Exact Match Domains Still Work (Most SEOs Are Wrong)

Edward Sturm

Abstract

Fifty-four minutes arguing that a tactic the field has written off is still working. Exact-match and keyword domains are treated here as a ranking advantage rather than a dead myth — with the qualification doing most of the work, because an advantage is not a shortcut and the authority still has to be earned. The structural fact underneath it is that the engine does not really know what a brand is and will not protect one.

• Betting an income on a single domain is a concentration risk; on a commercial results page you want to own more than one slot.
• Registering the keyword domain does not send anybody to the first position — the domain is a starting advantage, not a ranking.
• The next domain to buy is already in your own console, in the top keywords by impression.
• Returns on many domains diminish, because authority comes from activities rather than from registrations.
• Generative optimisation is not a different discipline — but if a security layer blocks the crawler you fall out of the answer entirely.
• Promoting yourself inside a generated listicle now gets you removed; targeting a variation of the search is the durable move.

Everything below follows the episode in order, from the opening claim through portfolio construction and the spam boundary, into the generative half, and out to what a page has to reach to be cited. For anyone building a personal brand the useful part is the reframe: the domain question is a portfolio question, and the same handful of slots is what both a results page and an answer are handing out.

Chapter summaries

00:27 - Exact-match domains remain a ranking advantage rather than a dead myth

Exact-match domains are still a ranking advantage, not a dead myth — a massive one the search engine has quietly played down.

"Absolutely. An advantage in ranking."

— 00:27

Exact-match domains are still a ranking advantage rather than the dead myth the field treats them as. The claim opens the episode and everything after it is the qualification: an advantage is not a shortcut, and the rest of the hour is about the difference.

00:40 - Parasitic publishing on a big platform still needs the work behind it

Parasitic SEO — publish on a big platform, link to yourself, and rank — does not just work; the normal rules of SEO still apply and real work is done to rank those pages.

"Right? It's like parasitic SEO. I think a lot of."

— 00:40

Parasitic publishing — putting a page on a large platform and linking back — still needs the work behind it. Borrowing somebody else's authority does not remove the requirement to have something worth ranking.

01:30 - Betting an income on a single domain is dangerous

Betting your income on a single domain is very dangerous; in a results page of ten to a hundred slots, one domain controls almost none of the narrative.

"So EMDs are a very powerful play."

— 01:30

Betting an income on a single domain is dangerous. On a results page you want to own more than one slot, which reframes the domain question from branding to portfolio. Concentration is a risk position rather than a virtue.

02:37 - The engine does not really know what a brand is and will not protect one

The search engine does not really know what a brand is and will not protect your brand name — it will sell it to other advertisers and let competitors rank for it.

"Makes it surprisingly easy. Google has a small problem."

— 02:34

The engine does not really know what a brand is and will not protect one. That is the structural fact the whole episode rests on: the protection people assume exists is not there, and the space it leaves is what an exact-match domain occupies.

03:12 - Registering the keyword domain does not send anyone to the first position

Registering the keyword or brand-variant domain does not send you to number one; you still develop authority and earn your click-through — but it is a narrow, small amount of work.

"The way Google knows that Nike and Nike.com are related."

— 03:12

Registering the keyword domain does not send anybody to the first position. The domain is a starting advantage and authority still has to arrive, which is the correction that separates this from the version of the tactic that fails.

03:55 - Two routes exist: make the keyword the brand, or keep them separate

Two routes exist — make a generic keyword itself your brand, or keep a branded domain and build external EMDs — and the second is preferred, though many did the first for twenty years.

"Are you suggesting somebody create their brand as an exact."

— 03:55

Two routes exist: make a generic keyword the brand itself, or keep a brandable name and run the keyword domain separately. They are different businesses rather than different tactics.

05:24 - Even a first-place brand should hold more than one slot on the page

For single-purpose commercial keywords, even a first-place brand should take the second and third slots with EMDs; a domain yields roughly two results, so controlling the first two shrinks the field.

"So I think with a lot of businesses, they."

