Media Mentions: How to Get Featured and Quoted
Sign in Subscribe
Playlist

SaaS SEO in 2026: Link Building, Topical Authority & the GEO Hype

Slone
Slone
— 26 min read
Takeaway

Authority in search is not a score on your domain or a pile of content on your own pages; it is a topic-by-topic property other people confer through the clicks, links and searches they aim at you. Build and link pages that already earn real traffic, test a small batch before committing, spend authority only in the topic where you earned it, and read your own analytics before believing anyone who says search is dead.

Leave a Comment
Related Articles

Edward Sturm

Abstract

An hour and a half taking apart the conventions the field runs on. The authority score is one made-up number describing something that varies by topic. The threshold rule for link building is a convention nothing enforces. Thin content is not a thing. Site-wide relevance is not a ranking factor. What is left after the demolition is a small, testable method: build where nobody is competing, route the authority you have, and stop spending authority you have not earned.

  • A low-authority link that sends real relevant traffic beats a high-authority one that sends nobody.
  • Authority decays roughly eighty-five per cent per jump, so it is nearly gone within three links.
  • Internal links are wiring: they route authority you already have and cannot create any.
  • Each question gets its own page, because the page name sets the relevancy score that decides what gets spent.
  • The assistants do not crawl for the best answer — they run a query and synthesise what already ranks.
  • Optimisation is subtraction: the fewest things that must run very well, not everything turned on.

Everything below follows the conversation in order, from the authority score through link building through networks, into relevance, entities and the generative hype, out through the question-page method, rotation and republishing, and finally into attribution, brand handcuffs and the slug. For anyone building a personal brand the useful part is the demolition: most of what stops people starting is a rule nobody can point to the source of.

Chapter summaries

00:44 - The domain-authority score is one made-up number where real authority is an array

Give up on the idea that a domain-authority score matters; it is a single made-up number for a whole site, whereas real authority is an array with a separate value per topic that sites inherit.

"I think that the the first thing is to give up on this idea that"

— 00:44

One made-up number for an entire site cannot describe something that varies by topic. Real authority is an array with a separate value per subject, inherited down from the site — which means a single score is describing a shape it cannot represent. The metric is the problem.

The rule that you can only get links from sites above a certain authority score is a myth; a giant site earns its high score from countless small ones, and there is no negative score to avoid.

"you can't get links from less than da40. And I it it just really undermines the"

— 01:18

A giant site earned its score from countless small ones, and there is no negative score to protect against. The threshold rule is a convention that spread because it sounds cautious, not because anything enforces it.

01:45 - Authority is lumpy and flows unevenly across a site

Authority is lumpy and flows unevenly across a site; a brand-new page can start at zero, and an off-topic page on an on-topic site may simply not rank.

"rest their DA in their homepage or in their blog page."

— 01:45

It concentrates in some places and not others, so a brand-new page starts at nothing and an off-topic page on a topical site may simply not rank. Authority is not a property of the domain you can draw down anywhere.

Do link building through your real-life network — friends, chamber of commerce, a weekly tech meetup, clients, and the companies you integrate with — and start one if none exists.

"even start link building. I think go back to your network, go back to your friends,"

— 02:24

Friends, the chamber of commerce, a weekly meetup, clients, the companies you integrate with. If no such group exists locally, start one — which is a slower answer than buying links and the only one that compounds.

The spam lists offering to sell you backlinks are free keyword research; opening them sometimes surfaces a URL with near-zero difficulty and real monthly traffic you can outrank, so the clicks act like a backlink and you never buy the link.

"one of the things I love doing when people email those lists"

— 03:21

Opening them sometimes surfaces a URL with almost no difficulty and real monthly traffic that you can outrank directly. The clicks then do what the link would have done, and nothing was bought.

04:15 - The domain gets evaluated rather than its score

Evaluate the actual domain, not its authority score; the high score in a tool may be built from estimated traffic and rank on a site that is waning, not growing.

"They get those lists from people who own link farms or private blog networks"

— 04:15

A high score in a tool can be assembled from estimated traffic and rank on a site that is declining rather than growing. Looking at the actual domain takes a minute and the number takes none, which is why the number wins.

04:47 - A friend's link only helps once that page itself ranks

A link from a business friend's or client's page only helps if you first make that page rank — do its keyword research and get it links — and then link contextually so it flows authority and context back.

