GEO vs SEO: James Dooley Says Only 1% of SEOs Can Actually Do It

Edward Sturm

Abstract

An hour and a half of two respected practitioners disagreeing in public about whether being found inside an AI answer is a separate discipline or the same work under a new name. Neither position is a hedge. What emerges between them is more useful than either: the activities overlap almost completely, the budgets do not, and the label decides who gets hired regardless of who can actually do the job.

• Owning the whole first page is a different objective from ranking one site first, and it commands much larger fees.
• High-competition search tactics move rankings and do not move answers, which is the cleanest evidence the two systems differ.
• Claim, frame and prove needs corroboration off the site — awards, reviews, testimonials — not another page.
• The fan-out now points at specific review and directory sites per industry, which relocates the work entirely.
• Owned assets beat rented profiles, because rented profiles get deleted and demoted.
• Fifty pages produce two that rank; the choosing cannot be done in advance.

Everything below follows the conversation in order, from the opening disagreement through budgets, positioning and what is actually performed, into the tactical repertoires of both camps, the amplification mechanics, and out through indexing, crawl archives and trust models to what each of them would do first. For anyone building a personal brand the useful part is the corroboration argument: what you say about yourself is the least load-bearing thing on the page.

Chapter summaries

00:00 - Two respected practitioners hold opposite views on whether this is one discipline or two

Two respected practitioners hold opposite views: one says getting found in AI is a different discipline, the other that it is SEO renamed.

"Two of the greatest, most respected search"

— 00:00

The disagreement is the episode. One holds that being found in a model is a different discipline with its own department; the other that it is the same work under a new name. Neither position is a hedge, which is what makes the hour worth following.

00:34 - The case for a separate department is that first position does not earn citations

Getting found in AI is different enough to warrant its own department; people who rank first in the search engine often can't earn AI citations.

"I just want to start by saying I have absolutely put myself on a death wish"

— 00:34

The evidence offered for the split is specific: people who reliably hold first position in search often cannot earn a citation in an answer. If the same competence produced both, that would not happen.

02:31 - Reputation management means owning the whole first page rather than one position

Reputation management means owning the whole first page, not just ranking one site first; "AI visibility = SEO + reputation management".

"OM is generally reputation management."

— 02:31

It is a page-level objective rather than a position-level one. Owning every result for a brand term is a different job from ranking one site first, and the formula offered is visibility as search plus reputation. Owning the page is the unit.

02:58 - Anyone wanting reputation management still goes to a search agency

Anyone wanting reputation management still goes to an SEO agency; the overlap is near total — every reputation agency is an SEO agency ranking multiple pages and domains.

"Do you think that if someone wants ORM, they would go to an SEO agency"

— 02:58

Every reputation agency is a search agency ranking multiple pages across multiple domains. The overlap is close to total in practice, whatever the labels on the doors say — which is the strongest version of the one-discipline case.

05:34 - "Build a great brand and the model will come" is not how a model gets designed

"Build a great brand and the model will come and find you" is hogwash; that is not how engineers design a model.

"of the challenges with SEO versus GEO being two different things is how the market is bifurcated, right?"

— 05:34

It is not how a model gets designed. Engineers build retrieval and weighting; they do not build a mechanism that goes looking for admirable brands. The comforting version of the advice describes an outcome and calls it a method.

07:12 - Different pots of money fund search work and reputation work

Different pots of money fund SEO versus reputation and branding; some who do reputation and AI visibility don't do SEO at all.

"I so partially agree with what you're saying."

— 07:12

The budgets are separate, and separate budgets create separate suppliers. Some of the people doing reputation and visibility work do no search work at all, which is an argument about procurement rather than about capability.

10:12 - The real crux is positioning rather than what is actually performed

The real crux is positioning versus what is performed: one side says the activities are identical, the other says there are genuine specialties.

"actual uh what is being performed."

— 10:12

Separating the brand label from the task list is what clears the argument, because the two sides are not disagreeing about the work. The reframe comes from the host, to keep the debate honest.

10:33 - Buyers looking for the label search for the label

If you searched "best AI-visibility agency" and an SEO agency ranked, you'd hire them — but buyers looking for the label search for the label.

"Let me let me ask it a different let me ask it a different way."

