SEO Legend Charles Floate's A-to-Z Blueprint (Zero to Enterprise)

Edward Sturm

Abstract

An hour and a half laid out as a full sequence: how to launch a site, earn the engine's trust, build links that help rather than sink you, structure the tree, diagnose a decline, and where the money has moved. The guest says he has not put the whole thing in public before, and the order is most of what makes it useful — almost every failure described is this sequence attempted out of order.

• A hundred links on one page sinks a new site; the same hundred across a hundred pages lifts all of them.
• Relevance at the page level is what makes a link good; domain relevance counts toward link-profile consensus instead.
• The traditional algorithm is linear with relevance — the machine does not infer connections that are obvious to a person.
• Intent gets read off the ranking pages, not off your own judgement about what a query means.
• The sitewide quality score is the mean of the page scores, which is why fixing a template lifts everything.
• Information gain is the consensus of what every ranking page says plus something nobody else has.

Everything below follows the conversation in order, from the staged launch through link selection and competitor reverse-engineering, into relevance, semantic triples and the refresh tactic, out through site structure, enterprise versus small-site work and decline diagnosis, and finally into where the money is now — including a demonstration of a fabricated brand reaching first position in three days. For anyone building a personal brand the useful part is the ordering: nothing here is secret, and almost all of it is sequenced wrongly in practice.

Chapter summaries

00:43 - The episode is offered as a full playbook, laid out publicly for the first time

On a brand-new site, step one is to get the search engine to trust and understand the brand as a real entity before anything else, and this matters more in the AI age because the engine's knowledge graph now feeds the language models.

"always preached I was anti stacking and just get Google to tr trust and"

— 01:55

It is offered as the whole sequence at the individual-site level, start to finish, which the guest says he has not laid out publicly before. What follows is unusually ordered for this kind of conversation, and the order is most of the value.

02:18 - Citation and directory profiles earn trust authority rather than link authority

Building citation profiles, social profiles, directory and reference links earns trust authority, not link authority, because those high-authority sites are mostly no-follow.

"Um, in terms of after getting your initial entity recognized, building those citation profile,"

— 02:18

Those high-authority sites are mostly no-follow, so what they pass is trust rather than link value. Conflating the two is why people are disappointed by directory work and why they stop doing it before it has done its job.

02:43 - The launch builds everything in staging and puts it live in one day

The "0 to 100" launch: build the entire site in a staging environment the engine cannot index, then on campaign day one put it all live at once and force-index it through search-console sitemaps plus all the trust-signal links.

"authority. In terms of the link authority from a newer website, I will always"

— 02:43

Everything gets built where the engine cannot see it, and on campaign day the whole site goes live at once, force-indexed through the console alongside all the trust links. The launch is an event rather than a ramp.

03:29 - Fifty to a hundred and fifty trust links over three days spikes the crawl rate

Putting 50–150 trust links live over one to three days spikes the crawl rate — but only if those links are themselves getting indexed.

"signal links live. Because again, if you're putting 50, 100, 150 links live over"

— 03:29

Crawl rate follows links that have been discovered and indexed, and an unindexed link sends no signal at all. Forced-indexing services, internal indexing or press-release indexing are used to get them picked up — which is what the spike actually depends on.

04:06 - Link volume scales with the site and the cadence rather than a fixed number

Link volume scales with site scale and publishing cadence, not a fixed number.

"root authority, going back to the publishing cadence and the actual site scale, depending"

— 04:06

There is no correct number. It scales with the size of the site and how fast it publishes, which means any fixed target somebody quotes is borrowed from a different site.

04:52 - A hundred links on one page tanks a new site and across a hundred lifts it

Building 100 links all to one page on a brand-new site tanks the whole site; spread across 100 different pages, the same links can lift every page and the root authority with them.

"and how many links you're building to each page. If you're building 100 links"

— 04:52

A hundred links pointed at one page on a brand-new site sinks the site. The same hundred spread across a hundred pages lifts each of them and the root along with it. Distribution is the whole difference.

08:01 - What a link points at is the first variable in whether it is any good

The first variable in whether a link is good is what it points AT: the same link can be toxic to a brand-new site and rocket a page on a huge-authority domain to number one.

"So again, it depends on what you're linking to is is what I"

— 08:01

The same link can be toxic to a brand-new site and push a page on a large domain to first position. What the link points at is the first variable, and it changes the sign of the effect rather than its size.

08:39 - A good link needs niche relevance, and only at the page level

A good link has niche relevancy, which only needs to exist at the page level; domain-level relevancy instead counts toward link-profile consensus.

"what is a good link that threshold is going to depend on what you"

— 08:39

Niche relevance only has to exist at the page level. Domain-level relevance is a different thing — it counts toward the consensus in your link profile rather than toward the relevance of the link itself.

09:01 - The engine wants the same trusted domains that power the rest of the niche

The engine builds clustered sets of trusted domains for each niche and wants to see those same trusted referring domains powering your domain — consensus is who links to you, not just what your page says.

"consensus isn't just at an onpage level. It's not just the content that is"

— 09:01

The engine builds clustered sets of trusted domains per niche and wants to see those same domains powering yours. Consensus is about who links to you, not about what your own page claims.

10:29 - A backlink on a page that never indexes is worth nothing

A backlink on a page that never gets indexed is worth nothing, and the crawled-but-not-indexed problem applies to backlinks exactly as it does to your own pages.

