There’s a lot of conflicting advice regarding the best way to optimize and structure content for GEO (generative engine optimization).
Some experts suggest using schema markup and chunk optimization. While others call these methods a waste of time.
So, I decided to find out for myself.
I spent the past year testing different tactics to understand how AI models read, segment, and summarize web content.
Here’s what I learned.
The fundamental difference between optimizing content for SEO vs GEO
During my experiment, one thing I noticed is that LLMs often cite pages that don’t even rank in Google’s top 10.
For example, when I asked ChatGPT for the most beautiful places in the world, the top source was an article from Travel + Leisure.
The surprising part? That page didn’t even rank on the first page of Google search results; it sat at the bottom of page two.
My findings align with recent AI SEO studies:
- Only 12% of URLs cited by ChatGPT, Perplexity, and Copilot rank in Google’s top 10 search results.
- 80% of LLM citations don’t even rank in Google’s top 100 for the original query.
- Only 14% of URLs cited by AI Mode rank in the top 10.
That tells us one crucial thing: Large language models (LLMs) don’t retrieve and rank content the same way Google does.
Instead of evaluating an entire page against hundreds of ranking factors, AI systems focus on meaning and context. They look for clear, self-contained information that can be confidently reused in an answer regardless of where the page ranks.
This explains why pages that never reach Google’s first page can still appear in ChatGPT, Perplexity, or AI Overviews.
The easier it is for AI models to understand and interpret your content, the more likely it will be parsed, summarized, and cited.
How I structure content for GEO in 2025
So, how do you structure your content for AI inclusion?
Here’s the exact content strategy I used to help my clients earn citations in AI summaries.
Image source: Position Digital
Spell the brand name instead of “we” or “our”
This might sound like a small change, but it has a huge impact on how AI systems interpret and cite information about your brand.
Large language models rely on entity recognition to understand who’s speaking and attribute insights correctly.
When you say “we” or “our agency”, AI models often struggle to associate that statement with your brand.
But when you write “HubSpot recommends…” or “According to HubSpot’s findings…”, you’re reinforcing that entity every time.
I applied this approach while refreshing a blog post for one of my clients, HR Datahub.
Throughout the article, I consistently referred to the company by name: HR Datahub helps, HR Datahub brings, HR Datahub team, etc.
Implement a proper heading structure
It’s something that we always do in traditional SEO, but it’s even more critical for answer engine optimization (AEO).
This is because AI systems depend on clear hierarchy to understand context and relationships between ideas.
A well-structured heading layout not only improves user readability, but also enables LLMs to identify which sections answer which questions, making your content easier to summarize and cite.
Each H2 should represent a specific question or theme that aligns with user intent. Your H3s should then expand on that topic logically with supporting details or steps.
Image source: Semrush
Focus on one idea per heading
Avoid mixing multiple ideas within the same headline. If a heading reads like it could answer two different questions, break it into two smaller, more specific sections.
This helps AI systems map each paragraph to a clear topic and prevents confusion during information retrieval.
Use the BLUF (Bottom Line Up Front) principle
Each AI model has a limited token window, which means they’re more likely to retrieve information that appears early in your content.
So, don’t bury the most important piece of information at the bottom.
Instead, present the main takeaway first, then use the rest of the section to expand on it. Explain your reasoning, add supporting data, and include examples that reinforce your argument.
Here’s a good example of the BLUF principle from one of my recent articles.
This section immediately starts with a concise summary of why tracking AI brand visibility is important, before providing more data.
Keep paragraphs short and self-contained
AI systems read and process information in chunks, not pages.
If your paragraphs are too long or cover multiple ideas, models struggle to decide which sentence best answers a query.
In contrast, short and self-contained paragraphs give LLMs clear boundaries and make it easier for them to extract relevant insights without losing context.
Using an example from the same article, here’s the paragraph I wrote about the importance of tracking branded searches:
ChatGPT actually picks up that exact paragraph (albeit with some rewriting) and use it in one of its summaries.
Use lists and tables
AI loves structured information and segmented data.
