Last updated July 2026
When someone says “do the technical AEO stuff,” they usually mean all three of these tactics at once. That is a mistake. Each one targets a different bottleneck. Conflating them leads to wasted effort (and sometimes counterproductive moves, like removing schema because you think llms.txt covers it).
This piece untangles the three. For each one, you get a clear definition, the honest evidence on whether it works, and a note on where the research is still contradictory.
At a glance
| Tactic | What it fixes | What it does NOT do | Strongest proof signal | Effort |
|---|---|---|---|---|
| Schema markup (JSON-LD) | Parse ambiguity: crawlers misread your content type, entity type, or Q&A structure | Drive citations on its own; replace content quality | Contradictory (see evidence box below) | Medium: developer time or a schema tool |
| llms.txt | Crawler prioritization: AI bots waste crawl budget on low-value pages | Block crawlers; generate structured data; signal entity identity | Emerging; no controlled study yet | Low: one static file |
| Entity optimization | Knowledge Graph gaps: AI engines do not recognize your brand as a known entity | Fix on-page content quality; replace backlinks or citations | Strongest directional evidence of the three | High: ongoing earned media and consistency work |
Schema markup: the crawler parsing layer
Schema markup (specifically JSON-LD, the format Google recommends) is structured data you embed in a page’s <head> or <body> to tell crawlers how to interpret what they are reading. Without it, a crawler decides on its own whether your page is a product listing, a how-to guide, a Q&A exchange, or a generic article. Schema removes that guesswork.
For AEO, the most relevant schema types are FAQPage, HowTo, Article, and Product. They tell Google AI Overviews, Gemini, and other engines that a block of text is a question-and-answer pair, not a paragraph of prose. That distinction matters for answer box retrieval.
What to actually do with schema
Add FAQPage schema to pages with real Q&A blocks. Add HowTo schema to step-by-step content. Add Article schema with dateModified to keep your content freshness signal current. Validate with Google’s Rich Results Test before publishing.
Tools that support schema generation within an AEO workflow include Otto SEO and Surfer, which both flag missing structured data as part of broader content audits. Temso includes schema generation as one step in its full AEO fix workflow. For teams focused specifically on the writing-and-optimization loop, Writesonic generates FAQ schema alongside content drafts.
The honest benchmark: schema is table stakes for correct crawl interpretation. The citation uplift debate is real and unresolved. Do not skip it, but do not expect it to move citations on its own.
llms.txt: the crawler prioritization layer
llms.txt is a plain-text file you place at your domain root (for example, yourdomain.com/llms.txt). It lists the URLs and content descriptions you want AI crawlers to visit first. Think of it as a curated table of contents for your site, written specifically for large language model crawlers rather than for humans or traditional search bots.
It was proposed by Jeremy Howard in September 2024 and is not a standard backed by any major search engine. Google, OpenAI, and Anthropic have not publicly confirmed they read or weight it. What it does offer is a signal: if an AI crawler does consult it, it will find a clean, prioritized list of your best content rather than having to infer that from your sitemap or navigation.
What llms.txt does NOT do
Three common misconceptions:
- It does not block AI crawlers. That is
robots.txt. - It does not generate structured data or tell a crawler how to interpret content. That is schema.
- It does not establish your brand as a known entity. That is entity optimization (see below).
If you have heard that “llms.txt replaces schema,” that is wrong. They operate at entirely different layers.
What to actually do with llms.txt
Create the file. List your most authoritative pages first. Include a one-line description of each URL so a language model can understand what the page covers without fetching it. Update it when you publish major new content or deprecate old pages. It takes an hour to set up and minutes to maintain.
There is no dedicated tooling category for llms.txt yet. It is a file you create and maintain through your own workflow or developer setup.
Entity optimization: the knowledge layer
Entity optimization is the process of making your brand, products, and key claims recognizable as distinct, consistent entities in the knowledge systems that AI engines draw on. For Gemini, that means being present and consistent in the Google Knowledge Graph. For ChatGPT and Perplexity, it means being mentioned frequently and consistently across authoritative third-party sources.
This is the hardest tactic of the three and has the strongest directional evidence.
Entity optimization has three practical components:
-
Consistency across your own site. Your brand name, product names, and key claims should be identical across every page. Inconsistent naming (different capitalizations, shortened forms, outdated product names) makes it harder for any system to recognize you as a single entity.
-
Third-party mentions on authoritative sources. AI engines weight sources they trust. Wikipedia entries, industry directory listings, press coverage, review sites like G2, and Quora answers citing your brand all reinforce your entity identity. This is earned media work, not a technical fix.
-
Knowledge Graph presence. For brands where Gemini is a priority engine, having a verified Google Knowledge Panel is a direct signal. Consistent NAP (name, address, phone) data across directories, a Wikipedia entry, and correct schema markup on your own site all contribute.
What to actually do with entity optimization
Start with a brand audit: search your brand name in ChatGPT, Perplexity, Google AI Overviews, and Gemini. Note how each engine describes you. Note what it gets wrong. Note which sources it cites when it mentions you. Those sources are your distribution priority.
Knowatoa specializes in tracking how AI engines represent your brand across platforms and surfacing the gaps. For teams that want entity monitoring within a broader AEO platform, Temso tracks brand representation across eight AI engines and flags hallucinations alongside citation gaps. Profound provides citation maps that show which third-party domains are driving your entity mentions.
Entity optimization is the highest-effort tactic of the three, and it is the one with the least direct technical leverage. You cannot file a configuration change and be done. It requires ongoing earned media work and content distribution.
The multi-schema example
Here is a minimal example of layering FAQPage and Article schema on a single blog post. Both types can live on the same page.
[
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Schema Markup vs llms.txt vs Entity Optimization",
"datePublished": "2026-07-17",
"dateModified": "2026-07-17",
"publisher": {
"@type": "Organization",
"name": "AEO Rankings"
}
},
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "Does schema markup improve AI citations?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Controlled evidence found no meaningful citation lift after adding schema, while observational research found structured data correlated with higher citation likelihood. Treat schema as infrastructure and measure the result."
}
},
{
"@type": "Question",
"name": "Does llms.txt replace schema?",
"acceptedAnswer": {
"@type": "Answer",
"text": "No. Schema tells crawlers how to interpret content they have found. Llms.txt tells AI crawlers which URLs to prioritize. They operate at different layers and can coexist."
}
}
]
}
]Place this in a <script type="application/ld+json"> block. Validate with Google’s Rich Results Test before publishing. You can combine as many @type blocks as are genuinely applicable to your page, but do not add schema types your page does not actually represent.
How the three tactics stack
None of these is a replacement for content quality or third-party citations. They are infrastructure. Think of them as:
- Schema: labels on your shelves so the crawler can find what it is looking for.
- llms.txt: a note on your door saying which shelves are most important.
- Entity optimization: making sure the crawler already knows your name before it walks in.
All three contribute to the same outcome: an AI engine can find your content, understand what it says, and trust your brand as a source. Skip one layer and you create a bottleneck at that layer.
For the full picture on what tools cover each layer of the AEO stack, see the full AEO tool ranking or the AEO glossary for definitions of terms used here.