— 10:33

If a search agency ranked for the new label, the buyer would hire them — but buyers looking for the label search for the label, which means the label decides who gets seen. Reputation fees also dwarf search fees, which decides who can afford to compete.

11:52 - Visibility is achievable with no site at all, which one side calls failing at the basics

You can get AI visibility with zero SEO — review-site listings and reviews, no website — which the other side calls simply failing at SEO.

"In other words, can you do geo without doing"

— 11:52

Listings and reviews with no website at all can produce visibility, which one side offers as proof of a separate discipline and the other calls simply failing at the basics. Both readings fit the same fact.

12:25 - The fan-out now points at specific review and directory sites per industry

The query fan-out now points at specific review and directory sites per industry (site: operators), not the search engine itself.

"I'd say that's your review base."

— 12:25

The fan-out now runs against specific review and directory sites per industry rather than against the open web. That relocates the work: being in the right directory matters more than outranking anyone.

13:30 - Two directory profiles plus social can lift visibility with the site not ranking

One directory listing won't guarantee ranking in a model, but two directory profiles plus social can lift AI visibility with the site not ranking at all.

"let's take that. Let's just stop right there cuz that that's really interesting."

— 13:30

Two directory profiles and some social presence can lift visibility while the site itself ranks for nothing. One listing alone guarantees nothing, but the combination does something the site was not doing.

14:58 - Against nine competitors hiring a search agency, bet on the search agency

Against nine competitors who hire an SEO agency, bet on the SEO agency; ~90% of SEO and AI-visibility work overlaps and you must do everything.

"give me a scenario where an electrician or a plumber or a lawyer isn't on Yelp"

— 14:58

Against nine competitors all hiring a search agency, back the search agency. Around nine-tenths of the work overlaps, and the remaining tenth does not compensate for missing the rest, so everything has to be done anyway.

16:51 - The split is emotional rather than technical

The split is emotional, not technical: buyers of the new discipline were taught to hate SEO — that backlinks are black hat and page rank is mysticism.

"Thank you very much for the kind words. I'm not going to go easy on you."

— 16:51

Buyers of the new discipline were taught to dislike the old one — that links are disreputable and authority is mysticism. The objection is emotional, which is why technical argument does not resolve it.

18:41 - "Search is dead" is an empirical question settled by data

"SEO is dead" is an empirical question settled by data, not emotion; much new-discipline marketing is confident disinformation people lap up.

"it's an empirical question, right?"

— 18:41

Whether search is finished is settled by data, not by conviction. A great deal of the marketing around the new label is confident and wrong, and it is absorbed enthusiastically because it flatters people who never learned the old thing.

19:40 - An agency that ranked for the new label would simply get hired

An SEO agency that ranked for "best AI-visibility agency" would get hired; the only barrier is whether they'll hold their nose and add the tag.

"if somebody like a law a lawyer, a law firm, a venture capital firm, a startup needs"

— 19:40

The only barrier is willingness to wear the tag. An agency that ranked for the new term would be hired on the strength of the ranking, which is a slightly absurd conclusion and appears to be true.

20:44 - Answer optimisation leans far harder on off-page signals and third-party validation

AI answer optimisation is far more off-page signals and third-party validation than SEO; a good SEO ranks with topical authority and few links, then must echo it offline.

"I think they need to from a brand positioning point of view."

— 20:44

It leans harder on off-page signals and third-party validation than search does. A strong practitioner can rank on topical authority with few links, and then has to echo that story offline before an answer engine will repeat it.

22:03 - The case that it is different rests on different competitors, tools and reports

The case that it's different: different competitors, different tracked keywords, different tools, different KPI reports.

"I want to throw some stuff at you, right?"

— 22:03

Different competitors, different tracked terms, different tools, different reports to the client. It is an argument from the shape of the work rather than from its substance, and it is not nothing.

22:37 - The practice barely changed — what changed is asking clients to echo the story elsewhere

He does SEO the exact same way with more AI work; the only real change is asking clients to collaborate with other companies to echo the same story.

"Do I have to answer this honestly? >> Yeah."

— 22:37

The practice is the same with more machine work layered on. What genuinely changed is asking clients to collaborate with other companies so that the same story appears in more than one voice.

