Five Layers of SEO. One Percent of the Traffic.

The diagram is right about the work

Let me be fair to it first, because the framework is useful and I would rather improve it than dunk on it.

It captures something real. There are now multiple distinct surfaces where a brand can be visible, they have different mechanics, and they are measured in different places. A team that thinks only about blue-link rankings in 2026 is missing most of the board. Laying that out as layers, each with a goal, a focus and a success metric, is a clearer teaching tool than most of what the industry produces.

The problem is not the layers. It is the equal signs between them, and the implication that each one is a separate practice needing its own budget line, its own specialist and its own retainer.

That implication is where clients lose money. So let us take the layers one at a time and check each against what the platforms and the published data actually say.

Layer one: SEO is not a layer, it is the floor

The diagram puts SEO at the top as the first of five. In practice it is underneath all four of the others, and Google states the dependency explicitly.

To appear in Google's generative AI features, a page must be indexed and eligible to be shown with a snippet, meeting the standard Search technical requirements, and the site must not have opted out of AI features. That is not a philosophical point about foundations. It is a hard gate. A page that fails technical SEO cannot be an answer, cannot be cited, and cannot convert anyone, because it will not be retrieved in the first place.

Google explains the retrieval mechanics in the same document. Its AI features run on retrieval-augmented generation, also called grounding, which "relies on our core Search ranking systems to retrieve relevant, up-to-date web pages from our Search index," plus query fan-out, where the model issues several related queries to gather more results.

Read that sentence twice if you sell GEO services. The generative layer does not have its own index, its own crawler or its own ranking logic on Google. It reaches into the same index your technical SEO work feeds. Which means the single highest-leverage action for AI visibility on Google is the least glamorous one available: make sure your important pages are crawlable, indexed and not accidentally excluded. Our technical SEO checklist covers exactly that ground.

What this means for the diagram. Redraw it. SEO is not the first of five bands. It is the slab the other four sit on. Every dollar spent on the upper layers is contingent on it, and no amount of AI-specific tactics compensates for a page that is not indexed.

Layer two: AEO is real, and it is twenty years old

Answer engine optimization is a genuine discipline, and it long predates the acronym. Its lineage runs through featured snippets, the position-zero chase and voice search. The unit of success shifts from being one result in a list to being the answer.

The data supports treating this as important. SparkToro research using Similarweb clickstream data, reported by Search Engine Land, found US Google searches ended without a click 68.01 percent of the time in the first four months of 2026, up from 60.45 percent in 2024. Worth carrying the caveat: that count includes clicks to Google-owned properties like Maps and YouTube but excludes follow-up searches, and SparkToro notes its data providers have changed over the years, so the long-term line is not a like-for-like comparison.

Still, roughly two thirds of searches ending without a click is the strongest argument in the whole framework for optimizing to be the answer rather than merely a result.

The intent mix is shifting underneath that number too. Search Engine Journal's review of two studies on how AI is reshaping search intent is worth reading alongside the zero-click figure, because the queries that survive as clicks are not a random sample of the queries that used to. Informational lookups are the ones being absorbed. Commercial and navigational intent is holding up better. That distinction matters far more for planning than the headline percentage, and almost nobody segments for it.

What this means for the diagram. AEO earns its place, with one correction. The diagram lists "zero-click visibility" as a success metric. Zero-click is a state of the market, not something you can raise or lower. Measure appearances in answer features and branded search lift instead, which are things your work can actually move.

Layer three: GEO is real, and Google says it is still SEO

Generative engine optimization has a precise origin. It comes from a 2023 paper by Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan and Deshpande, later accepted to KDD 2024, which framed the problem honestly and reported that its methods could boost visibility by up to 40 percent. That figure is an upper bound measured on a purpose-built benchmark, not an average on live Google, and the authors themselves noted efficacy varies by domain. If you want the fuller background, start with what GEO actually is.

Here is where the naming gets interesting, because the two biggest players disagree in public.

From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.

That is Google's guide addressing the acronyms by name. Microsoft went the other way, launching AI Performance in Bing Webmaster Tools in February 2026 and describing it as "an early step toward Generative Engine Optimization (GEO) tooling."

Both are telling the truth about their own systems, which is why the taxonomy argument never resolves. I worked through that tension at length in our piece on AEO versus GEO versus SEO.