"if you're getting a link from a friend's website"

— 04:47

Do the keyword research for that page, get it links, and only then place a contextual link back. A link from a page nobody reaches transfers nothing, however friendly the arrangement was.

A small satellite site can rank widely on very few, ordinary backlinks.

"I shared a a an image last night"

— 05:16

Very few links, none of them remarkable, and a small satellite ranks across a range of terms. It is the counterexample that undermines most of the volume advice in the field.

07:15 - Pages answering the questions nobody targets rank easily

Building pages that answer People-Also-Ask questions ranks easily because almost nobody targets them; you satisfy intent, avoid bounce-back, earn good clicks, and funnel that authority into more important pages.

"So you were getting authority by putting up people also ask questions."

— 07:15

Almost nobody targets these questions, so they rank easily, satisfy the intent, avoid the bounce back to the results, and earn clean clicks. That authority then routes into the pages that matter. The cheap ground feeds the expensive.

An external link carries more context and weight than an internal one, purely because it is external — so a tiny parent site can still push a satellite up.

"the parent site barely has 500 clicks a month."

— 08:22

An external link carries more context and more weight purely by being external, which is why a very small parent site can still lift a satellite. The relationship between the two pages is doing work the sizes do not predict.

Understanding the method and refusing the self-imposed handcuffs matters more than backlink counts; most people talk themselves out before they experiment.

"the methodology >> and not putting in place all these handcuffs is far more important."

— 08:53

Most people talk themselves out of the experiment before running it. Understanding the method and declining the invented constraints matters more than the count of links, and the constraints are almost all self-imposed.

Given the choice, prefer a lower-authority backlink from a niche page that sends you real, relevant traffic over a high-authority one from a page that ranks but sends nothing.

"So, I put up this question on X"

— 10:24

Given the choice, take the smaller link from a niche page that sends real relevant visitors over the large one from a page that ranks and sends nobody. As a single metric the score is brake horsepower — impressive until you need the thing it ignores.

11:20 - The money-and-health category is applied far more widely than it should be

The your-money-your-life category is applied far too widely; most business, dental, fintech and finance topics do not trigger the strict treatment true medical topics do.

"I think people over um have a very wide net of YM."

— 11:20

Most business, dental, financial and technology topics do not trigger the strict treatment that genuine medical topics do. The category has been stretched into a general excuse for caution.

12:55 - Site-wide relevance is not a real ranking factor

Site-wide relevance is not a real ranking factor; deciding it would demand human editors the engine cannot afford, since even one field splits into many sub-fields nobody agrees on.

"I don't care about sitewide relevance. I imagine imagine how manual that process would be"

— 12:55

Deciding it would require human editors at a scale nobody can fund, and a single field splits into sub-fields that practitioners themselves do not agree on. The mechanism is not available, so the factor is not real.

13:52 - Page relevance is a downstream metric rather than an input

Page relevance is not that important either; the original page-rank formulation is really about the words in and around the anchor, and relevance is a downstream metric that tightens as authority tightens.

"I don't even think page relevance is that is that important."

— 13:52

The original formulation is about the words in and around the anchor. Relevance is downstream — it tightens as authority tightens — rather than an input you can set on the page.

15:33 - An entity does not exist until people search for it

You cannot fabricate yourself into a known entity with schema; an entity does not exist until people search for it, and stops existing when they stop.

"It's kind of like for my own name."

— 15:33

An entity comes into being when people search for it and stops existing when they stop, which no amount of structured markup can substitute for. You cannot declare yourself into the graph, and the hardest links to win are exactly the ones that would prove otherwise.

Broken-link building works: publish question, glossary or technical pages, find sites with dead outbound links, and offer yours as the replacement.

"Another thing about link building is um broken link building."

— 17:04

You give the other site a reason to link by fixing a real defect on their page, which is what separates it from asking. An expensive outreach firm scans for broken links and emails the admin, the editor and the SEO contact — and the outreach is slow and costly.

21:55 - An author bio does very little on its own

An author bio does little; even a recognised author only earns loose association, not a verified vetting, and an author page has authority only if people search the name and click it.

"It sounds a bit idealistic. I mean, it sounds like a journalist's job, right?"

— 21:55

The engine associates by confidence rather than flying out to certify that authors are correct. Authors make mistakes, lie, and are not peer-reviewed, and the plan to become a certified expert is long and arduous — you do not need one to rank, because the engine democratised publishing.