24:11 - Claim, frame and prove needs corroboration off the site, not another heading

He uses a "claim, frame, prove" model that needs third-party corroboration off-page — not a new page or H2 — repeating who you are, what you do, why you're brilliant via awards, testimonials and reviews.

"So we use profound peak um radar kit."

— 24:11

Claim, frame, prove — and the proving part happens off the site. Not a new page or another heading, but awards, testimonials and reviews repeating who you are and what you do, in places you do not control.

25:24 - The budget pain is that echoing the site off-site costs more than the retainer

The budget pain: telling an SEO agency on a small monthly retainer to echo the site off-site costs more and may not move rankings, so clients spend that money with a separate agency.

"the other big pain point from when people talk about SEO versus GEO or AEO"

— 25:22

Echoing the site across other properties costs more than the retainer covers and may not move rankings at all. So the client spends that money with a different agency, which is how the separate discipline acquires a budget.

25:55 - The fan-out changes every few days, and the word "best" is fading from it

The query fan-out changes every few days; the word "best" appears far less than before as the engine avoids listicle-bait.

"it's difficult and I think I know I keep coming back to the brand positioning,"

— 25:55

The model is steering away from self-referencing best listicles, which reshapes the keyword set underneath. Best is reported forty-two per cent lower in the fan-out than the prior month — and listicles and comparison content still work; these are brand-new keyword sets rather than dead ones.

28:22 - The real gap is creativity, and listicles are the thing most practitioners forgot

The real gap is creativity: rather than ride a directory's coattails, bring that traffic through your own site — listicles and comparison pages are SEO most SEOs forgot or avoid.

"a creativity issue right so when"

— 28:22

Recreating the comparison on your own property keeps both the traffic and the moat. He outranked a large tool for its own brand plus alternatives and pricing because it cannibalised itself across two pages — a tougher decision, doable, and still the same discipline.

29:44 - Positioning a model as immune to optimisation was a strategic disaster

Models positioned themselves as "SEO proof," a disaster; no how-to guidance from the AI companies, while the search engine published a guide as a clever head start.

"because somebody else is ranking because they've got another page doesn't mean they hold on to it."

— 29:44

Positioning a model as immune to optimisation was a disaster, and it was compounded by publishing no guidance at all — while the search engine published a guide and took a head start out of it.

30:52 - The quiet play is to take the label, rank first, and let rivals argue

The quiet play: add the AI-visibility tag, be a good SEO, rank number one, and get found while rival SEOs argue about not tagging it; the label may be ruined by agencies promising entity-stacking, schema and mass PR that doesn't deliver.

"is it just a creativity issue where someone is saying, 'Well, that's not part of my SEO"

— 30:52

Take the label, do the work well, rank first for it, and collect the enquiries while rival practitioners argue about whether the label is legitimate. The risk is that the label gets ruined by agencies promising things that do not deliver.

32:50 - High-competition tactics move search rankings and do not move the answers

High-competition SEO tactics — redirect techniques, click-through manipulation, social virality spikes — move search rankings but do not move the AI overview or AI-assistant mentions.

"I completely agree with what you're saying there, by the way."

— 32:50

Redirect techniques, click manipulation, engineered social spikes — all of them move search positions and none of them moves an answer or an overview. That is the cleanest available evidence that the two systems are not the same system.

34:33 - The other side of the same niche runs poisoning, argument searches and mass releases

The AI-visibility side of the same niche does data/model poisoning, back-and-forth argument searches, and mass press releases (some with no links) that manipulate AI answers — proof of two different strategies.

"However, the GEO specialists for I gaming are doing data poisoning."

— 34:33

The same niche's visibility specialists run data poisoning, back-and-forth argument searches and mass releases carrying no links at all. Two entirely different tactical repertoires, aimed at two different machines.

35:35 - Conversion work is part of the discipline, and a tactic feeling wrong is not a reason

Conversion — landing-page persuasion and calls to action — is part of SEO; don't let a tactic feeling "black hat" veto it, because emotions driving tactical decisions is the classic error.

"I think it depends again on your definition of SEO"

— 35:35

Persuasion and calls to action are part of the work, and a tactic feeling disreputable is not an argument against it. Letting emotion decide the tactics is the classic error, and it is expensive in both directions.

