{"id":1434,"date":"2025-08-13T18:18:42","date_gmt":"2025-08-13T18:18:42","guid":{"rendered":"https:\/\/dreamwarrior.com\/blog\/?p=1434"},"modified":"2025-12-09T21:07:57","modified_gmt":"2025-12-09T21:07:57","slug":"how-to-build-up-your-generative-ai-optimization-engines-part-2","status":"publish","type":"post","link":"https:\/\/dreamwarrior.com\/blog\/how-to-build-up-your-generative-ai-optimization-engines-part-2\/","title":{"rendered":"How to Build up your Generative AI Optimization Engines – Part 2"},"content":{"rendered":"

Step 1: Be Crawlable<\/span><\/h3>\n

This<\/span> may even sound simple, but this is a vital preliminary process. Your best bet for getting seen in large language models is to enable them to crawl your site. There are many LLM crawlers, such as OpenAI and Anthropic.<\/span><\/p>\n

Specific crawlers are likely to act so predatorily that they can provoke scraping and DDoS mitigation actions. If you have automated blocks against aggressive bots, align with your IT team to make sure that you do not block the LLM crawlers that you do want to access.<\/span><\/p>\n

When you deploy a CDN, such as Fastly or Cloudflare, your LLM crawlers should not <\/span>be blocked<\/span> by default.<\/span><\/p>\n

Step 2: Continue Gaining Search Engine Optimization Rankings<\/span><\/h3>\n

The best strategy for GEO is relatively simple: target traditional SEO. Rank highly on Google (Gemini and AI Overviews), Bing (ChatGPT and Copilot), Brave (Claude), and Baidu (DeepSeek).<\/span><\/p>\n

SEO remains crucial in Google Search and other search engines. However, the introduction of AI search with its advanced models is changing the game of generating and providing search results by eliminating the need to match keywords merely and instead offering highly context-sensitive and personalized search results.<\/span><\/p>\n

Step 3: Target the Query Fanout<\/span><\/h3>\n

The current generation of LLMs does a little more than simple RAG. They generate multiple queries. <\/span>This<\/span> is called <\/span>query fanout<\/span><\/a>. LLMs analyze user queries and process entire sequences of text, allowing them to <\/span>generate<\/span> AI-generated responses that <\/span>are tailored<\/span> to user intent.<\/span><\/p>\n

As an example, I also recently asked ChatGPT, “What is the newest Google patent that SEOs talk about?”, and it did two web searches on the queries “latest Google patent discussed by SEOs patent 2025 SEO forum” and “latest Google patent SEOs 2025 discussed” (which returned identical results as the former query).<\/span><\/p>\n

Tips:<\/span><\/strong> Check typical query fanouts to your prompts and strive to acquire ranking on those keywords too.<\/span><\/p>\n

The most common fanout patterns I encounter in ChatGPT involve using the word “forums” to inquire about a topic in the public and attaching “interview” to questions about people. Also often mentioned is the current year (2025).<\/span><\/p>\n

Caution: fanout patterns can change across LLMs and may change over time as well. What we are seeing today may not be relevant in a year.<\/span><\/p>\n

Step 4: Keep Consistency Across Your Brand Mentions<\/span><\/h3>\n

It is a rather basic but essential exercise for all people, whether in a business context or otherwise. Ensure your presence remains consistent across all platforms, including X, LinkedIn, your web page, Crunchbase, and GitHub.<\/span> Always maintain the same profile description, regardless of where you are.<\/span><\/p>\n

When listing your occupation on multiple social sites, such as “GEO consultant on small business,” do not alter it to “AEO expert” on Github or “LLMO Freelancer” in your press releases.<\/span><\/p>\n

On ChatGPT and Google AI Overviews, people have claimed to see positive results within a couple of days by simply utilizing a consistent self-description throughout the web. The same can be said about PR coverage, too. The broader and better the PR coverage you get on your brand, the more you will get referred back to users by the large language models.<\/span><\/p>\n

Step 5: Avoid JavaScript<\/span><\/h3>\n

With SEO, we preferred that our developers to use\u00a0 as little\u00a0 JavaScript as possible.\u00a0 With GEO\/AEO, we <\/span>\u00a0insist on it!<\/span><\/p>\n

The majority of LLM crawlers are unable to process JavaScript. You are out if your main content <\/span>is buried<\/span> in JavaScript.<\/span><\/p>\n

Step 6: Embrace Social Media & UGC<\/span><\/h3>\n

LLMs appear to be heavily dependent on Reddit and Wikipedia. Both sites provide user-generated content on nearly any subject. And with various levels of community-based moderation, a lot of the junk and spam already gets weeded out. Platforms are now dealing with the issue of maintaining a balance between human contributions and the volume of AI-generated content and posts.<\/span><\/p>\n

Donors can manipulate both platforms, but the general content trustworthiness of those systems is much more customary than it is on the internet. Besides, the two are constantly updated.<\/span><\/p>\n

Reddit provides valuable insights to LLM labs on how individuals can engage in online discourse, the terminology used in various topics, and knowledge of specific specialty areas.<\/span><\/p>\n

It is only logical to suppose that user-generated content that <\/span>is moderated<\/span> on websites such as Reddit, Wikipedia, Quora, and Stack Overflow will remain relevant in large language models.<\/span><\/p>\n

I do not suggest spamming those sites, but whenever you have such an opportunity to influence how your brand and competitors <\/span>are represented<\/span> there, it might be worth considering.<\/span><\/p>\n

Step 7: Create For Machine-Readability & Quotability<\/span><\/h3>\n

Write in a manner or style that large language models (LLMs) will understand readily and tend to cite. Although there is nobody who has it right yet, the following approaches appear to work:<\/span><\/p>\n

Write indicating and factual words. Instead of typing something like “We are fairly confident that the shoe is good on our customers”, type something like this: 96 percent of buyers have self-reported being happy with this shoe.<\/span><\/p>\n

Add schema. It has <\/span>been argued<\/span> many times. Very recently, Fabrice Canel (Principal Product Manager at Bing) stated that schema markup can assist LLMs to comprehend your content.<\/span><\/p>\n

If you desire to <\/span>be quoted<\/span> in an already publicized AI Overview, then you are supposed to possess content that is approximately of the same length as the already available one. Having a high cosine can also assist, but you are not just supposed to copy the current AI Overview. And nerd-fans: yes, you can of course use dot product instead of cosine similarity, given normalization.<\/span><\/p>\n

In case you use some technical terms that you wish to explain in your content, do so, ideally in a single sentence.<\/span><\/p>\n

Include summaries of lengthy paragraphs of text, lists of reviews, tables, videos, and other forms of content formats that may be difficult to cite.<\/span><\/p>\n

Step 8: Optimize your AI-generated content<\/span><\/h3>\n

To be cited for some topics in some LLMs, it helps to:<\/span><\/p>\n