Why Darwin Works: Modern Search Is No Longer One Query, One Ranking, One Result

Fair warning before you keep reading: this is nerd-level content.

Way farther into the weeds than most business owners will ever need to go.

You do not need to understand Reciprocal Rank Fusion, query fan-out, retrieval systems, RAG, source selection, query graphs, or any of the other nerd shit in this article to use Darwin.

Honestly, you probably will not understand all of it.

That is fine.

The important part is that we do.

Darwin has been in development for more than a year, and the whole thing really started with something I noticed about a year ago.

AI search was not behaving like traditional Google search.

For most of the history of SEO, the model was pretty simple:

Somebody types a keyword. Google ranks 10 pages. You try to be one of them.

But when I started watching how AI search behaved, I could see that was not really what was happening anymore.

You could ask 1 question, but the system appeared to be researching several different versions of that question, pulling results from different searches, comparing them, and then using that information to construct the answer.

That sent me down a pretty deep rabbit hole.

And that led me to Reciprocal Rank Fusion — RRF — which, funny enough, isn’t some new AI thing at all. In fact, it predates the current generative AI boom by more than a decade.

RRF was introduced in 2009 as a way to combine multiple ranked result lists into 1 stronger result set.

And once I understood what RRF was doing, a whole lot about where search was heading started to click.

Darwin came out of that.

Not because Darwin is some RRF trick.

Not because we found a loophole in ChatGPT.

And not because we think there is some secret ranking formula nobody else knows.

Darwin came out of understanding the larger architecture:

Modern search is no longer necessarily 1 query, 1 ranking, 1 result.

And if search does not work that way anymore, websites should not be built that way either.

THE OLD SEARCH MODEL

Let’s use a contractor example.

Somebody in Raleigh needs a failed driveway culvert replaced.

Traditional SEO would probably start with:

culvert contractor Raleigh NC

You build a page around that phrase.

You optimize the title.

You mention Raleigh.

You talk about culvert installation.

You build some links.

Then you wait and try to move up the rankings.

That worked because the search interaction itself was pretty constrained.

The customer searched.

Google returned a ranked list.

The customer clicked.

Done.

But people do not have to search like that anymore.

That same customer can now ask:

Who can replace a collapsed culvert under my driveway in Raleigh?

Or:

Water keeps washing out the end of my driveway. Who fixes that?

Or:

I need somebody near Raleigh who can dig out an old driveway pipe and install a larger one.

Or:

Every time it rains, gravel from my driveway washes into the road. Who should I call?

Those people may all need the exact same company.

But the language is completely different.

The customer does not even have to know what a culvert is.

They can simply describe the problem.

That is a massive change.

Traditional keyword SEO looks at those as separate searches.

Modern retrieval systems can understand them as different expressions of the same underlying need.

Darwin is built around that underlying need.

Not just the obvious keyword.

1 QUESTION CAN BECOME MANY SEARCHES

This is where it gets interesting.

Google has publicly described a process called query fan-out.

The basic concept is simple.

You ask 1 question.

The system can break that question apart and issue multiple related searches behind the scenes so it can investigate different parts of the problem.

Conceptually:

USER QUESTION

        ↓

"What do I need to fix this problem?"

        ↓

QUERY FAN-OUT

        ↓

Query A
Query B
Query C
Query D
Query E

        ↓

MULTIPLE RESULT SETS

        ↓

RETRIEVAL

        ↓

RANKING

        ↓

SOURCE SELECTION

        ↓

FINAL ANSWER

So go back to our customer saying:

The pipe under my driveway collapsed and water keeps washing gravel into the street. Who around Raleigh can fix this?

That 1 question can create searches around:

driveway washout repair Raleigh
culvert replacement Raleigh
residential drainage contractor Raleigh
collapsed driveway pipe repair
stormwater contractor Raleigh
culvert installation near Raleigh

The customer asked 1 thing.

The system can research the problem from several directions.

And that completely changes what a website needs to do.

You are no longer only trying to rank for the exact thing the customer typed.

You are trying to be relevant across the searches created from that question.

That is exactly what Darwin is built to do.