Guest and collaborated posts work best between genuinely related companies — integration partners doing joint releases, mutual posts, integration pages and contextual anchors — but the page must rank and get traffic to help.

"that's exactly what I'm talking about when I talk about between friendly companies,"

— 23:26

Integration partners running joint releases, mutual posts, integration pages and contextual anchors — that is a real relationship producing real links. The page still has to rank and get traffic, or the relationship produced nothing measurable.

25:45 - A blog selling posts across unrelated topics helps nobody

A blog that gets bored and just sells guest posts across unrelated topics helps nobody; a real guest post is a contributed or collaborated piece between two companies demonstrating a relationship.

"If if you write for like uh someone builds a blog and then they"

— 25:45

A blog that drifted into selling placements across unrelated topics helps nobody, and the post that does work is a genuinely contributed piece between two companies with something to demonstrate. The distinction is whether a relationship exists.

Relationship-first networking builds links naturally; run a local business meetup, discourage selling, and referrals and legitimate cross-links follow.

"the idea that you know"

— 26:28

Trust formed before any pitch converts into recommendations and reasons to link. A no-business-cards Thursday-morning coffee network and a weekend unconference whose members' own blog posts made the event rank are the two cases.

28:55 - Referrals beat pitching, and the room rewards one real conversation

Referrals beat pitching: bouncing round a room announcing your service wins nothing; arriving early, holding one real conversation and investing in the relationship produces clients through trust.

"what's funny? the when I started doing SEO, the way that I was getting clients is"

— 28:55

Bouncing around a room announcing a service wins nothing. Arriving early, having one real conversation and investing in that relationship produces clients — and a known brand also reduces bounce-back and lifts click-through, which behaves like a trust signal whether or not it is one.

31:26 - A hype and disinformation wave is running through the field

A generative-search hype and disinformation wave is running, with firms paying practitioners to declare search dead and recycle old myths as new AI-optimisation rules.

"to questions from the SEO subreddit"

— 31:26

Vendors profit from a manufactured shift, so they fund the narrative that it has already happened. A forum's moderation drama, offers to pay for SEO is dead takes, and recycled experience-and-schema myths are the visible symptoms.

33:39 - The assistants do not crawl for the best answer, they synthesise what already ranks

AI assistants do not crawl a forum site to find the best answers; they run a query that returns what the search engine already surfaces and synthesise it, so they are not mirroring the engine's knowledge graph.

"quick conversation one Friday"

— 33:39

They run a query, get what already ranks, and synthesise it — they do not crawl a forum looking for the best answer, and they are not mirroring anybody's knowledge graph. The pipeline is more ordinary than the marketing.

35:22 - The prompt is not the query

The prompt is not the query; the assistant runs on an expanded query fan-out, and if you rank for that fan-out you rank inside the assistant.

"I think it's just because they don't understand the prompt is not the"

— 35:22

The assistant runs on an expanded fan-out rather than on the words typed. Rank for that fan-out and you rank inside the assistant, which makes the target knowable rather than mystical.

36:44 - The answer is live-crawled on every fan-out rather than recited from an index

AI search live-crawls pages on every query fan-out rather than reciting a stored index, which is why an edit to a page you rank in shows up in the synthesised answer almost immediately.

"how is this building an index and then over 90 days it went and retrieved these files 37 27,000 times?"

— 36:44

Editing a page you already rank with shows up in the synthesised answer almost immediately, which is a testable claim and an easy one to test.

37:57 - The model companies cannot build crawling infrastructure, so they lease the result

AI companies cannot build the crawling infrastructure search engines have, so they leverage the search engines; training learns vocabulary and heuristics, but the graphics-style processors that do the model maths are not the crawling, disk and network infrastructure indexing needs.

"needed for training when when an LLM is being trained, it learns vocabulary."

— 37:57

Building crawling infrastructure is not the same problem as training a model — the processors that do the maths are not the disks and networks that indexing needs. So the model companies lease the result rather than duplicating it.

40:35 - The analytics answer the question before the assumption does

Before assuming your audience moved to AI, look at the analytics; reported AI-assistant traffic is a tiny, sometimes declining slice, which is a lack of adoption, not a lack of coverage.

"I had this discussion recently and um what I would say to people"

— 40:35

Reported assistant traffic is a small and sometimes declining slice, which describes a lack of adoption rather than a lack of coverage. The analytics answer the question before the assumption gets a chance to.