"this um this quote. It's from uh your best backlink indexing tools article and"

— 10:29

An unindexed link is invisible to the ranking system, so it transmits no authority whatever it cost. The crawled-not-indexed problem applies to link building as much as to content, and many campaigns build links that never index.

11:22 - Force-indexing every backlink looks like manipulation

But you should NOT force-index every backlink you build, because links all indexing in the same hour looks like manipulation; you want them to index naturally over time.

"uh campaigns that end up creating links that don't ever actually get indexed. Does"

— 11:21

Links all indexing within the same hour is a pattern nothing natural produces. The point is to let them arrive over time, which means the temptation to accelerate is the thing to resist.

12:25 - The real competitors are almost never the ones a founder names

Reverse-engineer your real search competitors, which are almost never the market leaders a founder names; a brand-new business is not competing with the category's giants.

"Yeah. So the the very first thing I tell every client is that"

— 12:25

The names a founder gives are the category leaders, and a new business is not competing with those. The real competitors are whoever occupies the results for the queries that would actually bring buyers.

13:10 - The biggest head term is not the day-one target

Do not chase the biggest head keyword from day one; build the long-tail and variant queries first, which often carry higher buyer intent and ROI than the core term.

"You need to go and look at what are your realistic search competitors. And"

— 13:10

Variant and long-tail queries frequently carry higher intent and better return than the core term, and they are reachable. The head term is a destination rather than a starting point.

13:56 - Authority has to be understood before the money pages can rank

Get the engine to understand your authority first; once pages start immediately indexing for the variant queries, you have unlocked the topical authority needed to rank the big, resource-intensive money pages.

"to, yes, become the topical expert in that niche, but I might not and"

— 13:56

Once pages start indexing immediately for the variants, the authority exists to attempt the big, resource-intensive money pages. Doing it the other way round spends the resource before the authority is there to support it.

15:25 - In topical-authority mode the next page becomes an arithmetic decision

In topical-authority mode every new page indexes and ranks predictably, so you can do ROI maths from existing page data to decide which page to add next.

"Yeah. So once you hit that topical authority mode basically you know that"

— 15:25

Every new page indexes and ranks predictably, so the next page becomes an arithmetic question answered from the data you already have. That predictability is the actual prize of topical authority.

17:20 - The core method is reverse-engineering what already works

Link-building tactics are effectively unlimited if you are creative, but the core method is reverse-engineering what already works using a content-exploration tool, then having an AI surface the outliers.

"there there's an an almost unlimited number of link building tactics if you're creative"

— 17:20

Copying proven winners removes the guesswork and lets return be measured before the spend. A niche's content gets filtered by which pieces gained links fast or overperformed, and the outliers get extracted — which is a different operation from asking a model for ideas.

18:31 - Copying what worked for a competitor is the only measurable link strategy

The number-one link strategy is copying what already worked for competitors, because it is the only approach where the ROI is measurable in advance.

"campaigns. My number one strategy in link building is copying what works because it"

— 18:31

It is the only approach where the return can be estimated before the work is done. Everything else is a hypothesis, and the competitor's profile is a record of hypotheses that were already tested.

20:17 - Time is the small operator's advantage over the big spender

The small operator's advantage is time: inspect each candidate site by hand for faked traffic, faked authority via redirects, and bad-neighbourhood rankings — the due diligence the big spenders skip.

"scaled side. For the much smaller side, you have a time advantage that the"

— 20:17

Inspect each candidate by hand for faked traffic, authority faked through redirects, and rankings in bad neighbourhoods. The large spenders skip this because they cannot afford the hours, which is the small operator's only structural advantage.

22:20 - Reverse-engineer several competitors and target the consensus domains

When reverse-engineering competitors, use several of them, not one, and target the consensus domains that link to most of them first.

"And so now you have a potential competitor, a real competitor, a SER competitor."

— 22:20

A domain linking to all of your top competitors is the niche's trusted source and therefore the highest-value target. If all five top competitors carry a link from one niche blog, that is the link to get; the same overlap is read across larger competitors too.

23:59 - Trace which links power the root and which power the page

Identify which links power the root authority versus individual pages, and trace the juice — sometimes a page's real power arrives through an internal link from a supporting page, so you must look at the whole site's link profile, not just the ranking URL.

"about which links are actually powering the root authority and which links are actually"

— 23:59

Sometimes a page's real power arrives through an internal link from a supporting page, which means reading only the ranking URL's profile gives the wrong answer. Trace the juice across the whole site.

24:52 - The tools carry heavy spam and their best filters hide the good links

Backlink tools carry heavy spam and their "best" filters hide genuinely good links; a page is sometimes powered by a single un-repeatable old news link, so find alternative strong domains and use your competitors' profiles as a negation set.

"can access the trained AIS, agent A is decent, but it's like $99 a"

— 24:52

The best filters hide genuinely good links, and a page is sometimes powered by a single old news link nobody can repeat. Use the competitor profiles as a negation set and go looking for the alternatives.

28:18 - At page level the traditional algorithm is linear with relevance

At the page level the traditional ranking algorithm is LINEAR with relevancy: a topic is either directly associated with your site or it is not — the machine does not infer human-obvious connections.

