageable.
- How often does the content change? If prices and campaigns change weekly, a broad file goes stale quickly.
- Who will maintain it? If one person manages it, a short and selected file is safer.
- What do you want to show the AI system? Not the entire archive, but the most valuable source pages for answers.
- Is your URL structure clean? If you have parameterized, filtered, and duplicate pages, fix the canonical structure first.
- Is the site multilingual? If content differs across languages, structure the file by language.
Technical checklist
- First, extract all important URLs from your site: service, category, guide, policy, and contact pages.
- Define one primary purpose for each URL: information, conversion, support, or reference.
- If there are fewer than 30 pages and the site is stable, consider llms-full.txt; if there are more, start selectively with llms.txt.
- Include only canonical, indexable, and up-to-date pages in the file.
- Leave out old campaign pages, test URLs, and duplicate content.
- Keep the file in the root directory and align updates with your content refresh process.
- Build a structure that does not conflict with robots.txt, sitemap.xml, and canonical tags.
- Review the file at regular intervals against logs and your content inventory.
Example scenario: how to decide for 3 site types
Scenario A: A law firm site with 12 pages. The content is stable, the pages are clear, and the site is conversion-focused. In this case, llms-full.txt may make sense, because gathering all critical pages in one file is easy.
# llms-full.txt
Site: Ornek Hukuk Bürosu
Amaç: Kurumsal bilgi ve hizmet sayfaları
- /hakkimizda
- /avukatlar
- /hizmetler
- /ticaret-hukuku
- /is-hukuku
- /iletisim
Scenario B: A blog with 180 articles. Not every article matters equally. Here, llms.txt is the better fit, because you can select your best-performing guides, pillar content, and conversion pages.
# llms.txt
> This site covers SEO, technical setup, and AI search readiness for SMBs.
## Priority pages
- /ai-seo-rehberi
- /robots-txt-rehberi
- /sitemap-xml-nasil-kurulur
- /iletisim
Scenario C: An e-commerce site with 6,000 products. The product list already changes frequently. In this case, instead of writing a broad file, category pages, return policy, shipping information, and buying guides should be prioritized. In other words, the practical solution is often llms.txt and layered content prioritization.
Common mistakes
- Adding every page to the file and turning it into a dump area.
- Including old campaigns and noindex pages.
- Writing llms-full.txt once and not updating it for months.
- Putting the AI file in place of robots.txt; they do not serve the same function.
- Writing URLs in the file differently from the canonical URL.
In short: llms-full.txt can be a strong choice for small and stable sites; llms.txt is often a more controlled, more sustainable option for most medium and large sites. When deciding, focus less on the file name and more on content volume, maintenance capacity, and the quality of the context you want to give the AI system.