Profound just dropped the biggest AI search study ever conducted.

10 million AI search results. Every major engine. Real data.

And what they found changes everything we thought we knew about getting cited by ChatGPT, Perplexity, and Google's AI Overviews.

Here's what actually works (and what doesn't):

Each AI Engine Has Completely Different Taste

This was the biggest surprise. We assumed all AIs would favor similar content. We were wrong.

ChatGPT is obsessed with institutional authority. It loves Wikipedia, Forbes, Amazon, G2. If you want ChatGPT citations, you need to sound like an expert, not a friend.

Perplexity is the opposite. It craves user-generated content. Reddit threads, YouTube comments, LinkedIn posts, Yelp reviews. It wants real people talking about real experiences.

Google AIO doesn't care about domain prestige. It blends everything—corporate sites, social media, random blogs—as long as the technical implementation is clean.

Microsoft Copilot leans heavily B2B. Gartner reports, PCMag reviews, corporate case studies. It's the business professional of the AI world.

Translation: You can't optimize for "AI" anymore. You have to pick your battles.

Traffic Means Nothing to AI Citations

Here's the stat that broke my brain:

Citation volume has almost zero correlation with website traffic (r² = 0.05)

Pages getting 12 visitors a month are earning 900+ citations across AI engines. Meanwhile, high-traffic JavaScript-heavy pages are completely invisible to AI crawlers.

Your glossary page that nobody visits? It could be your biggest AI asset.

Your homepage with 10,000 monthly visits? If it's built in React without SSR, AIs can't see it.

JavaScript Is Killing Your AI Visibility

This one's simple and brutal: AI bots don't execute JavaScript.

All those fancy dynamic components? Invisible.
Content that loads via API calls? Invisible.
Text hidden behind modals or tabs? Invisible.

If you view-source your most important pages and can't see the content in raw HTML, neither can the AIs.

The fix: Server-side render everything that matters. No exceptions.

Listicles Are Dominating AI Citations

32.5% of all AI citations come from comparative listicles. Not blogs. Not how-tos. Listicles.

The top-performing formats:

• Comparative Listicles: 57.6M citations
• Opinion/Analysis: 17.5M citations
• Product Pages: 8.3M citations

"Top 10 [Tools] for [Use Case] in 2025" isn't just good for SEO anymore. It's AI citation gold.

Why? Because AIs love extracting clean, structured comparisons. Give them a numbered list with clear pros/cons, and they'll quote you all day.

Fresh Content Gets Indexed in 48 Hours

Here's something that shocked even me: AI search engines are indexing new content within 2-3 days.

Publish a well-structured guide on Monday, and by Thursday it could be showing up in Perplexity citations.

But here's the kicker: recency signals boost citation probability across all engines.

Adding "Updated 2025" to your title isn't just good practice—it's competitive advantage. Fresh timestamps = higher citation odds.

Micro-Niches Beat Broad Topics Every Time

Generic content gets ignored. Hyper-specific content gets quoted.

Instead of "Best Marketing Tools" (boring, competitive), try "Email Marketing Platforms for SaaS Startups Under 50 Employees" (specific, quotable).

Instead of "AI in Business" (vague), try "How Small Agencies Use AI to Automate Client Reporting" (precise, actionable).

The more specific your answer, the higher your chances of being THE answer an AI provides.

The Universal AEO Playbook

Based on 10 million data points, here's what actually works:

1. Structure Everything

• Questions in H2 tags
• Answers in the first 2 sentences
• FAQ schema on everything
• Clean URL slugs with keywords

2. Write Meta Descriptions That Answer Queries

Your meta description should BE the answer, not tease it.

Bad: "Learn about our amazing AI tools..."
Good: "AI sales tools automate prospecting, lead qualification, and outreach personalization—here are the top 10 platforms compared for 2025."

3. Go Static or Go Home

Server-side render your critical content. Use Next.js getStaticProps, Nuxt generate, or plain HTML. Just make sure AIs can crawl it without JavaScript.

4. Target Answer-First Queries

Don't write articles. Write answers. Every page should solve one specific problem completely.

The Takeaway

10 million AI search results don't lie.

The brands winning in AI search aren't the ones with the most traffic or the biggest budgets. They're the ones with the clearest answers, cleanest structure, and most confident content.

If your content can't be easily parsed by an AI crawler and confidently cited by a language model, you're already losing.

The good news? Most brands haven't figured this out yet. The playing field is still wide open.

But not for long.

Time to start optimizing for answers, not rankings.

— Sam