STOP THINKING ABOUT KEYWORDS AS A LIST

Traditional SEO loves spreadsheets.

Keyword.

Search volume.

Difficulty.

Position.

Pick 50 phrases.

Build pages.

Done.

But people do not search like spreadsheets.

They search based on what happened.

They search based on what they see.

They search based on what they think the problem is called.

Sometimes they know the industry terminology.

Most of the time they do not.

A homeowner may search:

culvert replacement Raleigh

Another homeowner may search:

pipe under driveway collapsed

Another:

why does my driveway keep washing out

Another:

who fixes drainage underneath a gravel driveway

Another:

need bigger pipe installed under driveway

Same service.

Different language.

Darwin does not care which version they use.

Darwin is built to cover the territory around the service.

The easiest way to understand this is to stop thinking about search demand as a list and start thinking about it as a graph.

Take:

CULVERT INSTALLATION

Around that central idea you have service searches:

culvert replacement
culvert repair
driveway culvert installation
road culvert installation
drainage pipe installation
culvert removal
culvert upgrades

Then problem searches:

driveway washing out
water crossing driveway
collapsed drainage pipe
flooded driveway
ditch overflowing
gravel washing into road
standing water near driveway

Then questions:

what size culvert do I need
how much does culvert installation cost
who installs driveway culverts
do I need a permit for a culvert
how long does culvert replacement take
what type of pipe should be used

Then locations:

Raleigh
Apex
Cary
Wake Forest
Garner
Durham
Clayton

Then materials:

HDPE
concrete
corrugated metal
plastic culvert pipe

Then buying intent:

price
replacement
repair
emergency
installation
contractor
estimate
permit
sizing

A customer can enter that graph from almost anywhere.

An AI system can move through several parts of that graph while researching the answer.

Every legitimate part of that graph Darwin covers becomes another place the business can be found.

The objective is no longer just:

Rank for “culvert installation.”

The objective is:

Own as much of the legitimate culvert-installation search graph as possible.

That is a much bigger game.

And Darwin is built to play it.

DARWIN BUILDS RETRIEVAL SURFACE AREA

This is probably the best technical description of what Darwin actually creates:

retrieval surface area.

Think about the average contractor website.

It might have:

Home
About
Services
Gallery
Contact

Maybe there is a generic excavation page.

Maybe drainage gets mentioned in a paragraph somewhere.

Maybe culverts are listed in a bullet point.

That website has a very small retrieval surface.

There are only so many searches where Google or an AI system can confidently look at that website and say:

Yes, this directly answers what this person is asking.

Now compare that to a website with legitimate coverage around:

Driveway culvert installation
Culvert replacement
Collapsed culvert repair
Driveway washout repair
Residential drainage installation
Standing water problems
Culvert sizing
Culvert replacement cost
Driveway drainage in Raleigh
Driveway drainage in Apex
Culvert installation in Cary
Stormwater drainage
Ditch restoration
Pipe replacement
HDPE culvert installation
Concrete culvert replacement
Driveway erosion repair

Those are not simply more keywords.

They are more retrieval surfaces.

More legitimate reasons for Google to return the company.

More legitimate reasons for an AI retrieval system to use the company.

More pathways into the website.

Darwin continuously expands those pathways.

That is the basic loop:

SEARCH → FIND → BUILD → RANK → REPEAT

Darwin watches real search demand.

It finds where the website is missing coverage.

It determines what those searches actually mean.

It confirms that the business really provides the service or serves the location.

Then it builds the correct resource around that opportunity.

Sometimes that is a new page.

Sometimes it is expanding an existing one.

Sometimes it is a location.

Sometimes it is a customer question.

Sometimes it is a problem somebody is trying to solve.

The goal is not another page.

The goal is another legitimate search your business can own.

Then that resource enters search.

It gets indexed.

It ranks.

It produces data.

Darwin finds the next opportunity.

And it does it again.

That is what self-evolving means.

NOT MORE CONTENT. MORE SEARCH COVERAGE.

This distinction is extremely important.

Anybody can build an AI content generator now.

We could make a system that spits out 10,000 pages.

That is not impressive.