44:28 - Thin content is not a real thing

Thin content is not a real thing; there is no minimum word count, information-gain is an unimplemented patent, and one-line pages rank fine — the thin-content fear is a hangover from writers paid by the word.

"So I don't think there's a problem with thin content."

— 44:28

There is no minimum word count, information gain is an unimplemented patent, and single-line pages rank perfectly well. The fear is a hangover from an era when writers were paid by the word.

45:44 - The real problem is authority rather than content

The real problem with content at scale is mass AI-generated pages, and the deeper issue is authority, not content — match your page rollout to the authority you actually have.

"problem with scaled content, uh whether you all write write it all with AI"

— 45:44

Mass-generated pages are the actual problem, and underneath that the constraint is authority rather than content. Match the rollout to the authority you have and the question of length stops mattering.

Authority decays sharply as it passes through links; roughly an 85% tax per jump means it is near-dead within three jumps and no authority survives four, and only if every jump is judged relevant.

"there's a website that does research into it"

— 46:40

Something like an eighty-five per cent tax on each jump means it is nearly gone within three and entirely gone by four — and only if every jump is judged relevant. Distance is expensive in a way most site structures ignore.

Internal links are only wiring; they route authority you already have but cannot create it, so a new, un-authoritative site with perfect internal linking still will not rank.

"remember I we did a podcast and I talked about people painting"

— 47:30

They route authority that already exists and cannot create any. A new site with immaculate internal linking and no authority to route still will not rank, which is the whole reason the wiring metaphor is worth keeping.

48:21 - Each question gets its own page

Give each question its own page, because the page name sets the relevancy score that decides how much of your authority is spent; putting several questions on one page dilutes each.

"I want to say you're not somebody who hates content that is written"

— 48:21

A full topical match on a dedicated page uses all the available authority where a shared page splits it. An assistant writes 120-word answers from a one-line prompt, one page per question. A high-authority site can instead cluster FAQs on one competitive page to lift click-through — a different goal from growing authority.

50:07 - Twelve to twenty question pages is the starting range

Start with roughly 12–20 question pages, wait to see what ranks, and build up from whatever does — but do not go beyond that range on a new site.

"I start with like 12 to 20 and I wait to see what ranks."

— 50:07

Twelve to twenty, then wait and see what ranks, then build from whatever did. Going beyond that range on a new site spends authority that has not been earned, and the waiting is the part people skip.

51:02 - A page that rises and falls has been rotated out

A page that rises then falls like a failed rocket launch has been rotated out; the maths of click-through history, topical authority and backlink authority scored it too low, so park or remove it and return once you have built authority for that area.

"typically, you'll actually see that the page will will rank like this and come"

— 51:02

A page that climbs and then falls away has been rotated out: the combined arithmetic of click history, topical authority and link authority scored it too low. Park it or remove it, and return once that area has authority behind it.

51:57 - A stuck page republished at a new URL can lock at the top

Once you have built topical authority over time, republishing a stuck page at a new URL — near-identical content, a redirect from the old one — can lock it at the top, proving the block was authority, not content quality.

"I always share this story about how I took this"

— 51:57

The new URL re-enters the ranking trial backed by authority the topic did not have the first time. A page that flickered onto page one for years held number one after a one-word-different URL and a redirect — and it only worked because the authority was built in the interim.

53:45 - "It takes months" is often just sitting and waiting

"SEO takes months" is often just sitting and waiting; there is no time appreciation in page rank or anything the engine has said — but click-through and bounce-back testing genuinely can take months, so keep publishing other pages and building links meanwhile.

"when people talk about SEO taking months and they sit there and wait and wait"

— 53:45

There is no time appreciation in the original formulation or in anything the engine has said. What genuinely takes months is click-through and bounce-back testing — so the answer is to keep publishing and building while it runs.

55:51 - A site turning serious should expect rises and rotations

When a long-poor-quality site starts taking SEO seriously, expect pages to rise then rotate after core updates; a fall within about three positions or to page two is normal, but more than ~6 weeks flags a problem you fix by rehoming to a new slug and supplying supporting FAQs and internal links from easy-ranking pages.

"we talked about this concept of like a sitewide quality score."

— 55:51

Pages rise and then rotate after core updates, and a fall of a few positions or to the second page is normal. Beyond about six weeks it is a problem, fixed by rehoming to a new slug with supporting questions and internal links from pages that already rank.

