Skip to main content Scroll Top

AEO Experts: The Six Skills, and the Two That Are Scarce

Updated September 2026 · Written and maintained by the Progression Agency strategy team

Three of the six transfer straight from technical SEO. The other three are where capability varies — and four questions surface the difference in ten minutes.

On this page · 11 sections
  1. What an AEO expert actually needs to be good at
  2. How to tell practice from vocabulary
  3. Warning signs
  4. Why the field is this uneven
  5. In-house, specialist, or your existing team
  6. What to expect in the first week
  7. Everything else we have written on search, AI and getting found
  8. The six skills, and where each one comes from
  9. The four questions, and how to read the answers
  10. Hiring configurations
  11. What the first week should produce

The short answerAnswer engine optimization needs six skills: server-log fluency, rendering diagnosis, information architecture, measurement design, editorial judgement at sentence level, and the discipline to state what cannot be known. The first three transfer directly from strong technical SEO. The last three are scarce, and they are why capability varies so widely behind identical vocabulary. Four questions separate practice from vocabulary, and the most informative is the last: what can you not tell us, even in principle.

The discipline is roughly three years old. Claims of a decade of AEO experience describe adjacent work, which is valuable and should be described accurately.

The skill mix AEO actually requiresThe skill mix AEO actually requires
The distinguishing bar is measurement design — the ability to build a falsifiable prompt set — and almost nobody arrives with it, because the discipline is three years old.

What an AEO expert actually needs to be good at

Six skills, and only three of them exist in the market in quantity. Server-log fluency, rendering diagnosis, information architecture, measurement design, editorial judgement at sentence level, and the discipline to state clearly what cannot be known.

The first three transfer directly from strong technical SEO. The last three are where the supply narrows sharply, and they are the reason capability varies so widely behind identical job titles and identical vocabulary.

The distinguishing skill is measurement design — building a prompt set that survives scrutiny. Almost nobody arrives with it, because there was no reason to have it three years ago.

Server-log fluency — Skill. The evidence base for crawler access.
Rendering diagnosis — Skill. Reading what a machine receives.
Information architecture — Skill. Templates, not pages.
Measurement design — Skill. Prompt sets that survive scrutiny.
Editorial judgement — Skill. What makes a passage quotable.
Epistemic discipline — Skill. Stating what cannot be known.
Where the supply narrowsWhere the supply narrows
Two skills are scarce and both are about rigour rather than technique. That is the practical reason capability varies so widely behind the same job titles.

Server-log fluency

Crawler access claims are unprovable without it. Less universal even among technical SEOs than you would expect, and it is the entire evidence base for whether AI crawlers actually reach a site.

Rendering diagnosis

Knowing what a scriptless fetch means, why it differs from what a browser shows, and what fixing it actually involves in a given stack.

Information architecture

Thinking in templates rather than pages. This is what makes a hundred-thousand-page site a four-template job rather than an impossible one.

Measurement design

Building a fixed, versioned prompt set with a stated aggregation rule, and storing verbatim answers so results can be audited. The rare skill.

Editorial judgement

Knowing at sentence level what makes a passage quotable. Technical practitioners frequently lack this; editorial people frequently lack the first three.

Epistemic discipline

Knowing and saying what cannot be known. Nobody outside the labs knows ranking weights, and a practitioner who is precise about that boundary is signalling more than one who is confident everywhere.

How to tell practice from vocabulary

Four questions. Which crawlers will you test for by name. Will you fetch our templates with JavaScript disabled before quoting. How will you aggregate multiple runs of the same prompt. And what can you not tell us, even in principle.

The fourth is the one most candidates have never been asked, and the answers to it are the most informative thing you will hear. A practitioner has a ready, specific answer. Someone working from vocabulary usually does not.

Four questions that separate practice from vocabularyFour questions that separate practice from vocabulary
The fourth question is the one most candidates have never been asked, and the answers to it are the most informative thing you will hear in an interview.
Which crawlers by name — Ask. Specificity is hard to fake.
Render test before quoting — Ask. Turns a guess into a scope.
Aggregation rule — Ask. Decided before results.
What cannot be known — Ask. Most informative answer you get.
Citations or mentions — Ask. Do they distinguish them.
Who owns measurement after — Ask. It should be you.
Signals of genuine AEO expertiseSignals of genuine AEO expertise
Six positive signals, four disqualifying ones. The sixth is the strongest single indicator: real practitioners are precise about the boundary of what anyone can know.
Names the crawlers — Good sign. Unprompted, individually.
Asks for logs first — Good sign. Before offering an opinion.
Tests rendering early — Good sign. Before discussing content.
States an aggregation rule — Good sign. Before seeing results.
Separates citation and mention — Good sign. Different things.
Names the limits — Good sign. The strongest signal of all.