"Yeah. No, I So, it it it does again, it's always depend, isn't"

— 28:18

A topic is either directly associated with the site or it is not — the machine does not infer connections that are obvious to a person. Treating relevance as linear is treating it the way the system actually treats it.

29:08 - A large block of off-topic content destroys the site's focus

Adding a large block of off-topic content and interlinking it destroys a site's core focus and its rankings — the composting case: a weight-loss site added roughly 120 gardening pages (a third of the site), cross-linked them, and traffic fell off a cliff.

"example, I had a client a few years ago that was in the weight"

— 29:08

A weight-loss site added roughly a hundred and twenty gardening pages, about a third of the site, and cross-linked them. The traffic fell off a cliff. The lesson is not that the pages were bad but that the focus was.

30:57 - Relevance gets treated linearly, the way a computer treats it

Treat relevancy linearly, as a computer would; AI is better at connections but still runs on token correlation, so it also does not treat unrelated tokens as related.

"it lost its core topical authority and its core site focus score and it"

— 30:57

Models are better at connections and still run on token correlation, so unrelated tokens do not become related there either. The linear treatment holds across both systems, which is why it is worth adopting as a habit.

31:52 - Ranking in search and ranking inside a model are separate goals

Ranking in the traditional search results and ranking inside the AI are two fairly separated goals right now, needing different structure and format.

"That's a big explanation. I know. Going on to the secondary part of optimizing"

— 31:52

They need different structure and different formatting, and treating them as one target produces something that underperforms at both. Right now they are separate goals with separate work.

32:44 - Different models have to be optimised differently

Different AI models must be optimised differently: one assistant is far more sceptical of product recommendations (by design in its system prompt), AI overviews are the easiest to feed with fake information, and another sits in the middle.

"difference of the way models actually work. Eg Claude is way more skeptical of"

— 32:44

One assistant is far more sceptical about product recommendations by design, the overviews are the easiest surface to feed, and a third sits between them. The differences are in the system prompts as much as the models.

34:10 - A semantic triple is how a claim becomes machine-verifiable

Semantic triples — a subject-object-predicate structure repeated until the machine treats a claim as fact — are how you make a claim machine-verifiable, and machines prefer the active order ("X founded Y") over the passive.

"Okay. Sorry. Yeah, I I always kind of just jump straight into everyone's"

— 34:10

Subject, predicate, object, repeated until the claim is treated as fact — and the active order outperforms the passive. The form is doing the work, which is why phrasing matters more here than in ordinary writing. Layout, elements and what sits above the fold matter alongside it, because they produce the engagement signals the algorithm reads.

36:23 - The sites ranking for everything are generic authorities, not topical ones

The huge sites that rank for everything are not topical authorities, they are GENERIC authorities — a different, higher tier a normal site cannot simply imitate.

"So a Red Bull, Forbes, those kind of things, they're not topical authorities"

— 36:23

They are generic authorities — a different and higher tier that a normal site cannot imitate by copying their behaviour. Mistaking one for the other is how sites end up publishing in every direction at once.

36:56 - A famous collapse was mostly an intent change rather than a penalty

A famous multi-topic traffic collapse was mostly an intent CHANGE, not a pure penalty: the results for those queries shifted from blogs to commercial and tool pages over about a year.

"authority. I also so HubSpot is its own unique case because I think well"

— 36:56

The results for those queries moved from blogs to commercial and tool pages over about a year. It looked like a penalty because the effect was the same, and the cause was demand rather than judgement.

38:47 - The tactic being pushed this year is to stop publishing and start refreshing

The number-one tactic being pushed this year is to stop publishing and start refreshing — update metadata, a couple of sentences, an image, then force-reindex the page.

"rather than refreshing content. So, one of the number one tactics that I have"

— 38:47

Update the metadata, change a couple of sentences, add an image, then force a reindex. It is the most-pushed tactic of the year and the reason is simply that it works better than publishing more.

39:15 - A refreshed page can jump where a new page cannot

A refreshed page can jump dramatically where a new page cannot: one services page went from roughly 22 visits a month to about 5,000 uniques after a two-paragraph update, two added list items, and a new date.

"charles.com/linkubuilding services, obviously hyphen link hyphen buildings hyphen services, you're going to see that"

— 39:15

One services page went from around twenty-two visits a month to roughly five thousand unique visitors after two paragraphs, two list items and a new date. The same content as a new page would not have done that.

41:04 - A root welded to one niche cannot be bridged into a second

On a site whose homepage and root are welded to one niche, you cannot bridge into a second topic, because the root's weighting pushes every subfolder and subpage toward that niche.

"So on his website, he couldn't basically is is what we came to"

— 41:04

The root's weighting pushes every subfolder and subpage back toward the original niche, so there is no quiet route into a second topic from a site that was built for one. The constraint is structural.

41:41 - Adding categories degrades the focus score for the profitable core

Adding categories degrades the site-focus score for your profitable core, so a pivot carries financial risk; the three options are to set the root up generically from the start, to invest heavily in category content and links to shift it, or a subdomain (not recommended).

"you so you can put in the work but there's always going to"

— 41:41

The pivot carries financial risk because the focus score for the profitable core degrades as categories are added. The three options are a generic root from the start, heavy investment in category content and links, or a subdomain — and the subdomain is not recommended.

