I get asked "what's your process" on almost every discovery call, and for years I gave some version of a verbal answer that changed slightly every time depending on my mood and how much coffee I'd had. Eventually I realized that inconsistency was itself a problem, if the process is genuinely repeatable, it should be documented the same way every time. So here it is, written down properly, for the first time.
Why a Fixed Sequence, Instead of "It Depends on the Client"
I want to address the obvious objection first, because I used to make this argument myself before I changed my mind. Every client is different, so shouldn't the process be different too?
The content and priority within each phase absolutely adjusts to the client. What doesn't change is the order. Here's why: you cannot build a topical map (Phase 2) before you know what's technically broken and what a competitor's real entity coverage looks like (Phase 1). You cannot execute content (Phase 4) on top of a technical foundation (Phase 3) that hasn't actually been fixed yet, you'll just be writing excellent content that AI crawlers can't reliably reach. And you cannot report meaningfully (Phase 5) on results from work that skipped verification at any earlier stage, because you won't actually know which variable moved the number.
The sequence isn't rigidity for its own sake. It's what prevents the most expensive kind of mistake in this work: doing real, skilled, expensive labor on top of a foundation that quietly undermines it.
Phase 01: Discovery, Deep Diagnostic Audit
This phase answers one question honestly: what is actually happening on this site right now, not what anyone assumes is happening.
It starts with a full Screaming Frog crawl, checking crawlability, indexation status, and whether JavaScript-rendered content is actually visible to search engines and AI crawlers, two different questions that get conflated constantly. Alongside that runs an entity gap analysis against real, currently-ranking competitors, not a generic checklist, what specific entities, topics, and concepts do they cover that this site doesn't. Core Web Vitals get baselined properly, LCP, CLS, INP, not just glanced at. And the existing topical coverage gets mapped against what actual coverage would need to look like, which almost always reveals the gap is larger than anyone on the client side expected.
I've run this exact discovery process more than 350 times now, and the single most common finding, more common than any technical bug, is a site with real content investment already made, sitting disconnected from how it should be internally linked and structurally organized. The content exists. Nobody connected it properly.
Phase 02: Architecture, Strategy & Topical Mapping
Once discovery shows what's actually there and what's missing, architecture is where the plan for closing that gap gets built, before a single new page gets written.
This means a full topical authority map, the pillar-cluster structure that determines what gets built and in what order. A keyword and query-intent matrix spanning awareness through decision through conversion, not just a list of terms, a map of where each one sits in the actual buyer journey. URL taxonomy decisions, which sound boring and matter enormously, a badly structured URL hierarchy creates internal linking problems that compound for years. And the internal link architecture itself, planned deliberately for how authority should flow through the site, not left to whatever order content happens to get published in.
I think this is the phase most commonly skipped or rushed by agencies under deadline pressure, and it's the one that determines whether Phase 4 execution compounds into real topical authority or just becomes a pile of disconnected blog posts.
Phase 03: Foundation, Technical Infrastructure
Foundation makes sure the technical ground is actually solid before content gets built on top of it, closing the loop on whatever Discovery found broken.
JSON-LD schema gets implemented properly, Person, Organization, Service, FAQPage, BreadcrumbList, and Article schema, with real @id cross-referencing rather than disconnected, isolated markup. Any JavaScript rendering issues found in Discovery get fixed at the architecture level. Canonical tags, hreflang where relevant, pagination handling, and duplicate content get cleaned up systematically rather than patched one page at a time. Core Web Vitals get actually optimized, not just measured. And crawl directives, robots.txt, sitemap structure, get configured deliberately for both traditional search engines and the AI crawlers covered in how AI crawlers actually work.
This is the least glamorous phase and also, in my experience, the one where the most real, measurable improvement gets made per hour of work invested.
Phase 04: Execution, Content & On-Page Work
This is where most of what a client actually sees happens, but only after the first three phases have made it worth doing.
Content gets written entity-first, structured for both classic SEO relevance and the answer-first, citation-friendly format that AI Overviews and generative engines actually favor, not written for keyword density the way it might have been built five years ago. On-page optimization gets applied systematically across the priority list built in Architecture. Schema gets applied per content type, not as one generic template stretched across every page. Off-page and citation outreach happens where it genuinely strengthens entity trust, not as volume for its own sake. And GEO-specific optimization, the entity and structured-data work aimed specifically at generative AI citation, gets layered in throughout, not bolted on as an afterthought once everything else is done.
Phase 05: Intelligence, Reporting & Continuous Refinement
The phase that turns a one-time project into something that actually compounds, and the one that gets cut first when budgets tighten, usually to the client's real cost later.
A live dashboard connecting GA4 and Search Console data gives a real, current view rather than a monthly snapshot built by hand. Monthly strategy reviews look honestly at what's working and what isn't, not just what looks good in a report. Competitor intelligence keeps tracking, because the gap analysis from Phase 1 isn't a one-time snapshot, competitors keep publishing too. Algorithm and AI-model changes get monitored and translated into what they actually mean for the specific client, not generic industry commentary. And ROI gets attributed all the way through, from organic sessions to leads to actual revenue, the number that a client's finance team, not just their marketing team, will believe.
The 5-Phase AEO & GEO Methodology (The Actual Process Behind Every Engagement)
What Happens When a Client Only Wants One Phase
This comes up often enough that it's worth addressing directly. Sometimes a client genuinely only needs one piece, a technical audit alone, or a topical map alone, without the full engagement. That's fine, and I'll scope it that way honestly when it's the right call.
What I won't do is pretend that a single phase delivered in isolation produces the same result as all five run in sequence. If Discovery reveals serious technical debt and the client only wants Execution (new content), I'll say so plainly: the content will likely underperform relative to its quality, because the foundation it's sitting on wasn't fixed. That honesty costs me some short-term revenue occasionally. It's also the only way this methodology stays worth documenting publicly, if it only works when I don't mention the exceptions, it isn't actually a methodology.
Frequently Asked Questions
Do all five phases happen on every engagement, even small ones? Direct answer: The depth scales with scope, a smaller engagement moves through all five phases faster and lighter, but skipping a phase entirely, rather than compressing it, is where quality problems start. How long does the full five-phase process typically take? Direct answer: Discovery through Foundation typically runs around two months for a standard engagement, with Execution and Intelligence continuing on an ongoing retainer basis afterward, since AI-search visibility isn't a one-time achievement. Can this methodology be applied to a site that already has strong traditional SEO? Direct answer: Yes, and it often should be. Strong traditional SEO doesn't automatically translate into AI-search citation eligibility, Discovery frequently reveals real entity and structural gaps even on technically excellent, well-ranking sites. What makes this different from a generic SEO audit-then-execute process? Direct answer: The explicit GEO and AI-crawler-accessibility layer built into Phases 1, 3, and 4 specifically, most traditional audit processes were built before generative AI search existed and don't natively account for entity signals or AI crawler behavior. Is Phase 5 (Intelligence) really necessary, or is it just ongoing reporting? Direct answer: It's more than reporting. It's the phase where competitor and algorithm changes get caught and responded to before they cost meaningful visibility, treating it as optional reporting rather than active monitoring is exactly how sites quietly lose ground between larger review cycles.This is the exact process every white-label engagement runs through, whether you're evaluating me as a partner directly or just trying to understand what good AEO/GEO work actually requires. See how this applies to your specific situation.