37:26 - An exact-match domain displaces others and builds a moat

Pull out an exact-match domain to displace others — it saves money and builds a moat; recreate one listicle about yourself, and if the user searches "cheap," you must have "cheap" on the page.

"if I go back to my toolkit and say I can bring out an exact match domain"

— 37:26

Pulling out a matching domain displaces competitors, saves money and builds a moat. And the detail underneath it: if the searcher types cheap, the word cheap has to be on the page.

39:05 - Good visibility work becomes search work by accident

Good AI-visibility work becomes SEO by accident: a founder found after three months all his buying traffic came through SEO; ranking pages, not off-site recommendations, brought the direct buyers.

"Is GIO a framework that's going to fall over?"

— 39:05

A founder discovered after three months that all his buying traffic arrived through search. Ranked pages brought the buyers; the off-site recommendations brought something else. Good work on one produces the other by accident.

40:42 - Only a small fraction of decent practitioners can actually do the additional layers

Only about 1% of decent SEOs can actually do the AI work, because it is SEO with additional layers; many who can't rank rebranded to make it look different.

"It's a nightmare completely."

— 40:42

Something like one in a hundred competent practitioners can actually do the additional layers, because those layers sit on top of the existing craft rather than replacing it. Many of the people who rebranded did so because they could not rank.

43:14 - Clients now hold bigger pots for visibility than for search

On investment boards, clients now hold bigger pots for AI visibility than SEO, spending less on SEO because clicks are down and treating the new work as branding and reputation that also lifts paid-ad returns.

"I'm sat on some of these investment tables"

— 43:14

The spend follows the perceived source of buyers, and buyers are seen shifting to AI answers. He sits on some of these boards, which is where the observation comes from rather than from a survey.

43:42 - An independent adviser who is not pitching for the work is a rare and valuable service

An independent adviser who isn't pitching for the work is a lacking, valuable service; buyers doing that due diligence deserve credit.

"Before we go any further down because I think you've brought"

— 43:42

Somebody advising who is not also bidding for the work is rare, and buyers who go looking for that deserve more credit than they get. The conflict of interest in this market is structural rather than occasional.

44:46 - The vetting criterion is capability with the layer, not the label on the door

Vetting a shortlist of unknown agencies, the criterion is "the best SEOs in the world who understand the AI layer properly," able to work the search engines and third-party sources — not a label.

"let's just say they've the CEO's got some recommendations from LinkedIn, his network."

— 44:46

Vetting an unknown shortlist, the test is capability with the additional layer on top of genuine craft — able to work the engines and the third-party sources — rather than which label the agency has adopted.

46:13 - Buyers narrow to two brands and ask a model to choose

Buyers increasingly narrow to two brands and ask an AI assistant "pick one, don't sit on the fence," letting the model decide — which is why looking good off-page drives profitability.

"I would go to somebody like yourself."

— 46:13

Buyers narrow to two brands and ask a model to pick, explicitly refusing the hedge. That is why looking good off-page drives profitability: the comparison is being adjudicated by something reading your corroboration.

47:49 - Third-party validation is bought from parasite and reputation specialists

For third-party validation he'd hire parasite-SEO and reputation specialists who own the first page with awards, reviews, case studies and "is the brand legit/safe" queries — and AI visibility spikes from that off-page corroboration.

"doing it to third party, I'd be going out to people like yourself,"

— 47:49

Parasite and reputation specialists own the first page with awards, reviews, case studies and the is-this-brand-legitimate queries. Visibility spikes follow that corroboration rather than preceding it.

49:13 - Owners who know nothing hire the wrong people and ask strangers instead of their network

Business owners who know nothing about SEO hire the wrong SEOs, waste months, and ask strangers on a forum instead of using their own network.

"I think there's this this shows the big challenge for business owners, right?"

— 49:13

Owners who know nothing hire badly, lose months, and then ask strangers on a forum rather than their own network. The network is the asset they already have and the one they consistently decline to use.

50:00 - The vetting paradox cuts both ways, and the answer is a good practitioner underneath

The vetting paradox: you need an adviser not so blinkered in SEO they miss the AI layer, yet slick "experts" with no SEO clue won't help — at the heart of a good AI-visibility firm you must find a good SEO.