Warning signs

Confident numbers for internal ranking weights. An explanation of why a model chose a competitor. Leading with llms.txt. A citation guarantee. Rank data presented as proof of AI visibility. And ten years of AEO experience in a three-year-old field.

None of these means someone is dishonest. Most of the time it means they are repeating what they have read rather than what they have tested — which is exactly what the four questions are designed to surface.

Confident ranking weights — Warning. Published by nobody.
Explains model choice — Warning. Not exposed by any provider.
Leads with llms.txt — Warning. A checklist, not a method.
Guarantees citations — Warning. Describes a non-existent product.
Rank data as AI proof — Warning. Wrong surface entirely.
Ten years' AEO experience — Warning. The field is three years old.

Why the field is this uneven

The discipline is three years old and the vocabulary spread considerably faster than the practice. Demand arrived in early 2025 when AI Overviews rolled out; supply is still catching up, and a CV cannot yet carry the weight of distinguishing one from the other.

That is why assessment has to be by method. In an established field you can read a portfolio. In a three-year-old one, how someone approaches the first week tells you more than where they have worked.

Why the field is this unevenWhy the field is this uneven
This is why the four questions matter more than a portfolio. In a three-year-old field, method is a better signal than history.
Transfers — Crawl and index knowledge. Directly.
Transfers — Site architecture. Directly.
Transfers — Speed and Core Web Vitals. Directly.
Needs adding — Log analysis for AI agents. Specific user agents.
Needs adding — Scriptless render testing. New standard.
Needs adding — Prompt-set measurement. Entirely new discipline.

What transfers from technical SEO

Crawl and index knowledge, site architecture, page speed, and the general habit of diagnosing before prescribing. This is most of the foundation.

What has to be added

Log analysis filtered to AI user agents, scriptless render testing as a standard practice, and prompt-set measurement — which is an entirely new discipline rather than an extension of an old one.

Why ‘ten years of AEO’ is a warning

Nobody has it. Adjacent experience is genuinely valuable and should be described as adjacent experience, and someone who describes it accurately is telling you something useful about how they handle the rest of their claims.

In-house, specialist, or your existing team

Developer capacity is the axis that decides most of this. Without it, the rendering half stalls regardless of who you hire, which makes it the first thing to establish rather than the last.

Where AI visibility is central to growth and you have developers, a specialist working alongside them moves fastest. Where it is peripheral and you have developers, upskilling your existing SEO lead is often the better economics — the foundation already transfers.

In-house, specialist, or your existing teamIn-house, specialist, or your existing team
Developer capacity is the axis that decides most of this. Without it, the rendering half stalls regardless of who you hire.
Hire a specialist — Central to growth. And you have developers.
Project-based — Central, no devs. Rendering will be the constraint.
Upskill in-house — Peripheral, devs available. Your SEO lead can learn this.
Advisory — Peripheral, no devs. Direction without delivery.
Either way — Own the prompt set. It must survive the engagement.
Either way — Name an internal owner. Measurement decays without one.

Central to growth, developers available

A specialist alongside your engineers. This is the fastest configuration and the one where the work actually ships.

Central to growth, no developers

A specialist on a project basis, with rendering scoped as an engineering procurement rather than an agency deliverable. Name that constraint early.

Peripheral, developers available

Upskill your existing technical SEO lead. Three of the six skills are already there, and the other three are learnable with a good brief.

Peripheral, no developers

Advisory. Direction without delivery, revisited when the picture changes.

Whatever you choose

Own the prompt set internally and name a person responsible for it. Measurement that belongs to a supplier leaves when the supplier does, and the baseline goes with it.

What to expect in the first week

A practitioner spends it on evidence rather than on recommendations. Logs pulled and filtered for AI user agents. Commercial templates fetched twice, scripted and scriptless. A prompt set drafted in your buyers’ language and run before anything changes.

