SEO vs GEO: How ChatGPT Search Really Works in 2025 (123 Real Citations)

I have to confess, I got curious when I read a discussion on Reddit in late 2025 that insisted ChatGPT's web search wasn't powered by Bing. Wait, what? If not Bing, then how does it search?

Additional evidence came from Backlinko's research, which concluded that ChatGPT pulled information from Google's index.

This got me thinking: If ChatGPT is using Google's search infrastructure, how does it decide which sources to cite? Do the same sites that rank well in Google also get cited by AI? And honestly, what does this mean for the future of SEO?

I had to find out. So I ran my own test.

Understanding the Search Landscape

Before diving into the results, let's understand what we're dealing with. There are fundamentally different search ecosystems at play:

Google

Google: Commands 91% of global search market share (Statcounter, 2025). Shows ranked lists of links (SERPs) based on relevance, authority, and hundreds of ranking factors.

Bing

Bing: Microsoft's search engine with 3.6% global market share (Statcounter, 2025). Uses similar ranking methods but with a less comprehensive index, especially for international content.

ChatGPT Search

ChatGPT Search: Doesn't show ranked lists. Instead, it reads and synthesizes information from multiple sources, then presents a direct answer with citations. Most importantly: ChatGPT retrieves about 10-20 sources but typically only cites 3-5 of them.

Methodology: Testing 100+ Real Citations

I spent a few weeks going through 123 real ChatGPT citations - everything from local coffee shops to product reviews to tech stuff. Basically anything people actually search for.

The Technical Setup

I used an AI playground setup that let me see exactly what ChatGPT does when it searches. The cool part? I could see the actual search queries it was making behind the scenes.

The interesting bit: ChatGPT doesn't search for exactly what you type. It transforms your question first.

For example:

• My question to ChatGPT: "What are the cheap coffee shops student friendly in Chicago?"
• ChatGPT's web search query: "cheap coffee shops Chicago student friendly 2025"

See what it did? Stripped out the "What are the" part, cleaned up the phrasing, even threw in "2025" to get fresh results. That's what actually gets searched.

ChatGPT's Citation Filtering Process

Something I noticed: ChatGPT pulls from way more sources than it actually cites.

It searches, gets a bunch of results, then picks and chooses which ones to actually mention. Most sources get filtered out.

Every citation I analyzed passed through two filters:

1. Search engine filtering: Which pages rank for the query
2. ChatGPT's filtering: Which of those ranked pages ChatGPT decides to cite

That second filter is huge. ChatGPT is basically being an editor, deciding what's actually worth citing. This is the key to understanding AI visibility.

Query Distribution

I tested all kinds of different queries to see how ChatGPT handles different types of searches:

• Local Business Recommendations (45%): Bakeries, restaurants, hotels, coffee shops, gyms, and other location-specific services
• Service Recommendations (21%): Language courses, programming courses, repair services, specialized studios
• Product Recommendations (14%): Electronics, vehicles, sporting goods, health products
• Real-Time Information (7%): Flight schedules, currency exchange rates
• Pricing & Financial Advice (7%): Pricing information for services, financial decision-making questions
• Activity & Experience Recommendations (7%): Hiking trails, outdoor activities

Pretty much what normal people search for - lots of local stuff, some product research, that kind of thing.

Research Sampling: 3 Real Examples

Here are 3 actual queries from my research showing how ChatGPT's citations compare to Google and Bing rankings:

The Verification Process

For every single citation, I checked:

• Whether the site appeared in Google's first 10 pages using the exact query ChatGPT used (not my original question, but ChatGPT's transformed query)
• Whether the site appeared in Bing's first 10 pages using that same transformed query
• The exact ranking position if found (page number and position on page)
• Whether "not found" sites were functional (accessible via direct URL or truly dead)

I tested in both English and Portuguese - wanted to see if language made a difference. Just used regular ChatGPT with web search turned on.

Main question: does ranking in Google/Bing actually matter for AI citations?

For practical strategies on optimizing your content for both traditional search and AI platforms, see my guide on advanced tactics for optimizing content for ChatGPT.

Key Finding #1: Nearly Half of Citations Don't Rank in Traditional Search

The most striking discovery: 48.4% of sites cited by ChatGPT didn't appear in Google's first 10 pages (100 results) when searched using the exact query ChatGPT used.

Here's the thing that a lot of AI-focused discussions miss: about half of ChatGPT's citations come from sites that rank in the first 10 pages of traditional search. But the other half? They either rank way beyond page 10 (position 100+) or aren't indexed at all. Invisible in normal search, but ChatGPT cites them anyway.

Key Finding #2: The Invisible Web That ChatGPT Sees

For Bing, this number was even higher: 8 functional sites cited by ChatGPT but not indexed.

