Modern Day Masters

Mastering On-Page SEO: Kyle Roof Shares the Secrets You’re Missing

SEO changes constantly, but the stuff that moves rankings is surprisingly consistent. In this interview, Kyle Roof (known for SEO testing and on-page experimentation) breaks down what actually matters on-page, what’s mostly “SEO theater,” and how to build pages that rank without being overly dependent on backlinks.

This post is a practical, operator-focused breakdown of the best takeaways, plus a clean “how-to” you can apply to real client sites.

Who is Kyle Roof (and why anyone should care)

Kyle focuses on:

  • On-page SEO systems
  • SEO testing (controlled experiments, not vibes)
  • Tooling that measures on-page requirements based on what’s already ranking

He also has a reputation for pushing boundaries and proving what Google rewards (and what it doesn’t). The point isn’t the story. The point is the mindset: measure what works, and repeat it.

Big Takeaway #1: “Silos” don’t rank. Internal linking strategy does.

Kyle draws a hard line between two concepts people mix up:

1) Physical silo (URL folders)

Example:

  • /colors/
  • /colors/red/
  • /colors/blue/

This is useful for humans.

  • It improves navigation and clarity.
  • It does not directly boost rankings just because it’s structured.

2) Virtual silo (internal links inside content)

This is where ranking power actually comes from.

The model:

  • You pick a “money page” (the page that should rank and convert).
  • You publish multiple supporting pages that target related subtopics.
  • Supporting pages link to the money page.
  • Supporting pages link to each other (daisy chain / tight topical cluster).

Kyle frames it as a “reverse silo”:

  • Most SEOs think from homepage down.
  • He thinks from supporting content up.
  • Supporting pages do the heavy lifting, then pass strength upward.

Why this matters in the real world:

  • You become less dependent on backlinks.
  • Your rankings become less volatile.
  • Updates hurt less when your site has internal strength.

Modern Day Masters action step

If you only do one thing from this entire post:

  • Build supporting pages that exist only to rank and push authority into ONE target page.

Be strict about it.

  • These pages should not link out to other categories/services.
  • Their only job is to support that one target.

This is where most people fail because they want every page to do everything.

Big Takeaway #2: Flat vs folder URL structure (the real reason shorter often wins)

Kyle calls out a common myth:

  • “Shorter URLs rank better.”

His take:

  • It’s not the short URL itself.
  • It’s distance from the root domain.

In plain English:

  • The deeper a page is nested, the more work Google has to do to crawl, understand, and prioritize it.
  • Too many folders can slow indexing and reduce performance.

Rules of thumb from the talk:

  • Try not to go deeper than 2 folders if you can avoid it.
    • Category > subcategory > page is usually enough.
  • Avoid meaningless folders like:
    • /blog/
    • /page/
    • /product/
      If you use folders, make them real:
  • Use folders that are actual keywords and represent real pages a user could land on.

Also important:

  • Your menu structure does not need to match your URL structure.
  • You can have a “Blog” in your nav without using /blog/ in the URL.

Modern Day Masters action step

When planning site architecture:

  • Optimize URLs for crawl + clarity, not “perfect visual hierarchy.”
  • Don’t add folders just because it “feels organized.”

Big Takeaway #3: How to find endless content ideas (without copying what everyone wrote)

Kyle’s best content ideation trick is simple and underused:

Subscribe to industry publications and write about what’s coming next

Every niche has:

  • Trade publications
  • Online magazines
  • Forums and newsletters

They publish what’s coming in 6–12 months.

You can write about it now.
That means:

  • You rank before the wave hits.
  • When the topic becomes popular, you’re already established.

Example from the host:

  • Lawyers and class action lawsuits
  • The first decent explainer wins, everyone else fights for scraps later

Modern Day Masters action step

For any niche you publish in:

  • Find 5–10 industry sources.
  • Build a “future topics” queue.
  • Publish ahead of demand.

Most SEO content is reactive. This is proactive.

Big Takeaway #4: “Entity optimization” is mostly just writing with the right contextual terms

This section was messy in the industry, so Kyle simplifies it:

  • LSI keywords
  • Contextual terms
  • Entity optimization

Call it whatever you want.
What matters:

  • Google needs contextual terms to understand what the page is actually about.

He gives a great example with “kitchen”:

  • kitchen remodeling context (tile, sink install, paint)
  • kitchen appliances context (fridge, stove)
  • emotional/family context (gathering, holidays)

Same word, different meaning.
Context creates the concept.

The “Google NLP API” trap

Kyle warns against thinking:

  • “If Google’s NLP API shows entities, those entities are what I need to paste into the page.”

That won’t work.
You can’t just paste the words.
You need the context that makes Google interpret the page correctly.

Modern Day Masters action step

When editing or generating content:

  • Stop counting exact-match keywords.
  • Start ensuring the page includes the full vocabulary of the topic.

Big Takeaway #5: Keyword stuffing isn’t your real risk. Missing context is.

Kyle’s blunt guideline:

  • You probably need the main keyword fewer times than you think.

Core placement:

  • URL
  • Title tag
  • H1
  • Once in the body

Then spend the rest of the page doing the real work:

  • Contextual terms
  • Subtopics
  • Supporting details Google expects to see

If your page lacks contextual coverage, it won’t rank well.

