Last Tuesday I watched a client's blog post — ranked #3 on Google for a high-intent keyword — get completely ignored by Perplexity, ChatGPT, and Gemini. Zero citations. Not even a mention.
The post was 2,400 words, well-structured, great backlinks. But when I asked three AI engines the exact question it answers, none of them surfaced it. They cited a competitor's 800-word piece instead.
That gap between traditional SEO and what AI engines actually pull? That's the whole game now.
What's actually happening under the hood
Generative engines don't crawl and rank pages the way Google's web index does. They're synthesizing answers from a retrieval layer that favors:
> Generative Engine Optimization (GEO) is the practice of structuring content so that AI engines retrieve, parse, and cite it in their synthesized responses.
Traditional SEO optimizes for ranking in a list. GEO optimizes for being the source the AI quotes.
The experiment I ran
I took 12 existing blog posts across three clients. All ranked on page one of Google for their target keywords. I tested each one by asking the corresponding question to ChatGPT, Perplexity, and Gemini.
Results across 36 queries:
The optimized versions weren't rewritten from scratch. I applied a specific set of changes:
1. Lead with the answer
AI engines extract the first clear statement that matches the query intent. I moved the core answer to the top of each post — no intro fluff, no "In today's digital landscape..."
Before: 180 words of setup before the actual answer.
After: Answer in the first 40 words, then elaboration.
2. Use explicit Q&A formatting
I added an H2 that mirrors the exact question users ask, then answered it directly underneath. This alone increased citation rates by roughly 3x in my sample.
3. Add structured data where it matters
Not just schema markup — though that helps. I mean structuring the content itself:
AI engines love extracting clean structures. A comparison table I added to one post got pulled verbatim into three different AI responses.
4. Cite sources — and make them real
Posts that referenced specific studies, named the researchers, and linked to primary sources were cited 2.4x more often than posts making unsupported claims.
This tracks with what we know about how generative models weight source credibility. Vague authority ("studies show") doesn't work. Specific authority ("According to a 2024 Stanford study of 12,000 pages...") does.
If you want to see how your content currently performs in AI engines, tools like the GEO Audit Tool can give you a baseline before you start optimizing.
What didn't move the needle
I tested a few things that I expected to matter but didn't:
The uncomfortable implication
If you're still optimizing purely for Google's blue links, you're optimizing for a surface that fewer people interact with every month. The GEO vs SEO landscape is shifting fast, and the content strategies that win in one don't automatically transfer to the other.
I'm not saying traditional SEO is dead. I'm saying it's now a necessary-but-insufficient condition.
The brands that figure out how to be cited by AI — not just ranked by Google — are going to own the next few years of organic discovery.
One thing to try this week
Pick your highest-traffic blog post. Ask ChatGPT or Perplexity the exact question it answers. See if your post gets cited.
If it doesn't, you now know where to start.
And if you want to understand why some content gets pulled into AI answers while other content doesn't, check out the AI Gravity Checker — it reverse-engineers what AI engines are actually looking for in your content.
FAQ
Is GEO replacing SEO?Not replacing — layering on top. You still need traditional SEO fundamentals (crawlability, backlinks, technical health). GEO is about making your content visible to a completely different retrieval system.
How long does GEO optimization take to show results?In my experiments, changes showed up in AI responses within 48-72 hours. AI engines re-crawl and re-index faster than traditional search engines.
Does GEO work for all content types?It works best for informational and comparison content. Brand pages, product listings, and purely transactional content see less impact because AI engines tend to cite informational sources for answer-type queries.