What this means for the diagram. Keep the layer, drop the idea that it is a separate budget. The measurement is separate, because Google's generative AI performance report gives impressions without clicks or queries while Bing's gives citations with sampled grounding queries. The work underneath is the same work.

Layer four: AIO is the one that does not survive

This is the layer I would delete.

As usually described, AI optimization means making content machine readable: semantic entities, Knowledge Graph signals, structured data, consistent brand information, multimedia, freshness. Compare that against GEO in the layer above and the overlap is close to total. Two labels, one activity.

Worse, several tactics commonly filed here are contradicted by Google directly. Its mythbusting section takes them one at a time.

Google still recommends structured data for rich result eligibility in classic search, which is a real reason to keep it, just not the reason usually given. We covered where that leaves structured data in AI search separately.

One part of the layer does hold up. Multimodal retrieval is changing what counts as optimizable content, and Search Engine Land's reporting that images have a new job in AI search reflects a real shift. Entity consistency also matters, because a model that cannot reliably resolve who you are will struggle to recommend you.

What this means for the diagram. Fold the surviving parts into GEO and drop the acronym. If a layer's defensible content is a subset of the layer above it, it is not a layer.

Layer five: SXO promises revenue the evidence cannot yet confirm

The bottom of the funnel is the most confident claim in the whole diagram: engagement, leads, sales, revenue. It is also where the published research is in open disagreement.

On one side, Visibility Labs analyzed twelve months of GA4 data across 94 ecommerce sites and found ChatGPT traffic converted at 1.81 percent against 1.39 percent for non-branded organic, a 31 percent advantage that held in ten of twelve months. Revenue per session was higher too, at 3.65 dollars against 3.30.

On the other, an analysis of 973 ecommerce sites covering 20 billion dollars in revenue found ChatGPT referrals converting worse than Google search, email and affiliate on both conversion rate and revenue per session. Other datasets have reported AI traffic converting at 18 to 20 percent and outperforming every other channel.

Search Engine Land has written the fairest summary of this mess, noting that every AI search study tells a different story and that each organization publishing the research has business interests that may influence the framing, if not the methodology.

Now the number that reframes everything. In that same 94-site study, ChatGPT drove 474,000 dollars against 32.1 million dollars from non-branded organic. That is 1.48 percent of organic revenue, rising to 2.2 percent in the second half of the year. Average order value was lower, at 204 dollars against 238. And AI referral traffic overall sits at roughly one percent of total web traffic, with ChatGPT accounting for around 87 percent of what does arrive.

There is also an attribution gap running the other way, which honesty requires mentioning. Many people get a recommendation in ChatGPT, then search the brand on Google and buy there, and that conversion is credited to branded organic. AI influence is almost certainly larger than AI referrals. It is just not measurable with last-click data, which is why Search Engine Land's coverage of how AI search traffic differs from organic is worth reading before anyone builds a forecast on it.

What this means for the diagram. SXO is the most durable layer of the five, because conversion work pays off no matter which surface sent the visit. We make that case in what search experience optimization actually is. But the revenue promise should be stated as a range with a caveat, not a checkmark. Google's page experience guidance remains the practical baseline here.

The success metrics column is where it really breaks

Look down the right-hand side of any version of this diagram and you will find metrics of wildly different quality presented in identical boxes.

Rankings, organic traffic and indexed pages are measurable, with caveats. AI Overview appearances and featured snippets are measurable. AI citations are measurable on Bing and partially measurable on Google. Then it degrades: AI mentions and brand references depend on which third-party tool you buy. And "AI understanding," "entity recognition" and "better AI visibility" are not metrics at all. They are aspirations with checkmarks next to them.

That matters more than it sounds. A framework that presents unmeasurable outcomes in the same visual language as measurable ones teaches people to accept vibes as reporting. It is also precisely how retainers get sold for work nobody can evaluate.

A better rule: if you cannot say where a number comes from and what would make it go down, it does not belong in the metrics column.

Apply that test to the five layers and the picture gets clearer fast.

Three passes, one half-pass, one fail. That is a more honest diagram than five identical checkmark columns. It is also the same test we apply when building reporting for executives.

Putting this to work

The layered model deserves a better version of itself. Same five ideas, drawn to scale, with the foundation underneath where it belongs and honest labels on the metrics. That diagram would be useful. The one being shared is a good teaching aid that quietly argues for a budget the evidence does not yet support.