— 05:24

For a single-purpose commercial keyword, even a first-place brand should hold more than one slot on the page. The alternative mode is also named — dozens of buyer-intent pages rather than pages answering questions.

07:41 - Fully automated domains are viable for fishing keyword data

Fully automated domains are viable for fishing keyword data — topics assigned, bots doing quality control — but need not be scaled to hundred-page sites; build content in phases as needed.

"Back to the podcast. Do you see any — I actually."

— 07:41

Fully automated domains are viable for fishing keyword data: assemble topics, publish, and see what the engine returns. The output is research rather than revenue, which is a legitimate use nobody budgets for — a map of who occupies the space falls out of it for free.

09:00 - Connecting a domain to the console is not required to optimise it

You do not have to connect a domain to the search console to do SEO; not connecting it hides the link between the domain and your brand — the engine only inspects a console when the offence is heinous enough.

"Absolutely no problem. I don't see any issues there."

— 09:00

Connecting a domain to the search console is not required to optimise it. A penalty against a scaled-content operator was also targeted rather than blanket, which is a more precise account than the one usually circulated.

10:51 - The sandbox is treated as a myth once authority is present

The sandbox — a waiting period before a new site can rank — is treated as not real: with the right authority a site can rank in an hour, going around the board as often as you like.

"It's also — I actually use it as an example."

— 10:51

The sandbox — a waiting period before a new site can rank — is treated as a myth once authority is present. The qualifier is doing the work: it is not that new sites rank instantly, it is that the delay is an authority problem rather than a clock.

11:37 - When a site gets struck for machine-scale content is an open question

When the search engine strikes a site for machine-scale content is an open question with no consensus worth agreeing to.

"It's very specific. Like it could be a job site or."

— 11:37

When a site gets struck for machine-scale content is an open question. Said plainly rather than filled in with a guess, which is the register that makes the rest of the claims worth weighing.

13:07 - The branding fears evaporate once the site earns money

The branding and trust fears around EMDs mostly evaporate once the site is well branded and people can find it and become leads; the real blocker is a culture that rejects change.

"Are there any — like, do you see."

— 13:07

The branding and trust fears around these domains mostly evaporate once the site earns money. It is a blunt observation about how objections are actually resolved in practice.

14:29 - The advantage starts with the name and runs through the targeted keywords

The EMD advantage starts with the domain name, then the targeted keywords, then the pages and topical authority built around it, plus inbound links matching the phrase and its long-tail variations.

"What signals give exact match domains an advantage when they do."

— 14:29

The advantage starts with the domain name and runs through the targeted keywords and the pages built on them. The name alone is the smallest part of it.

15:02 - No safe threshold is known for linking a parent domain to its satellites

There is no known safe threshold for linking a parent domain to its satellites; roughly fifty page-to-page links is fine, but too many deteriorate value and site-wide links negate most links.

"Sticking, when you do rank. Another good question that people."

— 15:02

Roughly half a practitioner's domains sit inside the parent brand and half do not, which is a portfolio decision made without a rule.

16:40 - A site that leans obviously toward its owner can backfire

A fake profile or comparison site that leans obviously toward you can backfire; the 99% best case is to build the EMD with your own branding as an owned asset.

"Think if you start setting up like a fake profile."

— 16:40

A site that leans obviously toward its owner — a fake profile, a comparison page with a predetermined winner — can backfire. The better build is a genuine branded asset, which costs more and survives inspection. On paid search, a high quality score is a main predictor of cost.

18:49 - A cheap memorable keyword domain can redirect to a buried section

A cheap, memorable keyword domain can 301-redirect to a buried section of your main site, giving a shareable address to send a social audience to your articles.

"I sent you yesterday."

— 18:49

A short branded phrase is sayable out loud in a video and forwards seamlessly, which replaces a URL too long to speak. It is a redirect rather than a ranking play — a keyword domain pointing at a long articles path, driving reads and email signups from an audience that would never have typed the real address.