Recommendations that arrive before that evidence are generic by definition, because nothing site-specific has been established yet.

How the expertise is actually builtHow the expertise is actually built
Steps one to three exist in the market in quantity. Steps four to six are where the supply narrows sharply, which is why capability varies so much behind identical vocabulary.
AEO expertise, in numbersAEO expertise, in numbers
Nobody has ten years in this. Anyone claiming to is describing adjacent experience, which is fine — but it should be described accurately.

Put the four questions to us

We will name the crawlers we test, fetch your templates with JavaScript disabled before quoting, state our aggregation rule up front, and tell you exactly which parts of this nobody can know.

/ai-visibility-audit

Everything else we have written on search, AI and getting found

AI, AEO and what is changing

Websites and design

Choosing and working with an agency

Social, content and brand

By industry and by situation

Want this done for your site?We build and maintain the search, content and paid programmes described on this page.

Get a free proposal

The six skills, and where each one comes from

Six distinct skills an AEO engagement needs, and the background each usually arrives from.

What transfers, what has to be added, and how common each is
SkillSourceMarket availabilityWhy it matters here
Crawl and index knowledgeTechnical SEOAbundantThe foundation retrieval depends on
Site architectureTechnical SEOAbundantTemplates are the unit of work
Page speed and vitalsTechnical SEOAbundantAffects crawl and indexation
Server-log analysis for AI agentsPartly transfersLess commonThe only proof of crawler access
Scriptless render diagnosisPartly transfersModerateThe most expensive failure mode
Measurement designNewScarceWhether any result can be audited
Editorial judgement at sentence levelNew in this applicationModerateWhat makes a passage quotable
Stated epistemic limitsNewScarcestSeparates tested knowledge from repeated reading

The four questions, and how to read the answers

Four interview questions, with what a strong and a weak answer sound like.

What a strong answer sounds like
QuestionWeak answerStrong answerWhat it reveals
Which crawlers will you test for?All the major AI botsGPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-ExtendedWhether they have read a robots.txt recently
Render test before quoting?We will review the siteYes — send three URLs, we will return scripted vs scriptless word countsMethod versus intention
How will you aggregate runs?Our data is accurateThree to five runs per prompt, median, decided before we lookWhether measurement is falsifiable
What can you not tell us?We can find out anythingRanking weights and why a specific source was chosen. Nobody canTested knowledge versus repeated reading

Hiring configurations

In-house, agency, freelance and hybrid compared on cost, speed and what each is bad at.

Developer capacity decides most of this
Your situationBest configurationMain constraintWhat to insist on
Central to growth, devs availableSpecialist plus your engineersCoordinationPrompt set owned internally
Central to growth, no devsSpecialist, project-basedRendering procurementRendering scoped separately, early
Peripheral, devs availableUpskill your SEO leadTime to learnA written method, not just training
Peripheral, no devsAdvisoryDelivery capacityA dated baseline you keep
Enterprise, anySpecialist plus internal ownerApproval pathsGovernance mapped in week zero

What the first week should produce

Concrete outputs you should expect within a week, whoever you hire.

Evidence before recommendations
DeliverableWhy firstWhat it rules in or out
AI crawler log extractOnly proof access is realWhether anything downstream can work
Scriptless render report per templateLargest cost variableWhether this is a content or engineering problem
Status codes by AI user agentInvisible in a browserCDN and WAF blocking
Dated prompt baselineBefore anything changesWhether later claims are provable
Template inventoryThe real unit of workHow large the job actually is
Stated limitsSets expectations honestlyWhat no supplier can deliver