"you need someone who isn't blinkered so blinkered in SEO that they're blind to what needs to be"

— 50:00

You need an adviser not so buried in search that they miss the new layer, and the fluent ones with no craft underneath are useless. At the heart of any good visibility firm there is a good practitioner, and that is what the vetting has to find.

53:14 - Visibility is pulling budget from the traditional and branding pots

AI visibility is pulling budget from traditional and branding pots; there's no rank tracking, only share of voice; a slice of a large package has become the biggest line.

"because of the KPIs and AI visibility, it seemed as very very different. We're using different tools."

— 53:14

The perceived new source of buyers reallocates the whole marketing budget rather than one line of it. A ten-thousand slice of a two-hundred-thousand package became seventy thousand a month for the AI work — and that work improves search, branding and conversion together, so it ties back in.

55:42 - The fastest tactic is still the press release

The fastest tactic for AI visibility is still press releases; force-index the URLs, and bigger wires can hit "top stories" and occasionally the discover feed.

"the three f press releases is definitely still number one."

— 55:42

It remains the fastest tactic. Force-index the URLs, and the larger wires can reach top stories and occasionally a discovery feed — which is an old technique doing new work.

57:00 - Virality can be manufactured when there is no following to draw on

Manufacture virality if you lack a following: cheap forum, Q&A and social ads, and pop-under traffic through a social-redirect link, to boost the URL.

"viability, when I speak to people like"

— 57:00

Cheap forum, question-and-answer and social advertising, plus pop-under traffic through a social redirect, manufactures the signal a following would otherwise supply. It is a purchase rather than an achievement, and it works the same way.

58:11 - A mailing list moved a page from fifth to first in four hours

Mailing-list virality: an episode at 300–400 views, sent to a large mailing list, drew ~3,500 clicks the search engine picked up and the page jumped from position four or five to number one in three to four hours.

"about viability. If when we go and release a an episode,"

— 58:11

An episode sitting at three or four hundred views was sent to a large mailing list, drew around three and a half thousand clicks, and moved from fifth to first within four hours. The engine responded to the traffic, not to the content.

59:28 - The pop-under mechanic buys the click volume the ranking responds to

Pop-under mechanics: tweet the URL, take the social-redirect link, run a pop-ads network through it, ~6,000 clicks in ~2 hours and the URL jumps; on a news site it pops into top stories, sometimes discover.

"Can you go deeper into that?"

— 59:28

Post the URL, take the social redirect link, run a pop-under network through it, and roughly six thousand clicks arrive in two hours. On a news site that can push into top stories and sometimes into a discovery feed.

1:00:31 - The durable play is to define the entity and then amplify every win

The durable play: define the entity, get a knowledge panel, build reputation (awards, reviews), then amplify every win across press, guest posts, social, force-indexing and virality so crawlers learn who you are.

"temporary boost. But then again, off air, I mean, you've spoken about press releases where long-term just doing press releases isn't a great strategy."

— 1:00:31

A stable, corroborated entity is what a model can retrieve without a live crawl. You have to be in the top twenty or thirty results to be part of the fan-out consensus at all, and most fan-out terms are long-tail.

1:02:04 - The append strategy targets the review site already ranking for the brand

The "append" strategy: append a review site to a page, write to outrank it for "[brand] reviews [site]," reinforce with tier-two links, force-indexing and virality.

"the video that you did on people also ask genius"

— 1:02:04

Intercepting the branded-plus-review query on your own property captures the consensus traffic instead of donating it. Some review sites come back as site: operators, which needs care; the appended ones either get outranked or joined on the first page.

1:04:21 - An alternative for speed is a term half yours and half a competitor's

An alternative for speed and longevity: start with low keyword-difficulty terms you can rank the same day, where half the phrase is yours and half a competitor's, engineered into a post.

"I think that I think those are are are great."

— 1:04:21

Half-yours long-tail phrases are winnable fast and stack into coverage. A difficulty around ten or eleven is rankable the same day against a field scored fifty.

1:05:44 - Owned assets beat rented profiles, because rented profiles get deleted

Build owned assets: buy exact-match domains, grow them, cornerstone; an extra domain mainly needs authority from the parent, and two domains echo in the same fan-outs; own your assets because rented profiles get deleted and deranked.

"The second thing is build up your own assets, right? Because that's the longer play."

