Not that long ago, organic visibility was all about improving rankings to generate traffic and conversion. The standard SEO tactics of keyword research, content creation & optimization, and iterating on initial results were a tried and true recipe for success over time. That world is changing fast.
Today, people are increasingly discovering content through LLMS like ChatGPT, Perplexity, and Google’s AIOs, which read, synthesize, and cite content directly in their generated answers. As of this post, research indicates that around 5% of searches are done on these LLMs vs traditional search engines.
So how do you write content that shows up in LLMs? Let’s break it down.
How LLMs Actually Read Content
Unlike classic crawlers that focused heavily on HTML structure, LLMs ingest raw text, break it into tokens, and map relationships between ideas. They reward content that is clear, structured, contextually rich and EASY to extract. If a reader can find the info easily on your site, an LLM is likely to, too!
Start With a Content Reality Check
As with traditional SEO, before writing anything new, audit existing content. With any audit, we look for “low hanging fruit” to optimize first. Rankings that hover on the edge of page 1 and are on page 2 (with volume) are easy targets for optimizations. The same holds true for LLMs.
Similar to an SEO audit, figure out:
- Which pages are already showing up in AI Overviews, ChatGPT, or Perplexity
- Where can content be better structured, optimized, more informative
- Are there grammar issues, duplicated content, or unanswered questions?
One Page, One Topic
Again, as with SEO, every page should have a clear purpose. Google ranks content on a url level – looking to see if a page is thematically relevant to a query. Same with LLMs. Using a descriptive title formatted as an H1 and keeping the page focused on a topic via an overview, subheads (formatted with h tags) and a call to action remains a best practice!
Write in Chunks
AI systems don’t usually pull entire pages. They retrieve passages called “chunks” and stitch them together into answers. In theory, each section on each page of your site could be captured as an answer or a mini article. Aim for roughly 100-300 words per section. Long enough to explain something properly, short enough to stay focused.
Lead With the Answer
Answer first, explain second. SEOs have always LOVED a good FAQ as a blog post topic. Definition/glossary pages, comparison graphs, FAQ are all great (and out-of-the-box) structured content when done correctly.
Anticipate the Follow-Up Questions
Modern AI search doesn’t stop at one query. Thematically, a page should focus on a topic and be answered more in depth with related queries (pro tip: use keyword research tools) as subheads. Your content should naturally address common :
- What is it?
- How does it work?
- Pros and cons
- How it compares to alternatives
- Real-world examples & case studies
Expertise Still Matters (A Lot)
AI can summarize what already exists, but it can’t replace us. Think of what value and insights you and your team offer (that are okay to share publicly). Expert commentary, opinions, actual stats, original surveys, dramatically increase the likelihood that your content gets cited instead of paraphrased away.
Make Your Content Easy to Cite
LLMs don’t just retrieve content, they cite it, as well as other websites you don’t even own (Indeed, Yelp, review sites, aggregators, etc.). Make sure that a. you are checking 3rd party sites that reference you and b. that your claims are validated, verifiable and from a place/person of authority very close to the citable content itself, Pro tip: Show author credentials and publish dates. Sound familiar? You must have been doing SEO all along 🙂
Build the Ecosystem, Hub & Spoke or Pillar & Cluster Style
As a refresher:
- Pillar pages cover the big topic
- Cluster pages dive into specific subtopics
- Everything links together clearly
Don’t Ignore Visuals (AI Uses Them Too)
LLM-powered search is increasingly multimodal. Images, tables, charts, and videos help when they’re informative, properly marked up, crawlable (text vs. image) and utilize proper file names, alt text, etc.
Authority Extends Beyond Your Website
LLMs consider entity authorityand expertise. This means consistent branding across platforms, active expert participation (not using a corporate profile) in trusted communities, mentions in reputable publications, original research. Monitoring and responding to brand reviews is critical as well.
Technical SEO Still Matters
SEO fundamentals still apply. Content should be discoverable and indexable to search engines and LLMs. Cleaning up old pages & 404 errors, ensuring an XML sitemap is present, minimizing redirect chaings, proper canonicalization and hreflang tags, site speed, all the things! And, of course – and maybe most importantly – Schema! GSC is your friend even with LLMs.
A Quick Note on LLMs.txt
There’s growing interest in the proposed llms.txt standard which is a Markdown file that highlights key content for AI crawlers. It’s still early days, and not all LLMs respect it yet. But early adopters, including us, are testing and using it.
Final Thought
Optimizing for LLM discovery isn’t about gaming algorithms, but there’s a lot to learn from traditional SEO. It’s about delivering unique expertise that is easy for engines to crawl that provides answers to questions people are searching for!
Do that well, and visibility tends to follow. Looking for a content audit that meets the LLM demands of today? Contact us!