19:48 - Exact-match domains turn spammy at the point of overstretch

EMDs become spammy when you overstretch, or use them as negative SEO against competitors; they are not free ranking — you still build authority and manage them as infrastructure.

"Really well. But I want to ask, we were talking about."

— 19:48

These domains turn spammy at the point of overstretch, or when used against somebody else. The boundary is named rather than left implicit, which is more useful than a general caution.

20:50 - Returns on many domains diminish, because authority comes from activity

Returns on many domains diminish: authority comes from activities like blogging, and it is hard to run more than two blogs, remember where articles live, or point single-domain social profiles at many sites.

"Well, is that a lot of people get authority from things."

— 20:50

Returns on many domains diminish, because authority comes from activities rather than from registrations. The work is the input, and a portfolio of empty domains is a portfolio of nothing. Splitting effort between the main brand and the satellites can be set as a rule.

24:17 - The redirect is a substitute for link building nobody has time for

The redirect trick is a substitute for link-building you do not have time to do; the correct move would be to build links to the article hub, but sending a large existing audience is the low-hanging fruit.

"I basically — the reason I'm even."

— 24:17

The redirect is explicitly a substitute for link building nobody has time for. Naming it as a substitute rather than as a strategy is the honest framing.

25:39 - A partial match is enough, and the expensive exact phrase is not needed

You do not need the expensive exact-phrase domain — a partial match is good enough, and the TLD, hyphens and country codes do not matter for ranking; memorability matters only for saying it aloud.

"Oh, it's great. I actually I had."

— 25:39

A partial match is enough, and the expensive exact-phrase domain is not needed. The top-level domain does not decide the ranking either, which removes two expensive assumptions at once.

27:06 - The console's own top keywords are where the next domain comes from

Choose an EMD from your own search console: the top keywords by impressions, where competition is strongest, and where paid spend is highest — then take a best-/top- variation if the exact phrase is unavailable.

"But actually, so for you, when you are buying an exact."

— 27:06

The next domain comes from your own console: the top keywords by impression are already telling you what to buy. It is free research that most people replace with a tool.

27:57 - Nobody clicks the sources in an answer, so the mention is the prize

In an AI answer, people almost never click the sources, so getting your brand into the first citation is enough — the satellite domain need not be memorable, only the brand must be.

"Or software.com, and those variations are almost always available."

— 27:57

In an answer, people almost never click the sources — so being mentioned is the prize rather than being linked. That changes what the target is, and most measurement has not caught up.

28:32 - Pages targeting unclaimed questions funnel authority to the money pages

Put up pages targeting "people also ask" questions nobody else targets, rank for them, and funnel that authority into pages aimed at competitive keywords.

"We have like so many amazing SEO tricks just."

— 28:32

Pages targeting questions nobody else has claimed funnel authority into the pages that carry commercial value. Same two-stage structure as elsewhere in the pillar: rank where it is cheap, spend where it is expensive.

29:09 - An exact-match domain is for authority that cannot be matched

Reach for an EMD when you are against authority you simply cannot match, or when you want to take over page one; then it becomes almost non-negotiable.

"If I'm up against someone whose authority I just cannot."

— 29:09

Reach for one of these domains when facing authority you simply cannot match. It is a tool for an asymmetric fight rather than a general-purpose upgrade, and picking the fight is the decision before it. A free-prefixed variant can also capture a term the branded domain has lost.

32:29 - Given a fixed budget the keyword domain beats the brandable one

Given a fixed budget, spend it on an exact-match domain rather than a brand domain — it is the fastest route to revenue, especially when targeting acquisition.

"Which is still there. If you had 10 grand to."

— 32:28

Given a fixed budget, the exact-match domain beats the brandable one on speed to revenue. It is a claim about sequencing rather than about which is better in the end.

33:23 - The claim that generative optimisation needs consensus is examined

The claim that generative optimisation differs from SEO because it needs "consensus in the results" rather than a single number-one applies mostly to listicles, and those AI list answers are close to random.

"All right, so I have a bonus question."

— 33:23

The claim that generative optimisation differs from search because it needs consensus is examined rather than accepted. The related pitch — that search is over — gets the same treatment.