— 1:05:44

Buy matching domains, grow them, build the cornerstone structure. An additional domain mainly needs authority from the parent, and two domains echo inside the same fan-outs. Rented profiles get deleted and demoted; owned ones do not.

1:07:16 - A model reads content as shapes rather than as sentences

A model reads content as shapes — it blows a concept into a mathematical form — so giving it a few reference points is enough to produce a whole story.

"the fast ways to do that is to discount all of the and disinformation, right?"

— 1:07:16

Supplying structured reference points lets the model generate the narrative you want rather than the one it would assemble. The first king was X, the second Y, the third me, on a linked second page, produced a full story — offered as how he pictures a model reading content.

1:08:47 - Success comes from failing at twenty pages and not stopping

Success comes from failing at 20 pages and never giving up; the "one thing" differs per search phrase, so play — people are too scared of models and the search engine, and there are no real consequences.

"reference points. And I think that if you look at the success stories that we have,"

— 1:08:45

Twenty pages that failed and no stopping is the actual shape of it. The one thing that works differs by search phrase, so the work is play rather than planning — and people are far more frightened of the consequences than the consequences warrant.

1:10:25 - The one message is to stop making excuses and do the thing

The one message: stop procrastinating, stop making excuses, get up and do the thing.

"if we have like one message from the start of this call,"

— 1:10:25

It is the least technical sentence in the episode and the one both sides of the argument agree on. Everything disputed in the preceding hour turns on method; this turns on whether the method ever gets applied to anything.

1:10:45 - A twelve-hour window was enough to confuse the models and restore the story

The "king of SEO" experiment: a rival's press release claimed the title, and within a ~12-hour window he confused the models and restored the story, using no schema, no chunking, no writing styles.

"Can you talk about what that was for people who don't know?"

— 1:10:45

A rival's release claimed the title, and inside roughly a twelve-hour window the models were confused and the story restored — with no schema, no chunking and no writing-style work. The speed is the finding.

1:12:50 - Fifty pages produce two that rank, and the time belongs to those two

Write 50 pages, find the two that ranked, and pour time into those two, rather than one page that may not rank; getting started is the first step to ranking.

"may as well start doing now."

— 1:12:50

One page may not rank at all, and fifty will produce two that do. Finding those two and pouring the time into them beats choosing in advance, because the choosing cannot be done in advance.

1:14:18 - The genius move is a contextual bridge up the traffic tiers

The genius move is a contextual bridge: rank a low-difficulty keyword, then move up the traffic tiers to the next — classic cornerstoning — and extend it across domains as daisy-chained guest posts.

"I want to expand a little bit on on the king of SEO"

— 1:14:18

Linking ranked pages across domains builds a semantic off-page network that repeats who you are from several directions. His content team runs those networks timed across articles rather than published at once.

1:15:56 - Temporal semantic updating claims the title by acknowledging the old one first

Temporal semantic updating: acknowledge the old entity was the leader, then claim a new one — the AI overview accepts it, where asserting the claim bare would not.

"loved about the king of SEO example right initially"

— 1:15:56

Acknowledge that the old entity was the leader, then claim the new one. The overview accepts the sequence where it would reject the bare assertion, because the sequence gives it a story rather than a contradiction.

1:18:59 - An index can be the whole web or the fraction that answers almost everything

Two ways to build an index: the whole vast index, or just the top 0.1% that answers 99% of questions — models build and refresh tables of that top slice like a ranking table.

"doing a live search? How's that working when they're not doing a live search?"

— 1:18:59

Either index everything, or index the fraction of the web that answers almost every question. Models build and refresh a table of that top slice, which behaves much more like a ranking table than like an archive.

1:21:46 - An open crawl archive may or may not feed training, and the test came back empty

An open web-crawl archive may feed model training; a test: 200 sites blocking every bot but that crawler, seeded with invented entities, asked daily "who is this" — 10 weeks, nothing back.

"Because I've got some interesting things here. I've not got the data on it yet."

— 1:21:46

Two hundred sites blocking every crawler except one, seeded with invented entities and queried daily for ten weeks, produced nothing. It does not settle the question, and it is a real test where most of the discussion is assertion.

1:24:00 - If it does feed training, harmonic centrality is the trust mechanism

If that crawl feeds training data it opens a can of worms: penalised sites still get crawled, and harmonic centrality — distinct from page rank — grows trust from a trusted seed set.

