Published by NewsPR Today | July 2025
The emergence of AI-driven search platforms such as ChatGPT Search is causing a profound change in the search engine optimisation (SEO) landscape.
Reciprocal Rank Fusion (RRF) is used in ChatGPT Search’s ranking and citation retrieval process, according to a recent analysis by Metehan Yeşilyurt, a Growth Marketing Manager at AppSamurai and a well-known SEO thought leader.
For digital marketers looking to optimise content for AI-driven search, this discovery—described in Yeşilyurt’s article on metehan.ai—offers vital insights. Through the use of RRF, ChatGPT Search gives topical authority top priority and rewards websites that provide thorough, multifaceted content. We examine RRF, its mathematical underpinnings, and its consequences for SEO tactics below.
Reciprocal Rank Fusion (RRF): What is it?
Cormack, Clarke, and Büttcher initially described the rank aggregation technique known as Reciprocal Rank Fusion (RRF) in a 2009 paper. It highlights documents that routinely rank highly across various search queries by combining the results of several queries into a single, cohesive ranking. RRF employs rank positions to produce a strong, relevance-driven ranking without the need for significant tuning, in contrast to conventional search algorithms that depend on a single query or score normalisation.
The RRF formula is simple but effective:
RRF Score (d) = Σ (1 / (k + rank(q, d)))
Where:
- d is a document (e.g., a web page).
- q represents each query in a set of queries.
- rank(q, d) is the document’s rank in the result set of query q (starting from 1).
- k is a constant (typically set to 60) that smooths the score, ensuring lower-ranked documents have diminishing influence.
For example, if a document ranks 1st, 3rd, and 10th across three queries with k=60:
- Score = (1/(1+60)) + (1/(3+60)) + (1/(10+60)) ≈ 0.0164 + 0.0159 + 0.0143 ≈ 0.0466.
Because they exhibit constant relevance across several queries, documents with higher RRF scores are ranked higher in the final list. Without requiring the normalisation of incompatible score scales, this method makes RRF robust, scalable, and efficient for combining various relevance signals, such as keyword-based and semantic searches.
Yeşilyurt’s Finding: ChatGPT Search’s RRF
Yeşilyurt’s investigation, which involved inspecting Chrome DevTools while using ChatGPT Search, uncovered backend configurations that confirm RRF (Reciprocal Rank Fusion) is actively used in the ranking process. He found specific parameters in the developer console, such as:
To view this in action, open Chrome DevTools, go to the Network tab, and inspect the background fetch requests triggered by ChatGPT Search. Inside the response payloads or request headers, you may find keys like:
{
"ranking_method": "reciprocal_rank_fusion",
...
}
This proves that the system uses RRF in its server-side ranking pipeline.
These parameters reflect a standard RRF setup, where multiple queries are merged to produce a unified ranking based on how consistently a document performs across them.
This means that ChatGPT Search doesn’t just respond to the exact query entered. For a phrase like “best coffee makers”, it may run related searches in the background — such as “coffee machine reviews”, “home espresso machines”, and “top drip coffee makers” — and combine the results using RRF.
Let’s look at a simplified RRF score comparison to understand the impact:
-
Page A ranks:
1st for “coffee makers”,
10th for “best espresso machines”,
20th for “coffee machine reviews” -
Page B ranks:
4th for all three queries
Assuming k = 60:
RRF Score for Page A:
(1 / 61) + (1 / 70) + (1 / 80) ≈ 0.0164 + 0.0143 + 0.0125 = 0.0432
RRF Score for Page B:
(1 / 64) * 3 ≈ 0.0156 * 3 = 0.0468
👉 Even though Page A ranked 1st in one query, Page B is more consistently relevant and receives a higher final score. This illustrates how RRF mathematically favors topical depth and authority across related queries.
As Yeşilyurt puts it:
“RRF mathematically proves why topical authority and topic clusters work so well.”
Why Being a ‘Topic Expert’ Is Your New SEO Superpower
B If you were exceptionally skilled at one thing, you could make it in the early days of Google. You were in business if you had the best page for the precise term “coffee makers.”
However, things have changed since the advent of AI search (such as in ChatGPT or Google’s new features). The AI searches for more than one keyword. It seeks to identify the real authority on a given subject.
Let’s Break It Down with an Example:
Imagine two websites:
- Website A is ranked #2 for the search “coffee makers.” That’s it. It’s a one-trick pony.
- Website B is ranked a little lower, at #5 for “coffee makers.” But, it’s also #3 for “coffee maker reviews” and #7 for “best home coffee makers.”
Who does the AI think is the real expert?
Website B, by a landslide.
The AI sees that Website B knows about the entire world of coffee makers—the reviews, the types, everything. It adds up all those little wins. Website A might be great at one thing, but Website B is the go-to source for the whole subject.
So, How Do You Become a Topic Expert?
It’s pretty simple. Stop thinking about single keywords and start thinking about covering a whole topic.
1. Build a “Library” of Content
Think of your main topic as the big, central book in a library (e.g., “The Ultimate Guide to Coffee Makers”). Then, write smaller articles that link back to it, like individual chapters or pamphlets on specific things:
- “Best Drip Coffee Makers Under $100”
- “How to Clean Your Espresso Machine”
- “French Press vs. Drip: What’s the Difference?”
When you link them all together, you’re showing the AI that you haven’t just written one article—you’ve built a complete resource.
2. Cover All the Angles
People search in weird ways. They might look for “best coffee machine” or “top-rated coffee makers,” or ask, “What coffee maker should I buy?”
Use keyword tools to find these different questions and phrases. Sprinkle them naturally into your articles. This shows you understand what people are asking, no matter how they phrase it.
3. Connect Your Dots
Don’t just let your articles sit there alone. Link them together! When you mention “espresso machines” in one article, link to your other article that’s all about them.
This builds a web of information on your site, making it easy for the AI to see how all your knowledge is connected.
4. Go Deep, Not Just Wide
Forget about writing short, 300-word articles that barely scratch the surface. That’s not what an expert does.
Instead, write one amazing, detailed guide that answers every possible question someone might have. A single, high-quality, long article is worth more than ten flimsy ones.
The bottom line? The game has changed. To win with AI search, you have to stop trying to trick a robot and start acting like a genuine, helpful expert.
SEO in the Age of AI: What’s Next?
Yeşilyurt summarizes it well:
“This is about getting ready for a world where search is AI-native and content authority triumphs over keyword stuffing; it’s not just about ChatGPT.”
This change is reflected in Google’s June 2025 core update, which gives content trustworthiness and relevance more weight than backlinks or keyword density.
You can increase your audience’s value and rank higher on AI search by coordinating your SEO efforts with RRF-friendly strategies, such as semantic depth, structured content, and broad topical coverage.
Final Thoughts
As AI search platforms continue to evolve, Reciprocal Rank Fusion (RRF) is becoming a critical concept for modern SEO. It rewards holistic, authoritative content and penalizes narrow, keyword-only strategies.
To thrive in this new landscape, marketers must think beyond traditional keyword SEO and embrace content strategies that mirror how AI understands and ranks relevance.
Stay tuned to NewsPR Today for deep insights into the future of SEO, content strategy, and AI-driven digital marketing.
Sources:
-
Metehan Yeşilyurt’s article: Is RRF the Secret to Dominating AI Citations?
-
Original research via Chrome DevTools and ChatGPT Search
-
Google’s June 2025 core update analysis