AI can generate words forever.

Darwin is not:

Generate page.
Generate page.
Generate page.
Generate page.
Generate page.

Darwin is:

Observe demand.
Find uncovered demand.
Understand the intent.
Confirm the business actually serves it.
Determine whether existing content already covers it.
Build the correct resource.
Measure what happens.
Find the next opportunity.
Repeat.

The feedback loop is the system.

The pages are simply the visible output.

That is why we keep saying:

NOT MORE CONTENT.

MORE SEARCH COVERAGE.

There is a massive difference.

NOW LET'S GET REALLY NERDY: RECIPROCAL RANK FUSION

This is the part that originally sent me down the rabbit hole.

Reciprocal Rank Fusion, or RRF, solves a specific information-retrieval problem:

If I have several ranked result sets, how do I combine them into 1 stronger ranked result set?

The formula looks like this:

                 1
RRF(d) = Σ  -----------
             k + r(d)

Where:

d is a document.

r(d) is where that document ranked in a particular result set.

k is a constant that reduces the impact of extreme rankings.

The original research used:

k = 60

Now there is 1 important technical distinction here.

OpenAI has publicly confirmed that ChatGPT Search retrieves current information from the web and returns source-backed answers.

OpenAI has not publicly documented every piece of the internal ranking architecture behind ChatGPT Search or said that RRF is universally used for everything.

So I am not saying we cracked some secret ChatGPT algorithm.

We do not need that claim.

RRF is important because it demonstrates something much larger:

REPEATED RELEVANCE HAS MATHEMATICAL VALUE.

Imagine a system runs 5 searches around the same customer problem.

A Darwin client ranks:

Query 1: #3
Query 2: #5
Query 3: #2
Query 4: #7
Query 5: #4

A competitor ranks:

Query 1: #1
Query 2: not found
Query 3: not found
Query 4: #9
Query 5: not found

Traditional SEO looks at Query 1 and says:

The competitor won. They are #1.

But across the entire problem, Darwin keeps showing up.

Again.

Again.

Again.

Again.

Again.

The competitor owns 1 narrow entry point.

The Darwin client owns the problem.

That repeated relevance is an enormous advantage.

Using the original RRF formula, let’s take these searches:

Q1: driveway washout repair Raleigh
Q2: culvert replacement Raleigh
Q3: drainage contractor Raleigh
Q4: collapsed driveway pipe contractor
Q5: residential culvert installation Raleigh

Darwin client:

Q1: #4
Q2: #2
Q3: #6
Q4: #3
Q5: #5

Competitor:

Q1: #1
Q2: -
Q3: #8
Q4: -
Q5: -

Darwin:

1/64
+ 1/62
+ 1/66
+ 1/63
+ 1/65

≈ 0.078

Competitor:

1/61
+ 1/68

≈ 0.031

The exact numbers are not the important part.

The important part is the pattern:

REPEATED STRONG RELEVANCE BEATS ISOLATED RELEVANCE.

Darwin is not exploiting RRF.

Darwin is building the kind of search coverage that performs extremely well in a world where multiple searches and multiple result sets are being combined.

That is the bigger idea.

COVERAGE DENSITY IS THE MOAT

Now take this another step.

Suppose Competitor A has 1 absolutely incredible page about culvert installation.

Fantastic page.

Ranks #1.

Competitor B has useful, legitimate coverage around:

culvert installation
culvert replacement
culvert repair
driveway washouts
drainage problems
culvert materials
culvert sizing
installation costs
local permits
service locations
common failure conditions
collapsed pipes
storm damage
standing water
ditch restoration

Competitor A may absolutely beat Competitor B on 1 specific search.

But once the search system fans the original question out and starts investigating the problem from several directions, Competitor B keeps appearing.

That is coverage density.

The system keeps running into the same business across the subject.

Darwin builds that density continuously.

And that is where competitors start getting left behind.

A traditional website may launch with 20 pages.

A year later it still has basically 20 pages.

3 years later it may still have basically those same 20 pages.

Darwin starts with a foundation and then keeps taking search territory.

Maybe it begins with 40 meaningful retrieval surfaces.

