Everything YouTube taught me about AI search

I just binge-watched every bit of YouTube wisdom about making your brand rank in AI search engines. The outcome? This guide.

It covers technical breakdowns, actionable advice, and real examples. No marketing fluff, just fundamentals.

"Has anyone actually seen results from focusing on LLM SEO?"

Yes, brands are seeing real results. Here is how you can too.

The SEO You Knew Is Dead

For decades, search was simple. You published content, Google ranked it, and users clicked through to your website.

That system is breaking down.

Over the last six months, AI has inserted itself between users and websites. ChatGPT's market share is up 400%. Google is down 2.15%, which is the first dip in a decade.

The way people get information is changing.

The AI Middleman

An AI model now sits in the middle of every search. User to AI to website. ChatGPT, Google AI Overviews, Perplexity, Microsoft Copilot. These answer engines are not just delivering results.

They are summarising your content. They decide if your brand, product, or service even merits being in the conversation. The AI model handles the direct interaction with the user. That final mile of information delivery is all AI.

This matters because your website no longer has a direct relationship with the user. The AI does.

Traditional SEO Metrics Do Not Work Here

Researchers analysed 41 million AI search results. The data tells a surprising story.

Traffic does not predict AI citations. 95% of AI citation behaviour cannot be explained by traffic. High-traffic sites often never get cited. Low-traffic sites can dictate entire industry narratives.

Backlinks do not predict AI citations either. 97.2% of AI citations cannot be explained by backlinks. In fact, there is an inverse relationship. Sites with 1 to 9 backlinks average 2,160 AI citations. Sites with 10 or more backlinks average only 681 AI citations.

The SEO strategies that worked for Google do not transfer to AI search. You need a different approach.

How AI Search Actually Works

AI search is not just another SEO channel. The architectures are fundamentally different.

ChatGPT and the Bing connection

ChatGPT uses Bing for the heavy lifting. When you type a long query, ChatGPT normalises it into 5 to 7 keywords. You have already lost control of the search at this stage. It is now ChatGPT's interpretation of what you asked.

That simplified query hits the Bing API. Bing returns just three things for each result. URL, title, and meta description. You can see exactly how ChatGPT searches the web with our Chrome extension.

But Bing is not the final decision-maker. The WebGPT Agent is. This agent chooses which web results to actually crawl. It parses the page HTML, picks out the most useful content and links, and then decides what to include in the answer as a citation.

Optimising for ChatGPT means optimising for WebGPT. Not ChatGPT itself. Not purely Bing. It is a distinct optimisation target.

Perplexity is completely different

Perplexity is semantic and vector-based. There is no traditional indexing. There is no traditional search engine under the hood.

Perplexity encodes your query and finds the single best-matched piece of content from its massive, proprietary, cached directory of the web. It does not do live searches for most queries. This means your content needs to already be in its cache before someone searches for it.

Each AI search engine has different biases

Google AI Overviews looks fairly domain-agnostic. It presents a relatively fair distribution of cited sources. Top cited sites include YouTube, LinkedIn, Gartner, Reddit, Quora, and Wikipedia.

Microsoft Copilot has a massive bias toward corporate and business content. Forbes alone has 2.1 million citations. Gartner has 1.3 million.

ChatGPT loves Wikipedia. Across 177 million AI citations, Wikipedia is the most popular domain. It is cited six times more than any other source.

Perplexity is UGC-focused. It loves user-generated content from platforms like Reddit, YouTube, LinkedIn, and Yelp. It wants to know what real people are discussing.

The New AI SEO Playbook

AI search is controllable. The playbook looks different, but you can still win.

If you are not indexed by Bing, you will never show up on ChatGPT. Get indexed first. But do not stop there. ChatGPT's results only overlap 26% with Bing search results. You still need a distinct AI optimisation strategy.

Put your content in HTML. If your site is heavily client-side rendered or relies on JavaScript for content delivery, crawlers will not see it.

Create an llms.txt file. This new protocol acts like robots.txt, but for Large Language Models. It is a markdown-formatted map that makes your content digestible for AI. It tells AI where your key content is, where to find it, and why it should click through. AI search engines already use llms.txt as frequently as sitemaps.

Use semantic URL slugs. WebGPT sees your URL slug. If your URLs are full of codes and random strings, you are losing on one of three fronts for AI signals. Make your URL slugs descriptive.

Spoil your content in the meta description. This sounds counterintuitive. But it works. Offer value upfront. Tell the AI what the content delivers. Do not tease. Give the answer. A meta description like "The best business credit cards include Ink Business Unlimited, Venture X Business and U.S. Bank Business Triple Cash Rewards" is a green flag for AI answer bots. It gets cited frequently.

Content Strategy for AI Search

Content is still king. But AI search favours formats that are easy to process.

32.5% of AI citations are listicles or comparative content. Blogs and opinion pieces account for only 9.9%. AI wants pre-digested, structured insights. Structure your content accordingly.

AI also has a recency bias. Recently updated content gets favoured. Update your content regularly. Make it feel fresh, even if it is evergreen. AI picks up and cites content on the scale of days and weeks, not months and quarters. Visibility can change quickly.

A fintech brand boosted citation share by 7.7% in a month by optimising existing content for AI queries. A remote staffing client saw visibility jump from 0.1% to 5.2% in a single day by releasing strategic content. Within a week, they were a top 10 brand in the multi-billion-dollar remote staffing industry.

Perplexity loves user-generated content from Reddit, YouTube, LinkedIn, and Yelp. Engage directly where the conversations are happening. A cloud computing brand's Google AI Overviews visibility jumped from 23.8% to 73% in a month just by commenting on a few influential Reddit threads.

The Opportunity

Most brands are about to lose the majority of their direct connections to customers. Generative search controls the narrative now.

But if you are reading this, you have an opportunity.

Understand what people are searching for in AI.

• Track your AI visibility.
• Monitor citations.

See exactly how your content and competitors shape the AI narrative.

The brands that invest in AI visibility now will be the ones that get cited. The ones that ignore it will not appear in AI answers at all.

---