"that's not to say that common crawl isn't being used."

— 1:24:00

Penalised sites still get crawled, which means the penalty does not propagate — and harmonic centrality, which is not the same thing as page authority, grows trust outward from a trusted seed set. That is a different trust model from the one people optimise for.

1:26:39 - Language structure arrives slowly and facts arrive fast

Two data types: language structure (retrieved slowly, so you wait for the next model version) and facts, where a brand repeated enough becomes a noun; the search engine trusts by distance from a trusted seed, layered with anti-spam and precision the crawl lacks.

"here's a couple of things that I think we don't understand"

— 1:26:39

Language structure is retrieved slowly, so you wait for the next model version; facts move fast, and a brand repeated often enough becomes a noun. The search engine layers anti-spam and precision on top of distance from a trusted seed, which an open crawl does not do.

1:31:40 - Off-page consensus can be built cheaply as a reputation tree

Build off-page consensus — a reputation tree — cheaply: press releases, PDF and image sharing blasts of your best reviews, podcast syndication across many platforms, review snippets cited on a forum, force-indexed with virality.

"So build off page consensus build that reputation tree."

— 1:31:40

Press releases, sharing blasts of the best reviews as documents and images, podcast syndication across many platforms, review snippets cited on a forum — all force-indexed and amplified. None of it is expensive and all of it is corroboration.

1:32:35 - Understanding the fan-out means checking the gaps across the surfaces

Understand the query fan-out: two AI assistants show it in their research, and you check the gaps across three search engines — is a video, forum or professional-network result ranking, and if not, get in.

"Understand the query fan out. If you don't understand the query fan out, learn it."

— 1:32:35

The fan-out is the map of sub-queries to compete on, so the gaps in it are the openings. Look across three engines for what is and is not ranking, then fill the gap — once the fan-out is understood, the mystery is gone.

1:33:24 - Great customer service turns happy customers into third-party coverage

Give amazing customer service, then ask happy customers for reviews — they do parasite SEO for you across review platforms, using the right language, and you get both AI visibility and sales.

"Everybody share in the comments who who they think the winner is."

— 1:33:24

Happy customers write reviews across the platforms in the language buyers actually use, which is third-party coverage you did not have to place. The service is the acquisition channel, and it produces the visibility and the sales together.

Personal Branding Lessons

A disagreement between two practitioners who agree on more than they admit. The moves below are the ones both of them would sign.

Own the whole first page rather than one position

Reputation work is a page-level objective rather than a position-level one, and it is priced accordingly. Ranking one site first leaves every other result to somebody else, and the results a buyer reads before deciding are rarely only the first one. Owning the page is the unit of work. 02:31

Do not let a tactic feeling wrong veto it

Persuasion, calls to action and conversion work are all part of the discipline, and a technique feeling disreputable is not an argument about whether it works. Letting emotion decide the tactics is the classic error, and it costs in both directions — the timid avoid what works and the reckless avoid what lasts. 35:35

Buy third-party validation rather than writing another page

Claim, frame, prove — and the proving happens off the site. Not a new page, not another heading: awards, testimonials and reviews repeating who you are in places you do not control. That is the part the answer engines weigh, and it is the part nobody can write for themselves. 24:11

Take the label and rank for it rather than arguing about it

An agency that ranks for the new term gets hired on the strength of the ranking. The only barrier is willingness to wear the tag while rival practitioners argue about whether the tag is legitimate. The argument is free to have and expensive to win. 30:52

Own the assets, because rented profiles get deleted

Buy the matching domains, grow them, build the cornerstone structure around them. An additional domain mainly needs authority from the parent, and two domains echo inside the same fan-outs. A profile on somebody else's platform can be removed or demoted without notice, and everything built on it goes with it. 1:05:44

Write fifty pages and pour the time into the two that ranked

One page may not rank at all. Fifty will produce two that do, and those two are where the effort belongs — after the fact, not before it. The selection cannot be made in advance, which is why the volume is the mechanism rather than the waste. 1:12:50

Acknowledge the old leader before claiming the new one

Asserting a new leader outright gets rejected. Acknowledging that the previous one held the position and then claiming the succession gets accepted, because the sequence gives the machine a story rather than a contradiction. It is a small structural change with a disproportionate effect. 1:15:56

Ask happy customers for reviews

Give service good enough to be worth talking about, then ask. Customers write across the review platforms in the language buyers actually use, which is third-party coverage nobody had to place. The service becomes the acquisition channel, and it produces the visibility and the sales at the same time. 1:33:24

Questions

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

Is getting found in AI a different skill from search optimisation, or the same thing renamed?