45:00 - A new niche gets tested through the closest real bridging entity

To test a new niche, bridge with the closest possible relation: find a real bridging entity the knowledge graph already associates with BOTH topics, publish that first, and only expand outward if it ranks.

"Yeah. Okay. So if you want to reinvest then essentially re like I"

— 45:00

Using an entity the engine already links to both sides borrows an association that would otherwise have to be built from nothing. A weight-loss gardening event connected in the knowledge graph to both topics becomes the first bridge page — and if the bridge page does not rank, there is no pivot without heavy root authority.

48:10 - Structure comes from live reverse-engineering, and the homepage targets the brand

Structure a site from LIVE reverse-engineering of what currently ranks, not a stale brief, and target the homepage for a BRAND keyword rather than a niche keyword.

"So the first thing in terms of that specific case is when he first"

— 48:10

Structure comes from what currently ranks rather than from a brief written months ago, and the homepage targets the brand rather than a niche keyword — because the homepage is where people expect to find out who you are.

49:55 - The core informational pages sit directly under the root

Put the core informational pages (home, about, contact) directly under the root so they reinforce the root, not one directory below where they reinforce a "company-info" folder you do not want to rank.

"And then you're putting are you putting these in the in the footer?"

— 49:55

Home, about and contact go directly under the root so that they reinforce it, rather than one level down where they reinforce a company-information folder nobody wants to rank. It costs nothing and it is frequently wrong.

52:11 - Intent is determined from the ranking pages rather than from judgement

Build a topical map from the core keyword, and determine intent from the RANKING PAGES, not from your own judgement: check whether the top three results are the same page-type, and if not, take the consensus with position one weighted highest.

"So I I topical map from the core, right? So uh from SEO"

— 52:11

Check whether the top three results are the same page type, and if they are not, take the consensus with first position weighted heaviest. Your own judgement about what a query means is the least reliable input available.

53:27 - Each query node gets triaged by the competition actually ranking it

Triage each query node by the competition actually ranking it, and go after the winnable nodes first for faster ROI in clicks, traffic, and revenue.

"there, I'm seeing which uh what is the competition level of the individual SERs."

— 53:27

Go after the winnable nodes first, judged by who is actually ranking them rather than by volume. The return arrives faster in clicks, traffic and revenue, which is what funds the harder nodes later. Pixel-tracking from launch and retargeting by paid impressions captures the rest, as long as earnings per click clear the cost.

55:55 - The topical map is built by feeding competitors' keywords back through the tool

A topical map is built by taking the core keyword, pulling the top competitors' ranking keywords, feeding those back through a research tool to find more overlapping keywords, and repeating per page — which scales exponentially.

"It's literally mapping the individual keyword. So, as an example, um, going back"

— 55:55

Take the core keyword, pull the competitors' ranking terms, feed those back through the research tool to find the overlaps, and repeat per page. The process scales exponentially, which is both its strength and its hazard.

57:16 - The bigger the site, the more red tape and the less control

The bigger the site, the more red tape and the less control: enterprise recommendations are often only about 20% implemented, slowly, by a dev team that never confirms the work is done.

"So, the number one thing with the the bigger site you get, the"

— 57:16

Enterprise recommendations get implemented around a fifth of the time, slowly, by a development team that never confirms the work happened. The constraint stops being knowledge and becomes organisational.

58:13 - Enterprise gains from every other department's work

Enterprise sites gain from every other department's work — PR links, hundreds of thousands of brand searches a month — so results are roughly 10x once implemented; enterprise SEO is easy to get results but hard to get them shipped.

"but also you get an advantage because you have more ability, more pages, more"

— 58:13

Public-relations links, hundreds of thousands of brand searches a month — the site inherits the rest of the company's work, so results are roughly tenfold once anything ships. Getting results is easy; getting them shipped is the problem.

58:57 - A small site needs far more manual work per page

Smaller sites need far more manual work per page — inspecting three competitors on desktop and mobile and through a VPN to check for geo changes — but this is possible per page and maximises the sitewide quality score.

"for the smaller sites, it's it's just a lot more actual work. So,"

— 58:57

Inspecting three competitors on desktop, on mobile and through a connection in another region to check for geographic differences — per page. It is only possible at small scale, and it is what maximises the sitewide quality score.

1:01:24 - Crawled-not-indexed is a fix on a small site and a foundation problem at scale

Crawled-not-indexed is a force-index fix on a small site but a technical-foundations problem at enterprise scale, where it is tens of thousands of pages at once.

"Again it depends on the size and scale of the website. Um if"

— 1:01:24

On a small site it is a force-index fix. At enterprise scale it is tens of thousands of pages at once and a technical-foundations problem. The same message means two different things depending on how many pages carry it.

1:03:01 - A single stuck page has a three-step sequence

For a single new page stuck crawled-not-indexed, the sequence is: ping it in search console, do a render-check to see what the crawler actually sees, then rule out a hosting or CDN issue.

"like first off ping it in search console in civ actually will reindex itself"

— 1:03:01

Ping it in the console, run a render check to see what the crawler actually receives, then rule out hosting or delivery-network interference. Three steps, in that order, and most cases resolve inside them.

1:04:32 - Over-dependence is solved by overlap rather than by diversification

For a business over-dependent on SEO, think content OVERLAP, not diversification: create one piece and reformat it per channel, because a written format does not transcend to video and vice versa.