Then it discovers 8 more.

Then 12.

Then 7.

Then 14.

Then 9.

Then 16.

Those numbers are examples, but structurally the point is the same.

The searchable footprint keeps expanding.

6 months later, a competitor is no longer competing against the Darwin website that launched 6 months ago.

They are competing against:

the original site + every legitimate search opportunity Darwin has discovered since launch.

A year later, that gap is larger.

2 years later, larger again.

That is why we say:

THE GAP KEEPS GETTING WIDER.

1 system keeps accumulating search territory.

The other sits still.

THE WEBSITE BECOMES A KNOWLEDGE SYSTEM

Historically, business websites have basically been digital brochures.

The company decides what it wants to talk about:

Home
About
Services
Gallery
Contact

Then somebody bolts SEO onto those pages afterward.

Darwin reverses that relationship.

The market helps determine what the website needs next.

SEARCH DEMAND
        ↓
GAP DISCOVERY
        ↓
INTENT ANALYSIS
        ↓
RESOURCE DECISION
        ↓
WEBSITE EXPANSION
        ↓
INDEXING
        ↓
RANKING
        ↓
PERFORMANCE DATA
        ↓
NEXT OPPORTUNITY

The website stops being a brochure.

It becomes a knowledge system representing the relationship between:

what the company actually does

and

how real customers actually search for it.

Darwin continuously closes the gap between those 2 things.

And the closer they get, the more search territory the business owns.

AI SEARCH MAKES THIS EVEN MORE IMPORTANT

Traditional Google search is only part of the equation now.

Modern AI search systems retrieve information from the web to construct answers.

That means there is another battle happening beyond simply trying to get a blue link.

You need to become:

a source the system retrieves.

Traditional search is largely asking:

Which page should rank?

AI-driven retrieval adds another set of questions:

What should I search?
Which pages are relevant?
Which passages answer this?
Which sources should influence the answer?
Which source should I actually show the user?

Before your business can become the answer, it has to enter that retrieval process.

Darwin creates more entry points.

Take 2 websites.

Website A says:

We provide excavation and drainage services throughout Raleigh.

That may technically be relevant.

Website B has legitimate coverage around:

• Culvert installation
• Culvert replacement
• Driveway washouts
• Residential drainage
• Standing water
• Ditch restoration
• Pipe sizing
• Culvert materials
• Cost factors
• Local service areas
• Common failure causes
• Storm damage
• Driveway erosion
• Drainage problems

Now somebody asks:

My driveway keeps washing out when it rains and I think the pipe underneath it may be too small. Who around Raleigh handles that kind of work?

Website A is relevant from 1 direction.

Website B is relevant from damn near every direction.

Darwin builds Website B.

Then it keeps building.

NATURAL LANGUAGE CHANGES EVERYTHING

For years, humans actually learned to talk to search engines.

Instead of typing:

My roof is 18 years old and after last night's storm I noticed water coming through the ceiling. Who around Atlanta can inspect it and tell me whether it needs to be repaired or replaced?

We learned to type:

roofer Atlanta

Why?

Because that was how Google worked.

AI changes that.

People can simply explain what is happening.

They can include the problem.

The location.

The urgency.

The symptoms.

What they think they need.

What they do not understand.

And then they can ask follow-up questions.

That explodes the number of possible search expressions.

A customer does not need to know the word culvert.

They can say:

The pipe underneath my driveway collapsed and now water keeps washing gravel into the street.

They do not need to know the term forestry mulching.

They can say:

I have a bunch of brush and small trees I want cleared without completely tearing the property up.

They do not need to know whether they need grading, drainage, excavation, or a French drain.

They can describe the problem.

Search systems can translate the problem into the service.

Darwin builds coverage on both sides of that translation.

The technical term.

The customer term.

The symptom.

The problem.

The service.

The question.

The location.

The buying intent.

That is how you own the search territory.

DARWIN DOES NOT HAVE TO PREDICT THE FUTURE

Traditional SEO is built around assumptions.

Somebody sits down today and decides which searches will matter.

They pick the keywords.

They build the pages.

And then the website stays stuck around those assumptions.

