The AEO Report

Vol. 1 · Issue 1

No. 05 · Editorial response

AEO is just SEO. Mostly.

Here's the 20% that isn't — and how we're going to prove it.

The AEO Report Editorial Published 2026-05-19

Jake Zward published a LinkedIn post last week with a clarifying argument: AEO is, mostly, just SEO done well. Schema isn't parsed. llms.txt has zero measurable impact. Every major LLM either retrieves from Google or was trained on data weighted by the same authority signals Google uses. The takeaway: "You don't need a new strategy every time there's a new acronym."

The post hit 98,000 views. It deserves a serious response — because it's mostly correct, and what's missing is the 20% that explains why a publication like this one exists at all.

Here's the honest map.

Where Jake is right

And the AEO industry has been getting away with overselling.

The dominant story sold by AEO vendors is that AI search is a different discipline with different rules, requiring different tooling, often priced at $2,000–$5,000/month. Most of that is overstated.

On retrieval

It's correct that ChatGPT's search step retrieves from Bing's index. It's correct that Perplexity's index is heavily seeded with Bing and Google results. It's correct that Gemini surfaces AI Overviews from Google's main index. The retrieval layer of every consumer LLM is, in 2026, still a search engine — and the search engines still rank pages on the signals SEOs have been optimizing for since 1998.

On authority signals

It's correct that getting cited by an LLM correlates very tightly with three things — getting mentioned on authoritative sources (the PR layer), ranking well in traditional search (the SEO layer), and being a recognizable entity (the knowledge-graph layer). All three of those are pre-existing disciplines with pre-existing playbooks.

On the vendor markup

It's correct that the "AEO tool" category is, at the moment, mostly rebranding traditional SEO and brand-mention tracking with a 3–10× pricing premium and a fresh dashboard. Some of those tools have genuinely useful prompt-coverage data. Most of what they sell as "AEO insights" is what every competent SEO platform has been surfacing for a decade.

Jake is right. The 80% of the work that moves AEO visibility is the same 80% of the work that moves SEO visibility. Anyone selling you a wholly separate discipline is, mostly, selling you a wrapper.

Where Jake is wrong

Or at least making claims that aren't yet proven.

Two specific claims in the post are stated as fact, without cited evidence, and are empirically testable.

Claim 1

"Schema isn't parsed."

This is too strong. We know with high confidence that:

  • →Bing's documentation explicitly states that schema markup influences indexation behavior and rich result eligibility, which feeds the ChatGPT retrieval layer
  • →Perplexity's citation cards visibly use Article and Organization schema metadata for source labels, author attribution, and publication dating (look at any cited source on a Perplexity result page — that's structured data being rendered)
  • →Google's AI Overviews surface FAQPage and HowTo schema content directly into answer text in observable cases

What's true is that schema doesn't move rankings in the SEO sense — it doesn't change what's first or third. What it does is help engines parse and present content cleanly, which influences whether you're cited at all, how you're labeled, and what's surfaced from your page.

"Schema isn't parsed" conflates two questions: does it move rankings (mostly no) and does the engine read it (verifiably yes). Conflating them produces a sound bite that's nearly the opposite of the truth.

Claim 2

"llms.txt has 0 measurable impact."

This is asserted without evidence. The honest answer is: we don't know yet. llms.txt as a standard has been published for roughly 18 months. Adoption is in the single-digit percentage of major sites. There hasn't been a clean, methodologically transparent test of whether llms.txt presence correlates with LLM citation frequency at scale.

"No measurable impact" without that test is a confident guess. A confident guess is different from a measurement.

Where the line actually is

Strip the labels.

The real question is: which specific tactics genuinely change LLM citation behavior in 2026, and which are vendor noise or skeptic noise?

Our answer, as of this writing:

High
Bing indexation coverage (massive), server-side rendering (massive), entity presence in canonical knowledge graphs like Wikidata (large), backlinks from DR-60+ domains in the same topical cluster (large), original quotable data on the page (medium-high)
Medium
Question-format H2/H3 structure (yes, this lifts citations specifically because LLMs prompt-match), self-contained 50–100 word lead paragraphs (citation-magnet effect), submitting to engine-specific tools (Bing Webmaster Tools, Perplexity Pages submissions)
Contested
Schema markup (we think medium-impact; Jake thinks zero; both are unproven), llms.txt (we think uncertain-but-probably-positive; Jake thinks zero; both are unproven), .md alternate file formats (probably noise)
Low
Most of what mid-tier AEO vendors sell as "AI optimization features"

That third category — the contested one — is the most interesting territory in the entire AEO conversation. Both camps are making confident claims without published methodology. Both are losing credibility every time they're proven partially wrong.

What we're going to do about it

An empirical test.

Over the next 8 weeks, we'll publish an empirical test of every disputed AEO tactic.

Methodology

If schema and llms.txt and the AEO-specific tactics show measurable lift, Jake is wrong about that 20% — and we'll have the dataset to prove it. If they show no lift, Jake is right and a chunk of the AEO industry is overselling — and we'll be the first publication to say so publicly.

Either outcome is good. The only bad outcome is more confident assertions without measurement.

Jake — if you read this — we're not picking a fight. We agree with most of what you wrote. We just think the 20% you wrote off without testing deserves a test. We'll send you an early read when the data lands.

Keep reading