"I tend to think of more so I again I will always go"

— 1:04:32

Create one piece and reformat it per channel, because a written format does not transcend to video and video does not transcend to writing. Overlap is cheaper than diversification and it produces the same protection.

1:06:31 - The goal is turning a one-time visit into a long-term asset

From the SEO angle, the goal is to turn one-time search visits into long-term assets — pixel data, email lists, communities, accounts — and the mechanism for that usually has nothing to do with SEO.

"from an SEO angle, I will always say, how can you turn your uh"

— 1:06:31

Pixel data, email lists, communities, accounts — the visit becomes an asset that does not depend on the next ranking. The mechanism for that conversion usually has nothing to do with search at all.

1:07:14 - For a site with traffic, conversion work returns more than more search work

For a business that already has traffic, the highest-return use of time is conversion-rate optimisation, not more SEO: moving conversion from 1% to 3% is a 300% gain you cannot get organically in a day.

"exactly. So, this is what I've been I've I've always been a massive"

— 1:07:14

Moving conversion from one per cent to three is a threefold gain available immediately, and no amount of organic work produces that in a day. For a site that already has traffic it is the highest-return hour available.

1:08:03 - Conversion does not have to mean a sale

Conversion does not mean an immediate sale: an email pop-up, lead magnet, account signup, free download, follow, or subscribe all convert the visitor into a long-term recurring asset — optimise for lifetime value, not one-time value, and be a marketer, not just an SEO.

"pages and co doesn't necessarily just mean that you're converting them right then and"

— 1:08:03

An email capture, a lead magnet, an account, a download, a follow, a subscription — each turns a visitor into a recurring asset. Optimising for lifetime value rather than one-time value is what separates a marketer from a technician.

1:08:47 - The only work worth doing is the work being paid and measured for

Only do the work you are actually paid and measured for: you can double a client's leads by changing button colours and look like a wizard, but you must be able to attribute recurring value in the monthly report.

"marketer rather than just explicitly an SEO. But I will also say, what are"

— 1:08:47

You can double a client's leads by changing a button colour and look like a magician, and if it cannot be attributed in the monthly report it does not count as work delivered. Be paid for what is measured.

1:09:59 - Clients get sold the outcome rather than the tactics

Sell clients the OUTCOME, not the tactics — not backlinks or topical maps but the money and results — because business owners do not care about or understand the jargon, exactly as a good landing page sells the outcome.

"it's important to to to think like a business owner. That's why I really"

— 1:09:58

Business owners do not care about backlinks or topical maps and mostly do not understand them. Sell the money and the results, exactly the way a good landing page sells the outcome rather than the specification.

1:11:08 - The first audit question is which areas actually matter

When a site is losing organic traffic, the first audit question is which areas you care about and which you do not — are you losing where it matters or where it does not.

"So the first thing I will say to them is okay so so actually"

— 1:11:08

Before any diagnosis: which areas matter and which do not. Losing traffic where it does not matter is not an emergency, and a great deal of audit effort gets spent on the parts of a site nobody was monetising.

1:12:17 - Then the decline gets classified as generic, categorical or technical

Then diagnose whether the decline is generic or a symptom of a specific category or a specific technical issue.

"there, why are they losing a lot of traffic? Is it a generic reason"

— 1:12:17

Generic, or a symptom of one specific category, or a specific technical fault. Three buckets, and the investigation that follows is different for each — which is why the classification comes before the investigation.

1:12:59 - On a large site a decline is a cascade rather than a cause

On large sites a decline is never one problem but a cascade of stacked issues, each degrading a different area, and unless you fix all of them the site keeps bleeding traffic.

"but again, when you start dealing with massive clients, it's not normally one problem."

— 1:12:59

Individually minor failures compound, and together they cause a calamity that no single fix resolves. Clients ask for the issue when there are many stacking issues cascading across areas of the site.

1:13:42 - The commonest cause of decline is intent change

The single most common cause of decline is intent mismatch and intent change — the query's intent shifted since the site was built three years ago, and often the original agency never researched intent at all.

"What are the most common issues that that you see? The most common"

— 1:13:42

The query's intent shifted since the site was built three years ago, and frequently the original agency never researched intent to begin with. It is the most common cause and the least often checked.

1:14:20 - Declines also come from development damage and organisational changes

Beyond intent, declines come from dev-caused damage (crawler-blocking JavaScript, header problems, unoptimised images, a 30MB video loaded on the page) and org-level changes (renaming a service that carried branded search traffic loses it, and the historical dip can hurt you for up to 18 months).

"problems generally you you run into devs just being you know devs. Um they"

— 1:14:20

Crawler-blocking scripts, header problems, unoptimised images, a thirty-megabyte video on the page — and at the organisational level, renaming a service that carried branded traffic. That historical dip can hurt for as long as eighteen months.

1:16:26 - The quality score is assigned at both page and site level

The engine assigns a binary quality score at BOTH the page level and the sitewide level, and both matter.

"Yeah. So Google essentially assigns a a site quality score to your page"

— 1:16:26

Quality is scored per page and aggregated for the site, so the two layers influence different outcomes. They are described as separate binary metrics that both mean something, rather than one number read at two altitudes.