35:36 - Being added to somebody else's ranked listicle is fragile

Being added to someone else's ranked listicle to piggyback visibility without ranking is a short-lived, fragile strategy; owning the position yourself with a supporting EMD is the durable move.

"Interesting is that a lot of geo brands have reached out."

— 35:36

Being added to somebody else's ranked listicle to piggyback visibility is fragile, because the position belongs to them. Owning the position is the alternative, and it is slower.

37:03 - Generative optimisation is not a different discipline

Generative optimisation is not a different discipline: the story that a different prompt yields different citations is fabricated — the list is still the list, and ranking for more than one variation is just long-tail.

"Right? So the idea that geo is somehow different from."

— 37:03

The story that a prompt drifts unpredictably describes long-tail behaviour that search has always had. It is the same problem in new packaging.

38:13 - A security layer blocking the crawler removes the citation entirely

If a security layer in front of your site blocks the AI crawler, you fall out of the AI's answer immediately when the engine tries to fetch you; an EMD helps because AI tools lack the engine's business-profile and console signals.

"I've been playing a lot with that."

— 38:13

If a security layer in front of the site blocks the crawler, you fall out of the answer entirely. It is the most concrete and least discussed failure in the whole generative conversation, and it is a configuration rather than a strategy.

40:58 - Training is overestimated, and most of an answer is live retrieval

Training is overestimated as a ranking input; roughly 90%-plus of an AI answer is live retrieval from search ranking in the moment, not model training, because the web is too large to store in a model.

"Whole thing about like, should you allow bots to train on."

— 40:58

Training is overestimated as an input: roughly ninety per cent or more of an answer is live retrieval. The practitioner will also trade protection for visibility, adjusting settings toward being fetched.

43:15 - Anyone can flood a brand term with cheap advertising

Because the engine does not link a brand name to its domain, anyone can flood a domain with cheap display ad traffic from a low-currency country — saturating it in an hour for a few dollars.

"That's crazy, that that."

— 43:15

Because the engine does not link a brand name to its domain, anyone can flood a brand term with cheap advertising. It is the same structural gap as the opening, seen from the attacker's side.

45:02 - Putting the author's own name into their own listicle now gets it removed

Putting your own name into your own AI listicle now gets you removed — it is anti the result you want — but posting under a separate handle on a forum evades the check, because the model can no longer trace the self-promotion.

"Was also — I just wanted to touch."

— 45:02

Putting the author's own name into their own generated listicle now gets it removed, and a separate handle is the reported bypass. The tactic is described as having flipped rather than as having been fixed.

45:46 - The self-promotion listicle has flipped from working to punished

The self-promotion listicle is a tactic that has flipped from de-indexed to actively harmful: agencies mass-produced them for clients, earlier research showed the engine de-indexing them, and now appearing can hurt.

"Why this is really crazy, because so many."

— 45:46

The self-promotion listicle has gone from working to punished, which is a description of a moving target rather than of a rule.

47:29 - The removal signal is unreliable, because the model does not count sources

The removal signal is unreliable: the "put their own name in their own list" message was relevant ~80% of the time, yet people known to be on no other list still appeared, and the model admits it does not count sources.

"But I don't know how good it."

— 47:29

The removal signal is unreliable, because the model does not count its sources and the answer can be fabricated. That undercuts the evidence for the previous beat, and it is said anyway.

48:42 - A model cannot explain how it works, because it does not know

You cannot trust a model to explain how it works because it does not know, and its list results are personalised: asking why someone is missing makes it insert them for you, so your view differs from another machine's.

"But these messages were new. Like I remember."

— 48:42

A model cannot explain how it works, because it does not know — and the lists it returns are personalised. Two reasons not to treat its self-report as data. An entire venture-funded industry exists to manufacture these listicles.

50:01 - Targeting a variation of the search beats the self-promotion listicle

Rather than a self-promotion listicle, target a variation of the search phrase — "top 10 [category] features" listing your own — and the engine does not know the difference.