Getting found in search

AI, AEO and what is changing

Paid media and lead generation

Websites and design

Choosing and working with an agency

Social, content and brand

By industry and by situation

Frequently asked questions

What makes someone an AEO expert?
Six skills: server-log fluency, rendering diagnosis, information architecture, measurement design, editorial judgement at sentence level, and the discipline to state what cannot be known. The first three transfer from strong technical SEO; the last three are where the supply narrows.
How do I tell a real AEO practitioner from someone using the vocabulary?
Four questions: which crawlers will you test for by name, will you fetch our templates with JavaScript disabled before quoting, how will you aggregate multiple runs of the same prompt, and what can you not tell us even in principle.
What is the most informative question to ask?
What they cannot tell you, even in principle. A practitioner has a ready, specific answer — ranking weights and model reasoning are published by nobody. Someone working from vocabulary usually does not.
Is ten years of AEO experience possible?
No. The discipline is roughly three years old. Adjacent experience in technical SEO is genuinely valuable, and someone who describes it accurately is telling you something useful about how they handle their other claims.
What transfers from technical SEO?
Crawl and index knowledge, site architecture, page speed and the habit of diagnosing before prescribing. That is most of the foundation.
What has to be added?
Log analysis filtered to AI user agents, scriptless render testing as standard practice, and prompt-set measurement — which is a new discipline rather than an extension of an existing one.
Which skill is genuinely scarce?
Measurement design — building a fixed, versioned prompt set with a stated aggregation rule and verbatim answer storage. Very few people arrive with it because there was no reason to have it three years ago.
Should I hire in-house or use a specialist?
Developer capacity decides most of it. Where AI visibility is central to growth and you have developers, a specialist alongside them moves fastest. Where it is peripheral and you have developers, upskilling your existing SEO lead is often better economics.
Can my existing SEO lead learn this?
Often yes, if they are strong technically. Three of the six skills are already there. The gap is usually measurement design and the habit of stating limits, and both are learnable with a clear method.
What if I have no developer capacity?
Say so early and scope rendering as an engineering procurement rather than an agency deliverable. Without it, the rendering half stalls regardless of who you hire.
What should an AEO expert deliver in week one?
Evidence rather than recommendations: an AI crawler log extract, a scriptless render report per template, status codes by user agent, a dated prompt baseline and a template inventory.
Why are recommendations in week one a warning sign?
Because nothing site-specific has been established yet. Advice that arrives before the evidence is generic by definition.
Should the agency own the measurement?
No. Own the prompt set internally and name a person responsible for it. Measurement that belongs to a supplier leaves when the supplier does, and the baseline goes with it.
What are the clearest warning signs?
Confident numbers for internal ranking weights, an explanation of why a model chose a competitor, leading with llms.txt, a citation guarantee, and rank data presented as proof of AI visibility.
Does a citation guarantee ever make sense?
No. Attribution happens inside the model, there is no ad inventory and no submission process. A guarantee describes something that does not exist.
Why does the field vary so much in capability?
Because the vocabulary spread far faster than the practice. Demand arrived when AI Overviews rolled out; supply is still catching up, and a CV cannot yet distinguish one from the other.
Is a portfolio useful for assessing AEO expertise?
Less than usual. In a three-year-old field, how someone approaches the first week is a better signal than where they have worked.
What is epistemic discipline and why does it matter?
Knowing and saying what cannot be known. Nobody outside the labs knows ranking weights or why a specific source was selected. A practitioner precise about that boundary is signalling that their other claims have been tested.
Do I need a specialist or a full-service agency?
Depends on coordination needs. A specialist has deeper practice; a full-service agency coordinates with content, PR and engineering. Both work if they answer the four questions well.
What should I insist on regardless of who I hire?
A dated prompt baseline recorded before anything changes, owned by you, with verbatim answers stored. It is the one deliverable that makes everything afterwards provable.
How long before a good practitioner shows results?
Crawler access shows in logs within days and rendering within about two weeks. Citation frequency takes months. A practitioner will tell you that up front rather than after the first invoice.
What does an AI visibility consultant cost?
Published hourly rates run $50-$200 for specialists and $200-$400+ for expert tier, with one-off audits published at $1,000-$4,000. Our pricing page covers the bands with sources.

Want this done for your site?We build and maintain the search, content and paid programmes described on this page.

Get a free proposal

Get a free marketing proposal

Tell us what you are trying to grow and we will come back with a plan, not a pitch deck. Same-day reply on weekdays.

Our website uses cookies from third party services to improve your browsing experience. Read more about this and how you can control cookies by clicking "Privacy Preferences".
Privacy Preferences
When you visit our website, it may store information through your browser from specific services, usually in form of cookies. Here you can change your privacy preferences. Please note that blocking some types of cookies may impact your experience on our website and the services we offer.
Privacy Policy
You have read and agreed to our privacy policy
Required
Contact Us