1:17:00 - The sitewide score is the mean, which is why templates matter

Model each page as a 0–100 quality score; the sitewide score is the mean of all page scores, so improving a template across many pages lifts the whole site a few points; page score drives individual rankings and sitewide score drives overall visibility.

"effectively put it as 0 to 100. Each page on your entire website has"

— 1:17:00

An averaged score means broad template improvements move the whole site, while an individual page still needs its own score raised to rank. A thin category page scoring around twenty drags the mean down; a template updated across ten thousand pages moves the sitewide score two or three points. The numbers are arbitrary placeholders for a leaked concept.

1:18:23 - There is no single lever, and the closest thing is intent matching

There is no single 20% that lifts a page's quality score, but the closest thing is intent matching — at a granular, consensus level: missing entities, unoptimised meta titles, missing schema, images, video, or a worse load time.

"It's it's going to depend on the page again, right? >> There's no"

— 1:18:23

Quality score is the sum of many granular matches to what the ranking consensus already has, so no single factor dominates. The failing page is the one missing the entities, the schema, the media and the load speed its competitors carry — intent and consensus matching at a granular level, not a lever.

1:19:24 - The opportunity most operators are sleeping on pays two and a half times more

The biggest opportunity most operators are sleeping on is AI-visibility services: the average client pays roughly $8,000 a month for it versus roughly $3,200 for SEO — about 2.5x — for work that is often easier.

"I like right now it's I I'm gonna hate myself for saying it, but"

— 1:19:24

The average client pays roughly eight thousand a month for visibility work against roughly three thousand two hundred for search work — about two and a half times — for work that is frequently easier. The gap is the opportunity.

1:20:25 - The newer work is easier because it skips most of the technical layer

AI-optimisation is easier because it skips most of the technical work (no indexing problems), and reduces to three stages — entity, consensus, and corroboration — plus retrieval, system-prompt, and model-architecture variables.

"Because you don't have to do all the technical work that you have"

— 1:20:25

Without the crawl-and-index technical layer, the job is largely establishing and corroborating an entity. Entity, consensus and corroboration are named as the core, with model architecture affecting retrieval and output — and a version change shifts the underlying behaviour.

1:21:52 - Getting into the cited pages is the single most important move right now

For AI overviews, getting your brand into the pages cited across the prompts you test is the single most important move right now; for the chat assistants themselves it matters far less long-term, because their retrieval and sourcing keep changing.

"probably uh the most important thing right now for chat GPT Gemini Claude Grock"

— 1:21:52

Getting the brand into the pages cited across the prompts you test is the highest-value move available right now. For the chat assistants themselves it matters less over time, because their retrieval and sourcing keep changing.

1:22:19 - Feeding a cache is a short-term hack that does not survive re-sourcing

Feeding an assistant's cache is a short-term growth hack: if you did not get into the training data, you will not keep the visibility once the cache re-sources.

"signing a bunch these pages right now for the next couple of weeks if"

— 1:22:19

If you did not make it into the training data, the visibility disappears when the cache re-sources. And whatever you do to get into training data lands you in the model after the next one, because the next one is already trained.

1:23:17 - The second-biggest training format is the PDF

After HTML/XML from common web crawls, the second-biggest training file format is PDFs, so where models scrape PDFs is a lever: a publicly-available research-paper repository is pulled by every model, and one large model trained on torrented material.

"Okay. I I could give you one example, right? Um the number one"

— 1:23:17

After crawled HTML, documents are the second-largest training format, which makes wherever models scrape them a lever. A public research-paper repository is pulled by every model, and at least one large model trained on torrented material.

1:25:49 - The overview has no arbitration step at all

AI overviews have no arbitration — no is-this-real check — so you can push a fabricated consensus into search, get it picked up by the traditional algorithm, and have the overview retrieve it as fact.

"Yeah. So, I noted that AEO views don't really have any arbitration, right?"

— 1:25:49

There is no is-this-real check. A fabricated consensus can be pushed into search, picked up by the traditional algorithm, and retrieved by the overview as fact. The missing step is the whole vulnerability.

1:26:42 - The demonstration reached the first position in about three days

The demonstration: an exact-match-domain "SEO awards" microsite, entirely AI-generated with a code assistant, plus a wiki-style page, a forum thread, and a couple of expert posts to reinforce the entity, hit number one in about three days before being slapped — and the overview called it the most prestigious awards ceremony in the field.

"output. I've realized that the AO views like certain formats plus the AI, sorry,"

— 1:26:42

An exact-match domain plus a coordinated entity build exploits the traditional algorithm's domain bias and the overview's lack of arbitration at the same time. Judges invented in a sentence, the whole site machine-built, first position in about three days — with the domain doing much of the work and the term carrying only about three hundred searches.

1:28:36 - No links were built to the fabricated site at all

No links were built to the fabricated site — an indexing tool got the pages indexed, they ranked on the EMD, and collaborators mentioned the brand with NO link at all.

"I did build links to them. I bu I used the indexal tool"

— 1:28:36

For an entity play, indexed pages plus brand mentions can rank with no backlinks at all. An indexing tool got the pages in, and friends wrote about the brand without linking to it.

1:29:15 - The mentions were left deliberately unlinked

The mentions were deliberately unlinked because an artificial link risks penalising the linking sites, and the domain did not need a link — an entity needs the MENTION, the consensus of mentions, and corroboration from multiple trusted domains to be pulled into the overview.