"Yeah. It's just — so let me ask."

— 50:01

Rather than a self-promotion listicle, target a variation of the search itself. The variation is a real query with real demand, which is the difference between ranking and being inserted.

50:57 - A features page can still rank for the bare head term

A "top [category] features" page can still rank for the bare head term, because competition reduces to how well a rival shifts authority to that page — if they rely on topical authority alone, they are beatable.

"But I can get it to rank for top CRM."

— 50:57

A features page can still rank for the bare head term, which is the point of the variation strategy: you do not have to beat the head term head-on to appear for it.

51:32 - The page title carries half the relevancy for document formats

For text, document and PDF pages the page title is 50% of the relevancy score and the slug the other 50%; you need not repeat the keyword in both — a slug matching the target can carry it.

"It's a numeric game, right? So there's also a difference with."

— 51:32

For text, document and PDF formats the page title carries about half the relevancy. The title is doing half the work on exactly the formats where nothing else is available to carry it.

52:44 - An answer needs a top-five or top-ten position rather than the first

Ranking for an AI answer means reaching the top five or ten, not number one, which makes generative optimisation slightly easier than SEO — and it often rewards lower-competition phrases.

"When you say rank in it, do you just mean."

— 52:44

Ranking for an answer means reaching the top five or ten rather than the first position, which makes generative visibility a lower bar than search. The episode closes on that comparison, having spent an hour arguing they are the same discipline.

Personal Branding Lessons

An episode about occupying more of a page than one property can. The moves below are mostly portfolio decisions rather than optimisation ones.

Stop betting the income on a single domain

Betting an income on a single domain is dangerous. On a results page you want to own more than one slot, which reframes the domain question from branding to portfolio: the risk is not that the one site is bad, it is that one site can only ever hold one position. Concentration is a risk position rather than a virtue, and the fix is structural rather than tactical. 01:30

Hold more than one slot for a commercial keyword

For a single-purpose commercial keyword, even a first-place brand should hold more than one slot on the page. The alternative mode gets named too — dozens of buyer-intent pages rather than pages answering questions — so the choice is between depth on one term and breadth across many. Either way the unit being bought is page real estate, not a ranking. 05:24

Pick the next domain from your own console's top impressions

The next domain comes from your own console: the top keywords by impression are already telling you what to buy. It is free research, and most people replace it with a paid tool that guesses at the same numbers. The data is describing demand that has already reached you rather than demand that might exist. 27:06

Accept a partial match rather than paying for the exact phrase

A partial match is enough, and the expensive exact-phrase domain is not needed. The top-level domain does not decide the ranking either. Two assumptions come off the budget at once, and what is left is the ordinary work of earning authority on whatever name you bought. 25:39

Check that no security layer is blocking the crawler

If a security layer in front of the site blocks the crawler, you fall out of the answer entirely. It is the most concrete and least discussed failure in the whole generative conversation, and it is a configuration rather than a strategy — which means it is both cheap to fix and easy to never notice. 38:13

Target a variation of the search rather than a self-promotion listicle

Rather than a self-promotion listicle, target a variation of the search itself. The variation is a real query with real demand, which is the difference between ranking and being inserted. The listicle tactic has gone from working to punished; the variation was never the kind of thing that could be punished. 50:01

Aim for the top five or ten rather than the first position

Ranking for an answer means reaching the top five or ten rather than the first position, which makes generative visibility a lower bar than search. That is a comparison worth holding onto after an hour spent arguing the two are the same discipline: same work, shallower depth required. 52:44

Redirect a cheap keyword domain instead of building links you have no time for

A cheap, memorable keyword domain can redirect to a buried section of the main site. It is explicitly a substitute for link building nobody has time for — named as a substitute rather than dressed up as a strategy, which is what makes it usable. It is the least glamorous tactic in the episode and among the most immediately actionable. 24:17

Questions

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

Do exact match domains still work, or are they an outdated SEO myth?