This connects to his AI content warning:

Why AI content often underperforms

Kyle’s claim (based on his tests):

  • LLMs often fail to include the contextual terms at the density and distribution Google expects.
  • So Google “doesn’t understand” the page, devalues it, and sometimes devalues the site overall if you publish lots of it.

This is a brutal failure mode:

  • The content reads “fine” to humans
  • But it’s context-poor for search engines

Modern Day Masters action step

If you use AI to scale content, your bottleneck is not writing speed.
Your bottleneck is:

  • ensuring correct contextual term coverage

If you don’t solve that, scaling AI content can scale damage.

Big Takeaway #6: EEAT won’t lift your rankings, but it can keep you from getting filtered

Kyle’s stance is one of the most useful clarifications:

EEAT elements (privacy policy, terms, etc.):

  • do not boost rankings
  • do not increase traffic
  • do not magically build authority

What they do:

  • help you “check the box” to remain safely in the index
  • reduce risk of suppression or filtering

In other words:

  • EEAT is defensive, not offensive.

Also, he notes:

  • real businesses usually don’t have EEAT problems
  • fake/affiliate/thin sites do

Modern Day Masters action step

Treat EEAT like insurance:

  • not a growth lever
  • a “don’t get kneecapped” lever

Big Takeaway #7: LLM visibility (ChatGPT, etc.) isn’t primarily earned by on-page SEO

This part matters because people are currently selling nonsense.

Kyle’s observation:

  • publishing a page does not reliably get you “inserted” into LLM answers

What does seem to work:

  • being mentioned on strong third-party sites
  • listicles
  • guest posts
  • press releases

The host confirms they’ve seen press releases move the needle for “best” type queries inside ChatGPT.

Translation:

  • LLM visibility looks a lot like classic off-site SEO + PR distribution.

Modern Day Masters action step

If your goal is “rank in AI answers,” your plan is not:

  • publish more blog posts on your own site

Your plan is:

  • get referenced by sources the model is already pulling from

That’s PR + placements.

Big Takeaway #8: Schema is still not a ranking factor (and links still work)

Kyle expected schema might finally become a ranking lever.
He tested it.
Result:

  • no ranking improvement

But another tester bought “spammy” links:

  • rankings improved

So despite all the noise:

  • links still move rankings
  • on-page still matters
  • the fundamentals haven’t changed, the KPIs expanded

Modern Day Masters action step

Don’t confuse “more modern SEO language” with “new rules.”
Most of the industry is renaming the same levers.

Big Takeaway #9: Word count matters, but it’s not an arms race

Kyle’s position:

  • word count absolutely impacts rankings
  • but you’re aiming for the “zone Google is rewarding,” not 10,000 words

Key nuance:

  • if competitors average 3,000 words and you have 500, you’re probably not competing
  • but going huge can backfire if you drift off-topic

He also drops a very practical detail:

  • “Word count” Google sees includes more than what you paste into a doc
  • pages can look short but have lots of indexable text in HTML

Modern Day Masters action step

Use SERP benchmarks:

  • Find the average content length in the top results
  • Build within that band
  • Maintain tight topical consistency

A practical on-page SOP based on this interview

Use this as a repeatable workflow for service pages, money pages, and supporting articles.

Step 1: Choose the money page

This is the page that should:

  • rank
  • convert
  • make money

Example:

  • “Lawn Care in Toronto”
  • “Water Heater Plumber in Copperas Cove”

Step 2: Create 6–20 supporting pages (one purpose only)

Each supporting page:

  • targets a related subtopic
  • links to the money page
  • links to other supporting pages
  • does NOT link out to other service categories

Examples:

  • fixing dead grass patches
  • weed control basics
  • seasonal lawn maintenance checklist

Step 3: Build the internal linking pattern

  • All supporting pages link to the money page
  • Supporting pages interlink (daisy chain, not every page links to every page)

Step 4: Optimize the money page like this

  • Keyword in URL
  • Keyword in title tag
  • Keyword in H1
  • Use keyword once in body (not 50 times)
  • Fill the rest of the page with contextual topic coverage

Step 5: Benchmark “contextual terms”

This is where tools can help, but the idea is simple:

  • look at what top pages mention repeatedly (services, problems, materials, steps, FAQs)
  • ensure you naturally cover the same concepts

If using AI:

  • do not trust it to do this automatically
  • force it to incorporate required terms and sections

Step 6: Defensive EEAT checks

Have the basics so you don’t get filtered:

  • real business information
  • privacy policy / terms as appropriate
  • clear contact and credibility signals

The failure modes you should watch for

If you try to apply this and it “doesn’t work,” it’s usually one of these:

  • You didn’t build enough supporting pages (one blog post isn’t a silo)
  • You linked supporting pages all over the site so they don’t concentrate power
  • Your content is too thin compared to the SERP baseline
  • Your “AI content” reads fine but is context-poor
  • Your money page is buried under too many folders and crawled slowly
  • You’re trying to rank outside your site’s current “tier” without extra reinforcement (content + links)

Closing: what to do next

If you’re building sites or updating client pages right now, the highest ROI move from this entire video is:

  • Pick 1 money page
  • Build a reverse silo around it
  • Use internal links to concentrate topical authority
  • Write for contextual completeness, not keyword repetition

That’s the play.

Darin Pirkey