"A because I knew that Google's engineers would say that this is an"

— 1:29:15

An artificial link risks penalising the site that carries it, and the domain did not need one. An entity needs the mention, the consensus of mentions, and corroboration from several trusted domains — the link is not part of the requirement.

1:29:58 - For an owner doing their own work, parasite lead generation is the opportunity

For a business owner doing their own SEO, the biggest opportunity is parasite-SEO lead generation: mass-produce parasite pages for long-tail variants with a call-to-action and phone number, and satellite micro-sites overlapping your existing keywords, until you occupy half the results.

"Parasite SEF lead generation 100%. You can just mass make parasite pages especially"

— 1:29:58

Mass-produce pages for long-tail variants with a call to action and a phone number, plus satellite micro-sites overlapping the existing terms, until half the results page belongs to you. It is the highest-leverage move available to somebody doing their own work.

1:30:55 - Ranking factors are niched down and root authority is still the first metric

Ranking factors are niched down (different signals matter for different queries), but root authority is still the single number-one metric: identical-quality content on a huge-authority domain gets thousands-of-percentile more visibility than on your own small site.

"It's it it it again it does depend on the niche because Google"

— 1:30:55

Different signals matter for different queries, and root authority remains the single first metric. Identical content on a high-authority domain gets a vastly larger share of visibility than the same thing on a small site.

1:31:55 - Ranking factors also depend on whether the page already ranks

Ranking factors also depend on whether you are already ranking: once ranked, you reinforce position with user-engagement signals that make you hard to overtake, but you can also over-optimise a ranking page and cause it to stop ranking entirely.

"copy piece. It also depends on if you're ranking or not. The ranking factors"

— 1:31:55

The factors that win a position differ from those that hold it, so the same page is judged twice by different criteria. User-engagement signals reinforce a position once held, and over-optimisation is seen to kill rankings routinely.

1:32:50 - Over-optimisation usually hides in the source rather than the page

Over-optimisation is often hidden in the source code, not the visible page: keyword stuffing in schema, meta tags, and hidden elements — a page showing 54 exact-match anchors may carry 214 in the source — and the fix is to cut the mentions in schema and table-of-contents down to what the page actually needs.

"They don't look at the source code, right? Um, and I've seen so many"

— 1:32:50

Keyword stuffing in schema, meta tags and hidden elements — a page showing fifty-four exact-match anchors may carry over two hundred in the source. The fix is cutting the schema and contents-list mentions back to what the page actually needs.

1:34:16 - Updates now mostly aim at feeding the overviews

Algorithm updates are not as powerful as they once were, and current updates mostly aim at one thing: feeding the AI overviews by reweighting the algorithm from ranking pages toward retrieving pages, and toward information gain.

"So I I I want to start by saying that actually Google updates"

— 1:34:16

Updates are less powerful than they were, and the current ones mostly reweight the algorithm away from ranking pages and toward retrieving them, and toward information gain. The target of the tuning has changed.

1:35:53 - Information gain is the consensus plus what nobody else has

Information gain is the consensus of what every ranking page already says about a topic PLUS new information no other page has — the overview looks for both the consensus passages and the unique ones.

"Yeah. So information gain is literally um consensus of information. So it's all"

— 1:35:53

It is the consensus of what every ranking page already says, plus something no other page has. The overview looks for both — the passages that agree and the passages that are unique — which is why either alone underperforms.

1:36:57 - Reconcile reverse-engineering with novelty at about ninety-five to five

Reconcile reverse-engineering with novelty at roughly 80/20 — really closer to 95/5: match the consensus and beat competitors' quality score first, then add the unique passages, data points, and insights the overview can surface.

"8020 is is doing a lot of hard work because it's really like 955."

— 1:36:56

Consensus is the entry ticket and novelty is the differentiator, so the bulk of the effort goes to matching before anything gets added. An eighty-twenty split gets corrected on the spot to nearer ninety-five to five — even eighty-twenty overstates the novel share.

Personal Branding Lessons

A full playbook told in sequence. The moves below are the ones that transfer out of the agency context.

Spread the links across many pages rather than concentrating them

A hundred links pointed at a single page on a brand-new site sinks the whole site. The same hundred spread across a hundred pages lifts each of them and the root along with them. Distribution changes the sign of the effect rather than its size, which is why this is the first decision rather than a refinement. 04:52

Build the long-tail variants before the head term

Variant and long-tail queries frequently carry higher intent and better return than the core term, and they are actually reachable. Once pages start indexing immediately for those, the authority exists to attempt the resource-intensive money pages. The head term is a destination, not a starting point. 13:10

Copy what already worked for a competitor

It is the only link approach where the return can be estimated before the work is done. Everything else is a hypothesis; a competitor's profile is a record of hypotheses that were already tested, at somebody else's expense. Use several competitors and go after the domains that link to most of them. 18:31

Determine intent from the ranking pages rather than from judgement

Check whether the top three results are the same page type, and if they are not, take the consensus with first position weighted heaviest. Your own reading of what a query means is the least reliable input available, and it is the one most plans are built on. 52:11

Put the core informational pages directly under the root

Home, about and contact belong directly under the root, where they reinforce it, rather than one level down where they reinforce a company-information folder nobody wants to rank. It costs nothing to get right and it is frequently wrong. 49:55