Whenever possible, break down long paragraphs and walls of text into smaller, digestible formats like:
- Bullet points to highlight individual ideas or concepts
- Numbered steps to guide AI systems through a sequence or process
- Tables or charts to show comparisons and data at a glance
This structure helps large language models (LLMs) quickly understand your key points and reuse them in summaries or citations.
Add a FAQ section
Back in June 2025, Chris Green conducted a small experiment on how content structure matters for AI search.
He found that Q&A is the best format to structure content for GEO, simply because it provides the highest semantic relevance to user queries.
Chris also revealed that structured content (headings and lists) is almost as effective as Q&A for non-question queries, while dense prose is unsurprisingly the worst format for AI inclusion.
We had major success with the Q&A format when we refreshed a client’s article about working in Saudi Arabia.
As part of the update, we added a new FAQ section based on real user questions in Reddit threads, expat forums, and Google’s autocomplete.
The blog post ended up getting citations in both AI Overviews and AI Mode for 17 keywords, several of which came directly from the new FAQ section.
Here’s the Oriel Partners case study if you want to learn more.
How I’ll adapt my GEO strategy for the future
AI systems are evolving fast, and the way they interpret, retrieve, and cite information will continue to change.
As AI gets smarter, here’s how I plan to adapt my GEO content creation strategy.
Keep content easy to read (and parse), but there’s no need to chunk
Headings, lists, FAQs, and concise paragraphs remain the foundation of GEO. Clear formatting helps both AI and readers grasp the points you’re trying to say quickly and easily.
But that doesn’t mean you need to “chunk” your content into one-sentence paragraphs or break every idea into fragments.
Why? Because each AI model uses a different chunking method, and you can’t control it.
And more importantly, over-segmentation can actually hurt flow and comprehension for both humans and machines.
So instead of trying to match a model’s internal logic, focus on making your content naturally coherent. Write for people first with clarity and structure. If it reads well to a human, it will parse well for AI.
Implement schema markup
I never really experimented with schema markup, because there were a lot of debates regarding its importance for GEO.
Some people argue that LLMs don’t really understand schema markup; they mostly read plain HTML text.
While it’s true that early AI models struggled with structured data due to tokenization limitations, today’s AI systems are a lot more advanced.
Modern retrieval pipelines routinely segment pages, read JSON-LD schema, and map entities across documents.
Schema isn’t a “ranking factor” per se, but it reduces ambiguity and hallucination. It ensures that AI will understand exactly:
- Who’s speaking (Organization/Person)
- What the page is (Article/Recipe)
- Where concise answers live (FAQ/HowTo)
Here’s a comprehensive guide that explains what schema markup is and how to add it to your website.
Use more multimodal content
AI search systems are increasingly capable of retrieving information from multimodal content.
So, I’ll definitely use more images, tables, charts, and videos in my future articles.
However, these systems still rely heavily on the text surrounding those visuals to understand and interpret them accurately. Alt text, captions, labels, and nearby explanations remain crucial signals for both AI and traditional search engines.
Here’s how to make your visuals machine-friendly:
- Use normal HTML for images: Avoid lazy loading or JavaScript-only image rendering, since many AI tools can’t see images that rely on scripts to load.
- Write clear image alt text: Describe what the image shows and include a few words about the topic so AI and search engines understand it.
- Add short captions: Place them right below or next to the image or video, and use them to explain what’s being shown.
- Use the correct markup: HTML tags like <figure>, <figcaption>, and <table> help machines read your visuals correctly.
- Don’t upload pictures of tables: Create real HTML tables instead because they’re easier for AI to read, summarize, and reference.
Further reading: If you’re not sure how to make video content for your agency, check out our B2B short-form video tutorial.
Over to you
Generative engine optimization is still new territory, and no one has all the answers yet.
But one thing is clear: well-structured content will continue to be the key to earning LLM mentions and citations.
Focus on clarity, hierarchy, and readability. Create content that’s easy for large language models to parse and engaging for real people to read.
This is where StoryChief’s Content Editor tool comes in handy.
You can use its AI assistant to write well-structured articles, or suggest improvements to optimize existing content.
It also gives you a readability score based on the Flesch Reading Ease test, so you can instantly see whether your content is clear and easy to follow.
When it’s all done, you can integrate with your favorite CMS and automatically publish your content.
Sign up for free and get started today.