1:01:06 - Low-volume high-intent niches are won on the click-through battle

In low-volume, high-intent B2B SaaS niches you win on the click-through-rate battle because the numbers are tiny; take the easy term, stack related items onto the ranking page, and move on from terms the big brands hold too tightly.

"I work in in um B2B SAS all the time"

— 1:01:06

The numbers are tiny, so the click-through battle is winnable in a way it is not at volume. Take the easy term, stack related items onto the page that ranks, and leave the terms the large brands hold too tightly.

1:02:26 - Comparison pages are vital and need not be review sites

Competitor and comparison pages are vital in B2B SaaS and need not be review or rating sites; list your own top features, or list rivals in no ranked order, and lean on the fact that wherever there is an A there is always a B.

"in intent and and there's a lot of like for example"

— 1:02:26

They are essential and they do not have to be review or rating pages. List your own top features, or list rivals in no particular order — and lean on the fact that wherever there is an A, somebody is looking for a B.

1:04:00 - The fear of too much optimisation is misplaced, with one exception

The fear of "too much SEO" is mostly misplaced, but automated internal linking is the exception to avoid; the story that spiders read your pillar-to-money-page structure is a cute fairy tale, not how it works.

"I want to actually So something that"

— 1:04:00

Most of the fear is misplaced, and automated internal linking is the exception. The story about crawlers reading a pillar-to-money-page structure is a pleasant fiction rather than a description of the mechanism.

1:05:47 - Authority cannot be spent before it exists

Stop spending authority you do not have; you cannot declare that blog posts will link power into a hub page before those posts and their neighbours themselves rank — ranking comes first, then you use the authority.

"Whoa, whoa, whoa, whoa, whoa. You're you're writing checks that your bank balance can't afford."

— 1:05:47

You cannot declare in advance that a set of posts will push authority into a hub page before those posts rank themselves. Ranking comes first, and the authority gets spent afterwards — in that order, always.

Keep internal links few and deliberate — never more than three, never from non-ranking pages, never pointing two synonym pages the same way — because the engine handles synonyms poorly and internal links decay.

"I don't think Google's really good at synony handing handling synonyms and semantics."

— 1:06:09

Over-linking and synonym collisions confuse an engine that already struggles with meaning. One acronym means both a security operations centre and a system on chip, and nothing in the link resolves which.

1:06:52 - Near-synonym pages cannibalise one another

Too many near-synonym pages cannibalise each other; no-indexing the too-similar ones can multiply traffic without new content.

"The second type of optimization is where you introduce too many synonyms and they cannibalize."

— 1:06:52

Too many near-identical pages compete with each other, and no-indexing the closest duplicates can multiply traffic without producing anything new. The gain comes from removal rather than addition.

1:08:15 - There is probably no structural over-optimisation penalty

There is probably no structural over-optimisation penalty; the engine is unlikely to enforce one page structure because it would make the web boring and cannot police free-form content and wordplay.

"much optimization. Does Google have a penalty for overoptimization? I I don't think so."

— 1:08:15

Enforcing one page structure would make the web uniform, and free-form content and wordplay cannot be policed at scale. The mechanism to punish structure does not plausibly exist.

1:09:25 - In controlled tests over-optimisation is real

In controlled tests over-optimisation can be real: the exact keyword in title, slug, meta description, heading, first sentence and alt text failed to rank, and removing it from the alt text alone started the ranking.

"I got to tell you, I've literally So, you and I, we both"

— 1:09:25

Piling the same keyword into every slot can trip a saturation signal that a single removal releases. The evidence is documented before-and-after keyword-placement tests rather than a claim about intent.

1:12:54 - Slow and judicious linking stays under any trip wire

Slow, judicious internal linking — two or three pages, no more than three links — probably keeps you under any over-optimisation trip; not every nail needs a sledgehammer, and more is not always better.

"I think that comes back to that sort of like slow judicious process"

— 1:12:54

Stuffing hurts, and exact-match anchors are riskier than they were — an amount once considered safe was hit by a recent update. Slow, judicious linking across two or three pages stays under whatever trip wire exists: not every nail needs a sledgehammer, and optimisation is subtraction.

1:14:21 - How much is right is relative to the keyword

How much SEO is right is keyword-relative: a competitive term with authoritative rivals demands real work, while a term nobody targets may need only the keyword in the page title.

"I like that distinction that you put in"

— 1:14:21

Effort should match how contested a term is rather than a fixed checklist applied everywhere. Backlink lists work as crowdsourced discovery of soft terms that tools often hide.