They still work — the position taken here is that the advantage is not only real but sizeable, and has been quietly played down rather than removed. A keyword-matching domain carries the target phrase in the place the engine trusts most, so it starts ahead for that phrase. The one caveat is that it is not automatic: the match is a head-start, and authority and click-through still have to be earned before the domain ranks. 00:27

Do you have to connect a domain to the search console to do SEO on it?

No. Connecting a domain to the console is optional, and leaving it disconnected hides the tie between a satellite domain and the parent brand, since ownership is not visible where there is no console link. The engine is described as inspecting a console only when the offence is serious enough to warrant a manual look, so a small, well-behaved satellite has little reason to draw one. 09:00

When does buying keyword domains cross into spam?

At two points. The first is overstretching — registering more domains than you can genuinely build authority for and manage as real infrastructure, because an extra domain is not free ranking any more than a branded one is free marketing. The second is using them as negative SEO against competitors, which, sold to an educated buyer, can blow up in your face. The honest test is whether you can still feed and manage each one. 19:48

How do you choose which exact-match domain to buy?

From your own data. The method is to read the search console for the terms already earning impressions, where competition is strongest, and where paid spend is highest — then take a best- or top- variation when the exact phrase is unavailable or absurdly priced. That points the spend at demand you have already proven rather than a term you merely like, the same discipline as any honest round of niche research. 27:06

Does the top-level domain — .com versus .net or a new TLD — matter for ranking?

No. The extension is described as irrelevant to ranking: hyphens, country codes and the newer TLDs all carry the keyword just as well, and you do not need the .com. What the engine rewards is the keyword match, not the suffix. Memorability is the only reason to prefer one string over another, and that matters only for saying the name aloud — in a video, say — not for where the page lands. 25:39

When is an exact-match domain worth it instead of ranking your main site?

When the authority in your way cannot be matched in a reasonable time. Against rivals with unreachable networks and branding, a fresh domain on a narrower phrase is often the only route in, and where the aim is to take over the whole first page it becomes close to non-negotiable. Given a fixed budget aimed at revenue — especially ahead of an acquisition — the stated preference is the matched domain over the brand domain, because it monetises a proven phrase fastest. Map the ground first with a competitive analysis. 29:09

Is generative engine optimisation really different from SEO?

Mostly not. The "consensus in the results" claim is narrowed to listicles, and those AI list answers are described as close to random, reading across ranked pages and returning a shifting subset rather than obeying a new law. The idea that a different prompt yields different citations is dismissed as a made-up differentiator — ranking for more than one variation is just what search always called long-tail. What actually rewards you is the same ordinary ranking, which is why understanding how the algorithm really rewards content still applies. 33:23

Why do self-promotion listicles in AI answers now hurt instead of help?

Because the tactic has flipped from ignored to penalised. Agencies mass-produced these lists, earlier research showed the engine de-indexing them, and the current state is worse: putting your own name in your own list can now get you removed, because the model flags self-reference it can attribute. Posting the same content under a separate handle on a forum evades the check, since the attribution is broken — but the signal itself is admitted to be unreliable. Standing on trust rather than trickery is the safer read, and reputation is the asset that compounds. 45:46

Can a security or CDN layer stop your site being cited by AI tools?

Yes, and quietly. In the case described, every AI-crawler setting was toggled on and the security layer in front of the site still blocked the crawlers, until removing it entirely restored the citations about a day later. The mechanism is that an AI answer is built by fetching ranked pages live, so a page the crawler cannot reach simply drops out of the answer. It is offered as one person's experience, not a comprehensive test, so the practical move is to verify the crawler can actually fetch you. 38:13

Do you need to rank number one to appear in an AI answer?

No — a top-five-or-ten position is enough, because the AI reads a shortlist rather than a single winner. That is named as the one way generative optimisation is genuinely easier than search, and it is compounded by the phrases these answers draw from often being lower-competition than the highest-volume terms. The practical conclusion is that a good search-optimisation footing is the right place to start, not a separate discipline, and standing out inside a shortlist is still a differentiation problem. 52:44

Sources

Exact Match Domains Still Work (Most SEOs Are Wrong)

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.