Which made me wonder:

How is ChatGPT finding and citing sites that traditional search engines either:

1. Haven't indexed at all, or
2. Have ranked so poorly (beyond position 100) they're effectively invisible?

This remains an open area of investigation, and one that will likely become clearer as OpenAI provides more transparency about SearchGPT's data sources.

Key Finding #3: Ranking Position Matters (When You're Indexed)

For sites that were indexed in both traditional search and cited by ChatGPT, I analyzed their ranking positions to understand if there's a correlation.

The pattern that emerged:

• Sites ranking in positions 1-10 had the highest citation probability
• Sites ranking in positions 11-30 still got cited, but less frequently
• Sites ranking beyond position 30 rarely got cited (unless they had unique information)

Looks like ChatGPT values relevance and specificity over pure authority - pretty different from how traditional search works.

The specificity advantage:

I noticed something interesting: Sites with highly specific, niche content got cited even when they ranked poorly for broader terms.

For example, a small local blog about "student-friendly coffee shops in Chicago's Logan Square" got cited even though it ranked beyond page 5 for the general query "coffee shops Chicago".

ChatGPT seems to value precision and relevance to the specific query over general domain authority.

The Review Platform Paradox

Here's something unexpected: Major review platforms like Yelp, TripAdvisor, and HomeAdvisor were significantly underrepresented in ChatGPT's citations compared to their traditional search rankings.

These platforms typically dominate traditional search results for local business queries. But in my testing, they accounted for less than 8% of ChatGPT's local business citations.

Instead, ChatGPT preferred:

• Direct business websites
• Niche local blogs and city guides
• Reddit discussions about local recommendations
• Independent review sites

SearchGPT vs Google vs Bing: Which Search Engine Dominates?

What really surprised me: Google beat Bing at predicting which sites ChatGPT would cite.

By the Numbers

• Google: 51.6% found in first 10 pages
• Bing: 45.9% found in first 10 pages

That 5.7 percentage point gap in the SearchGPT vs Google vs Bing comparison might seem small, but it actually matters quite a bit when you're trying to optimize for AI search.

The thing about SearchGPT vs Bing: even though ChatGPT's official partnership is with Microsoft (Bing's parent company), my data shows SearchGPT vs Google performance is stronger.

What This Means for International SEO

Interestingly, the gap widened for non-English queries. Portuguese-language searches showed even stronger correlation with Google indexation than Bing.

How to Optimize for AI Search: Practical SEO for ChatGPT Strategies

Alright, based on everything I found, here's what you should actually do:

1. Don't Compete Head-On With High-Authority Sites

Big authoritative sites dominate the broad queries. Don't try to beat CNN or The Verge at their own game - find a different angle.

Action items:

• Identify areas where major competitors don't create content
• Build unique angles based on: original research, local expertise, niche specialization, proprietary data or methods
• Target long-tail keywords with lower competition but high relevance to your expertise
• Focus on "different" rather than "better"—you don't need to beat CNN Brasil, you need to cover what CNN Brasil doesn't

2. How to Optimize for AI Search: Focus on Specificity, Not Just Volume

Action items:

• Create deeply specific, niche content answering precise questions
• Don't just chase high-volume keywords; create comprehensive resources for specific use cases

3. Leverage Community Platforms Strategically

My data showed strong citation patterns for Reddit and Wikipedia content (though these patterns are evolving). While you likely can't create your own Wikipedia page, you can strategically use Reddit:

Action items:

• Identify niche subreddits (not broad industry ones) where your target audience asks questions
• Ask hyper-specific long-tail questions that your content answers (authentic discussion, not spam)
• Subtly mention your brand/solution while seeking community feedback
• Focus on communities where detailed, helpful responses are valued (avoid meme-heavy subs)

4. SEO for ChatGPT: Rethink Local Search Priorities

For local businesses, here's the thing about review platforms:

Action items:

• Invest in your own website first, review platforms second
• Create detailed "About," "Services," and location pages with specific information
• Add blog content answering local-specific questions
• Include neighborhood names, landmarks, and hyper-local details
• Maintain review profiles, but don't neglect your owned properties

5. Understand SEO vs GEO (Generative Engine Optimization) Differences

With the rise of AI-powered search, the key difference in SEO vs GEO strategies is becoming clear. SEO vs GEO: traditional SEO focuses on ranking, while understanding Generative Engine Optimization (GEO) is crucial for those learning how to optimize for AI search engines like SearchGPT.

Why this works:

• Less competition in the long tail: Fewer sites target hyper-specific queries
• ChatGPT values precision: AI engines reward content that directly answers the exact question asked
• Lower barrier to entry: You don't need massive domain authority to rank for "best vegan bakeries in Austin's East Side"
• Coexistence strategy: Even as mastering both SEO vs GEO (Generative Engine Optimization) becomes essential to succeed in SearchGPT vs Google rankings, you can build AI visibility while growing traditional SEO authority