1:15:32 - Attribution keeps getting harder, so the journey gets captured first-hand

Attribution in SaaS keeps getting harder — the look-back window is gone and most organic arrives via branded paid search — so capture the journey yourself with UTMs on your own links and first-party storage exported to the CRM on consent.

"Very interesting question."

— 1:15:32

The look-back window is gone and most organic arrives through branded paid search, so the journey has to be captured directly — parameters on your own links, first-party storage, exported to the system of record on consent. And the call recordings hold the intent nobody wrote down.

1:22:01 - A brand should not be allowed to handcuff the content strategy

Do not let a brand handcuff the content strategy; glossary and informational terms belong out of paid search — make what, how, why and does your first negative keywords — while competitor-plus-comparison terms in paid search catch high-intent evaluators rivals cannot easily reproduce.

"that's very difficult is um a lot of companies like sort of like handcuff the content"

— 1:22:01

Glossary and informational terms belong out of paid search, which makes what, how, why and does the first negative keywords. Competitor-plus-comparison terms in paid search catch high-intent evaluators, and rivals cannot easily reproduce those.

1:22:55 - Thought-leadership-only usually means nobody checks the analytics

"Only thought leadership on our blog" usually means nobody checks the analytics; a slug written for social ("how we're crushing this industry") means nothing to the engine, which does not infer it stands for a real product category, so those posts get an initial spike then no traffic.

"is a problem like is it productled is it email driven is it social"

— 1:22:55

A slug written for social — how we are transforming our industry — means nothing to the engine, which does not infer that the phrase stands for a real category. Those posts spike once and then flatline, and nobody looks at the analytics that would show it.

1:26:16 - The primary keyword goes in the slug and the support goes in the title

To satisfy a brand-police client, you do not need the keyword everywhere — put the primary keyword in the slug and wrap supporting keywords into the title, then test.

"You don't need to have the keyword in everywhere."

— 1:26:16

The primary keyword goes in the slug and the supporting terms get wrapped into the title, then it gets tested. That is usually enough to satisfy both the brand team and the engine, which is a smaller compromise than either side expects.

1:26:48 - A forbidden keyword can live in the slug alone

If a forbidden keyword cannot go in the body, put it in the slug — but one bare word competes with irrelevant results while too many words narrow the page too far, so use the least on-point set and treat it like a Venn diagram you keep testing.

"So, for example, you're going to write a blog post about"

— 1:26:48

One bare word competes against irrelevant results and too many words narrow the page until nobody reaches it. The least on-point set is the target, and it is a Venn diagram you keep testing rather than a rule you apply once.

1:27:52 - A slug can change without changing the date or the subject

You can change a slug without changing publish date or subject (it must reindex), and you do not need the keyword down to the subheadings — repeating it does not raise ranking, because relevance caps at 100% and topical authority matters far more beyond that.

"cool if they just if they just are familiar with the page."

— 1:27:52

The slug can change without touching the publish date or the subject, provided it reindexes. And the keyword does not need repeating down through the subheadings, because relevance caps out and topical authority is what matters past that point.

Personal Branding Lessons

An episode mostly spent removing rules. The moves below are what remains once the invented constraints are gone.

Evaluate the domain rather than its authority score

A high score in a tool can be assembled from estimated traffic and rank on a site that is declining rather than growing, and its pages may be largely unindexed. Looking at the actual domain takes a minute and reading the number takes none, which is the entire reason the number wins arguments it should lose. 04:15

Given the choice, take the smaller link from a niche page that sends real, relevant visitors over the large one from a page that ranks and sends nobody. As a single metric the authority score is brake horsepower: impressive on its own, useless the moment you need the thing it ignores. 10:24

Give each question its own page

The page name sets the relevancy score that decides how much of your authority gets spent on it, so putting several questions on one page dilutes each of them. One question, one page, one name. It is a structural decision rather than an editorial one, which is why it is easy to get wrong. 48:21

Start with twelve to twenty question pages and build from what ranks

Publish the range, wait, see what ranks, and build up from whatever did. Going beyond that on a new site spends authority that has not been earned. The waiting is the part people skip, and skipping it is what produces the pages that rise and then rotate out. 50:07

Match the page rollout to the authority you actually have

The real problem with content at scale is not length or quality but authority — there is no minimum word count and one-line pages rank fine. Publishing beyond what the site can support is what fails, and it fails for reasons no amount of editing the pages will fix. 45:44