Darwin uses the market itself as feedback.

If customers start searching:

culvert replacement after storm damage

Darwin finds it.

If they start searching:

contractor to stop driveway from washing out

Darwin finds it.

If Wake Forest suddenly starts producing meaningful demand for culvert work, Darwin sees that.

If AI changes how people phrase these questions, Darwin sees that too.

We do not have to perfectly predict today how customers will search 2 years from now.

Darwin adapts when they do.

That is the self-evolving part.

THIS IS WHY DARWIN ACHIEVES #1 RANKINGS

There is no magic keyword.

There is no magic meta tag.

There is no single SEO trick.

Darwin achieves #1 rankings by going deeper into legitimate search demand than static websites do.

More services covered.

More locations covered.

More customer problems covered.

More questions answered.

More natural-language searches satisfied.

More retrieval surfaces.

Greater coverage density.

More ways for the business to enter the search process.

Your competitors are fighting over the obvious keyword.

Darwin goes after that keyword.

Then the searches around it.

Then the questions around those searches.

Then the customer problems behind those questions.

Then the locations.

Then the next service.

Then the next opportunity.

And it keeps going.

That is the advantage Darwin customers have.

And the longer Darwin runs, the larger that advantage becomes.

THIS IS NOT ABOUT GAMING GOOGLE

There is a huge difference between building search coverage and spraying garbage across a domain.

Automation used to manufacture thousands of junk pages is not Darwin.

Swapping city names into the same paragraph 500 times is not Darwin.

Creating pages for services a company does not provide is not Darwin.

Darwin builds where legitimate demand exists and where the business actually has a reason to be the answer.

There is a massive difference between:

manufacturing pages

and

building search infrastructure.

The first one is noise.

The second one is an asset.

And that asset compounds.

One new retrieval surface does not look like much.

Neither does the next.

But then there are 50.

Then 100.

Then 200.

Then entire clusters around individual services.

Then customer problems.

Then locations.

Then questions.

Then natural-language searches.

Eventually the website is no longer competing page against page.

It is competing system against system.

A static competitor has a website.

A Darwin customer has a search infrastructure that keeps expanding.

Those are not the same thing.

THE BIGGER IDEA

This is really what Darwin comes down to.

Search is becoming more conversational.

More contextual.

More multi-step.

More retrieval-driven.

More capable of breaking complex questions apart.

More capable of combining multiple result sets.

More capable of understanding that:

My driveway keeps washing out.

and:

Culvert replacement contractor Raleigh.

can represent the exact same commercial need.

That is a major shift.

And websites should not still be built around a search model from 10 or 15 years ago.

Search evolved.

So we built a website system that evolves with it.

Traditional SEO looks like this:

Research keywords
       ↓
Pick keywords
       ↓
Build pages
       ↓
Try to rank
       ↓
Wait

Darwin looks like this:

Observe actual search demand
            ↓
Find missing coverage
            ↓
Understand the intent
            ↓
Build the correct resource
            ↓
Rank
            ↓
Measure what happens
            ↓
Discover the next opportunity
            ↓
Repeat

That last word is the whole thing.

Repeat.

Search never stops changing.

Customers never stop changing the way they describe what they need.

New questions appear.

New locations matter.

New problems show up.

AI changes how information is retrieved.

Competitors change.

Markets change.

Darwin does not launch and sit there.

Darwin keeps going.

THE INTERNET BUILT STATIC WEBSITES.

SEARCH IS BECOMING DYNAMIC.

THE WEBSITE SHOULD BE TOO.

That is Darwin.

Not more pages just so we can say the website has more pages.

Not AI-generated garbage sprayed across a domain.

Not another SEO package.

Not a list of keywords somebody picked once and forgot about.

Darwin continuously expands the intersection between:

what your business actually does

and

how real customers actually search for it.

And when 1 customer question can become multiple searches, multiple result sets, multiple retrieval decisions, multiple rankings, and eventually 1 answer, being relevant across the entire problem is an enormous advantage.

That is why Darwin works.

SEARCH → FIND → BUILD → RANK → REPEAT.

Search evolves. Darwin evolves with it.