Reformat one piece per channel rather than diversifying

For a business over-dependent on one channel, the answer is overlap rather than diversification: create one piece and reformat it per channel, because a written format does not transcend to video and video does not transcend to writing. Overlap is cheaper and produces the same protection. 1:04:32

Spend the next hour on conversion rather than on more search work

Moving conversion from one per cent to three is a threefold gain available immediately, and no amount of organic work produces that in a day. Conversion does not have to mean a sale — an email capture, an account, a download or a subscription each turns a visitor into a recurring asset. 1:07:14

Sell the outcome rather than the tactics

Business owners do not care about backlinks or topical maps and mostly do not understand them. Sell the money and the results, exactly the way a good landing page sells the outcome rather than the specification. And only take on work whose value can be attributed in the monthly report. 1:09:59

Questions

Each answer ends at the moment in the recording where it is given.

How do I build authority and get links for a brand-new website from scratch?

Start by making the search engine trust and understand the brand as a real entity — citation, social, directory and reference profiles — before chasing ranking links, because until the machine can identify you, nothing else has anything to attach to. Then build the whole site in staging, push it live at once, and let trusted links index naturally over thirty to sixty days. Getting the basic identity consistent everywhere is the same first step behind finding and fixing your authentic brand. 01:56

What actually makes a backlink good or worthless?

Three things, in order: what the link points at, whether it is relevant, and whether it will ever index. The same link can be toxic to a new site and rocket a page on a huge-authority domain, because the receiving domain's trust sets the value. Relevance only needs to exist at the page level. And a link on a page the engine never indexes transmits nothing — effort only signals if the machine can see it, the idea behind costly signalling. 08:01

Who are my real competitors, and how do I reverse-engineer their links?

Not the category's giants — a brand-new business does not compete with them. Your real competitors are whoever actually ranks for your realistic queries. Use several of them, find the consensus domains that link to most of them, and target those first, because a domain trusted by the whole field is the niche's trusted source. The unoccupied ground is usually what everyone overlooks, the idea at the centre of benchmarking against the gaps rather than the strengths. 12:25

Why did adding new topics to my site tank my traffic?

Because the algorithm treats relevance as linear — a topic is either directly associated with your site or it is not, with none of the human-obvious inference in between. Adding a large block of unrelated content and cross-linking it destroys the site-focus score that powered every page, so the machine can no longer tell what the site is about. One documented case fell from roughly forty thousand a month to four or five thousand after diluting its core with a third of the site in off-topic pages. 28:18

Should I keep publishing new content or refresh what I already have?

Refresh first. The tactic pushed hardest this year is to stop overproducing and update declining pages instead — new metadata, a couple of sentences, an image, then force a reindex — because a refresh reactivates the historical trust a page already holds. One services page jumped from a couple of dozen visits a month to thousands after a two-paragraph update and a new date, something a brand-new page could not do. Build the cadence around updating, not just adding, the way a real content-calendar system plans. 38:47

How should I structure a site and its internal links?

Target the homepage for a brand keyword, not a niche keyword, so the site stays flexible as ranking page-types change over the years. Reverse-engineer what currently ranks rather than working from a stale brief. Put the core informational pages — home, about, contact — directly under the root, because every page reinforces whatever sits directly above it, and nesting them under a folder wastes their signal on something you do not want to rank. 48:10

My page says "crawled — not indexed" — what do I do?

On a small site it is usually a force-index nudge; on a large one it is a technical-foundations problem affecting tens of thousands of pages at once. For a single stuck page, the sequence is fixed: ping it in search console so it re-crawls, run a render-check to see what the crawler actually renders versus what you see, then rule out a hosting or CDN issue such as a firewall blocking images. A page can look complete to you and arrive broken to the machine. 1:01:24

I have search traffic but few sales — what should I fix first?

Conversion, not more traffic. Moving conversion from one per cent to three is a threefold gain you cannot get organically in a day, and conversion does not mean an instant sale — an email capture, a lead magnet, an account, a follow all count. The aim is to turn a one-time visit into a recurring relationship worth far more over its lifetime, the compounding logic behind network effects in a personal brand. 1:07:14

How do I get my brand cited in AI answers and overviews?

For the overviews, get your brand into the pages already cited across the prompts you test — that is the single most important move right now. For the chat assistants it matters less long-term, because their retrieval and sourcing keep changing, and a cached citation decays in a couple of months unless you get into the training data. An entity is pulled into an answer by corroborated mentions across trusted domains, not by a single link. 1:19:24

Sources

SEO Legend Charles Floate's A-to-Z Blueprint (Zero to Enterprise)

The sections above follow the episode's own order. Each timestamp is the point in the recording where the idea is discussed, so the post reads as a map of the conversation rather than a rearrangement of it.

Quotations come from the episode transcript with spoken filler ("okay", "right", "you know", "um", "actually", "kind of", "sort of") and false starts removed. No words are added: square brackets mark anything inserted for sense, and an ellipsis marks any cut. British spellings are restored, and speech-to-text errors were repaired against context only where the intended word is unambiguous. Where a line appears both in a cold-open teaser and in its place in the conversation, the timestamp cites the second — the point at which it is actually said.

Quotes are stamped with the moment they are said rather than with a speaker's name.