Never more than three, never from pages that do not rank, and never pointing two synonym pages the same way, because synonyms are handled poorly and internal links decay. Internal links route authority that already exists and cannot create any, so the restraint is arithmetic rather than caution. 1:06:09

Listen to the sales call recordings

Sales hears the objections, marketing hears people say they were searching for this, and the product manager finds a third story again. Recordings and transcripts recover the real intent nobody wrote down — and a user who could not find a feature on the product is an assist waiting to be picked up. 1:18:36

Put the primary keyword in the slug and the support in the title

When a brand team will not allow the keyword everywhere, it does not need to be everywhere. The primary term goes in the slug, the supporting terms get wrapped into the title, and then it gets tested. The slug can be changed later without touching the publish date or the subject. 1:26:16

Questions

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

No. The score is a single made-up number for a whole site, and a site is strong in some topics and absent in others, so one figure cannot represent it. The "you can't get links below a threshold" rule is called a myth — a giant site earns its high score from countless small ones, and there is no negative score to avoid. Judge the actual page's ranking, indexation and traffic instead. 00:44

Which is the better backlink: a low-authority page that sends traffic, or a high-authority page that sends none?

The one that sends traffic, even from a much smaller site, because real relevant visits correlate with ranking better than a raw authority score does. A big-name page that ranks but sends nothing hands over little; a niche page in your topic that sends real clicks acts like the link working. If you are weighing a mention, weigh who is actually searching for and clicking through to you, not the size of the logo. 10:24

Start with the network you already stand inside — friends, a chamber of commerce, a weekly meetup, clients, the businesses you work alongside — and if none exists, start one, because those links have an honest reason to be there. The spam lists selling backlinks are useful too, not to buy from but as free keyword research: they surface soft, trafficked pages you can outrank so the clicks act like the link. 02:24

They rank because almost nobody targets those exact questions, so you satisfy the intent, avoid the bounce back to results, and earn clean clicks that build authority you can funnel elsewhere. Each question gets its own page, since the page name sets the relevancy score that decides how much authority it can use. Start with roughly twelve to twenty, keep what ranks, and pull what clearly will not. 07:15

Why did my page rank and then drop, and what do I do about it?

A page that spikes then falls like a failed rocket has been rotated out: given a click-through trial, then demoted when the fuller scoring — click-through history, topical authority, backlink authority — judged it too low. Keeping it live does not recover it. Once you have genuinely built authority for that topic over time, republish the same content at a new URL with a redirect from the old one. 51:02

Are AI assistants replacing search, and should I move my strategy to them?

Look at your own analytics first. The reported traffic from AI assistants is often a tiny, declining slice — a lack of adoption, not a lack of coverage. The assistants do not crawl a forum to find answers; they run a query, live-crawl what the search engine already surfaces, and synthesise it, leaning on infrastructure they cannot build themselves. Ranking for the expanded query behind the prompt is what earns citations. Deciding when to jump is a first-mover-versus-wait call the data should settle. 33:39

Is thin content really a ranking problem?

No. There is no minimum word count, the much-cited information-gain idea is described as an unimplemented patent, and one-line pages rank fine when they answer a real, in-demand question. The thin-content fear is traced to writers once paid by the word. The genuine risk is mass AI-generated pages, and the deeper constraint is authority, not length — publishing beyond the authority you have earned leaves the pages unranked whatever their word count. 44:28

Is over-optimisation a real penalty, and how much SEO is too much?

There is probably no structural penalty for how a page is laid out — dictating one structure would make the web boring and cannot police free-form wordplay. But keyword saturation is real: the exact keyword in every on-page slot failed a ranking test until it was removed from the alt text alone, and heavy exact-match anchor text is newly risky. Restraint wins — a couple of links, not a sledgehammer on every nail — because more is not the same as better, a bias worth knowing before you pile options on. 1:08:15

How do I measure content-driven leads versus product-led growth in software?

Attribution has got harder: the look-back window is gone and most organic arrives through branded paid search, so the platform no longer shows the path. Own the record instead — UTMs on your own links, and a first-party store of the first-visit journey exported to your CRM on a consent tick — then read the sales-call recordings, where the objection the sales team hears and the "I was searching for this" the marketer hears are two different stories in one conversation. 1:15:32

Sources

SaaS SEO in 2026: Link Building, Topical Authority & the GEO Hype

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.