The debate never fully resolves it, and that is the honest answer: one side runs it as a separate department because ranking a page and earning an AI citation drew on different strengths in their own team; the other calls it search optimisation with more keywords and an off-site echo. Both agree the overlap is roughly ninety per cent. The split is largely about naming and budgets, not a genuinely new mechanism. 00:34

Why do I rank number one in search but never get cited by AI answers?

Because a citation is earned mostly off your own site. Ranking a page relies on topical authority and a few links; being repeated by a machine relies on third-party validation — reviews, mentions and the same story corroborated across independent sources. On-page work alone cannot produce that consensus. The fix named is to echo your claims off-site, on the review and directory sites the query fan-out actually reads, rather than adding another page or heading to your own domain. 20:44

How do I get an AI assistant to recommend me?

Own more of the first page a machine reads at the moment of choice. Buyers were described narrowing to two options and asking an assistant to pick one, so the brand with better off-page consensus — awards, reviews, case studies, "is this legit" answers all pointing the same way — wins. Build that corroboration deliberately, because the model resolves the final choice from what it can see about you, which is why how audiences and their tools decide who to trust sits underneath the whole thing. 47:49

What is the query fan-out, and why does it matter?

The fan-out is the set of sub-queries a model expands a single prompt into — the real map of what it checks before answering. It matters because it names the exact terms and sources you must appear on, and it shifts every few days; the word "best" was reported far less than the prior month as the engine avoids listicle bait. Two AI assistants surface the fan-out in their research view, and reading it turns AI visibility from mystery into a checklist, much like understanding how a platform's algorithm actually decides what surfaces. 1:32:35

What is the fastest way to get AI visibility?

Press releases, still, syndicated to seed many corroborating sources at once and force-indexed so they are picked up, with the bigger wires occasionally reaching top-stories placement. Where there is no existing following to carry it, the traffic can be manufactured — cheap forum, question-and-answer and social ads, or pop-under traffic routed through a redirect link to surge a URL. Every one of these is flagged as a temporary boost, explicitly not the best long-term strategy. 55:42

What is the most durable way to get AI visibility?

Make yourself a stable, corroborated entity a model can retrieve without a live crawl: define who you are, earn a knowledge panel, build reputation through awards and reviews, then amplify every win across press, guest posts and social with indexing on top. A stated threshold is that a page must sit in the top twenty or thirty results to be part of the fan-out consensus, and most fan-out terms are long-tail, so the durable play is breadth of corroboration rather than one big placement. 1:00:31

Do reviews on third-party sites help me show up in AI answers?

Yes — they were named the single easiest high-leverage move. Happy customers leave reviews across the major platforms, sometimes in the exact language you hoped for, and a machine reads that repeated praise as consensus. A single listing is not enough; two or three independent sources telling the same story is what lifts visibility, and it earns you sales at the same time, since the social proof of many voices agreeing works on people and models alike. 1:33:24

Can I influence what an AI says about me, and is that allowed?

You can, and the episode treats it as fair play with no real consequences for experimenting. A live case saw a claim flip a model's answer within a twelve-hour window using syndicated releases and a temporal framing — acknowledging what came before, then stating what is now. The durable, non-manipulative version is the same mechanism used honestly: get true, corroborating information about yourself onto enough independent sources that the machine reflects it. 1:10:45

How should I vet an AI-visibility or search agency?

Look past the label for a genuinely strong search practitioner underneath, with real case studies rather than one ranked vanity term. The paradox named: an adviser too narrow in search misses the AI layer, but a slick "expert" with no search skill is useless — so the rare overlap is what to hunt for. An independent adviser with no stake in winning the work is called the missing, valuable service, and doing that homework beats asking strangers, much like running a proper look at who you are really up against. 44:46

Sources

GEO vs SEO: James Dooley Says Only 1% of SEOs Can Actually Do It

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.