What Does “Not AEO-Worthy” Look Like? My Petty Self is About to Tell You

Soooooo, this was super fun. An outside SEO firm rewrote a client’s industry page and told the client our version “wasn’t AEO-worthy.” Bold claim. So I did what any self-respecting marketer with 20 years in B2B and a documented (well, it may as well be) obsession with AI search would do… I pulled up both versions and compared them. Trust me, I definitely have time to do this, there are not 345676540857 other things on my to-do list. Hyperfixation time tho.

Their version: clean prose, readable, no complaints on voice. Also no FAQ block. No FAQPage schema. No Article schema. No structured Q&A content written as natural-language search queries. No implementation handoff notes. Nothing an LLM would grab and cite.

Our version: all of the above BY DESIGN. Every structural element the current research says drives AI citation. Plus headers formatted as the questions buyers type into ChatGPT and Perplexity. Plus standalone FAQ answers in the 40-80 word extraction sweet spot. In fact, we have proprietary skills and SOPs that we’ve been working on since 2025. Like come at the queen, ya best not miss.

So which version is “AEO-worthy,” exactly? There is an Elaine from Seinfeld joke in there somewhere but I can’t find it.

If your SEO firm says your content “isn’t AEO-worthy,” ask them to open the code, not just read it out loud.

But First: Why Should You Even Care

If you’re still on the fence about whether AI search matters, I’m not gonna yell at you. But I am gonna show you the numbers and let you know that as a marketing agency I HAVE to stay on top of this stuff. We can have the debate about should or should we after I hit payroll.

Last month (June 2026), Google added a dedicated “Search Generative AI performance” report directly inside Search Console (via ReleaseBot). Impressions, pages, countries… all broken out for AI Overviews and AI Mode specifically. Google doesn’t build reporting dashboards for funsies. They build them for things that are eating the main product. When the biggest search engine on the planet says “here’s a whole new report for the AI version of your traffic,” that’s a friggin’ foghorn.

Meanwhile, agents are now reading more of the web than humans are. Bots, crawlers, LLMs scraping for training data and real-time answers… they’re the majority audience for most B2B content pages. So when you hire a firm to optimize your content based purely on what a human copywriter thinks sounds good, you’re kinda paying for advice that’s optimized for the smaller audience. It’s like hiring a decorator for a house that’s mostly visited by home inspectors. Make it pretty, sure. But like, maybe also make sure the wiring’s up to code?

Ahrefs analyzed 75,000 brands and found that brand mentions correlate with AI citation 3x more strongly than backlinks do (via BrandCited). Three times. The old playbook (get links, build domain authority, rank) is decoupling from the new one (get cited, get mentioned, get structured). If your SEO firm is still playing the old game exclusively… they’re not wrong exactly. They’re just incomplete. And that can be pricey when the ground is shifting this fast.

Uno mas. Google’s July 2026 core update (dropped July 7) is specifically targeting what they’re calling a “hollow expertise core” (via Shine Magazine).

Translation: content that’s AI-generated with no verified human expertise, no source diversity, no unique perspective. Penalized. Content showing real synthesis from multiple primary sources? Rewarded.

So the question isn’t just “does your page sound nice” anymore. It’s “can Google tell a human expert built this, and can it prove it to an algorithm.”

The Costume Change

Agencies are using “AEO” as a buzzword and clients can’t tell the difference because AEO is new enough that most people haven’t read the research or more accurately, stayed ON TOP of it, like this ish is changing on the DAILY. When someone says your content isn’t optimized for answer engines, that sounds like a technical critique. Sounds like there’s a measurable problem. But if the critique is “it reads a little stiff” or “I’d write it differently”… that’s a voice note. Legit one, sometimes I SUPPOSE. But it’s not an AEO argument. We are not the same.

Voice problems need a copywriter with brand context. AEO problems need structural changes to how the page is built. Apply the wrong fix and you end up with a page that sounds great and never gets cited by anything. It’s the content marketing equivalent of all hat, no cattle. IDK if that’s the case here, but I’d wager it’s more about following rules that were perfectly acceptable in May and are now shifting significantly. The people who succeed with this are going to be the ones who are willing to be wrong and willing to recognize how something that was best practice one week can be completely out of touch the next. Our agency does AEO precisely for this reason. It’s best to keep it under one roof if you can, especially so the people who are writing your copy can also make sure it’s optimized for the agents that are reading it, which is what we do. That’s why our last 10 leads have found us via AI engines. That’s why our clients are surfacing in LLMs all over the place.

What the Research Says (Named Studies, Not “Studies Show”)

I’ve been tracking this obsessively because we’ve built our entire page architecture framework around it. Here’s what holds up.

The GEO study out of Princeton, Georgia Tech, Allen Institute for AI, and IIT Delhi (Aggarwal et al., 2024) found that adding citations, quotations, and statistics to existing content raised visibility in AI-generated answers by up to 41% on average. The biggest gains came on pages that weren’t already dominating traditional search. That’s a structural finding, not a tonal one. Nobody’s LLM is thinking “this page sounds warmer, lemme cite it.”

Semrush analyzed 304,805 URLs being cited by AI tools. Q&A format ranked as the third most important content-quality factor for AI citations, behind clarity (+32.83%) and E-E-A-T signals (+30.64%). Third out of everything. That’s foundational people.

And here’s the one that should end every “schema is the magic bullet” conversation:

Ahrefs ran a causal study on 1,885 pages matched against 4,000 controls between August 2025 and March 2026 (via authoritytech.io). Adding JSON-LD schema alone produced no meaningful citation lift on ChatGPT or Google AI Mode. Google AI Overview citations declined slightly. Schema without genuine Q&A content underneath it doesn’t do anything. The content has to be there. It has to be structured as real questions and real answers. Written in the language people use when they’re searching.

Schema is a signal. The signal has to point at something worth citing. (It’s like putting a neon “OPEN” sign on an empty store. The sign works fine. There’s just nothing inside. At least not for our agentic friends…)

The FAQ Rich Result Is Dead. FAQ Schema Is Not.

Quick clarification because this is getting muddled everywhere. Google deprecated the visible FAQ rich result in the SERP (those accordion-style FAQ snippets under search listings). That happened May 2026, but folks are already using it to claim FAQ schema is dead. Feels like an overcorrection to me.

Nah. The underlying FAQPage schema is still read and used by AI Overviews, ChatGPT, and Perplexity as a citation input (AutomateLab’s analysis walks through this). Conflating the SERP feature with the AI-citation signal is like saying newspapers don’t matter because newsstands closed. Distribution changed. Content still gets read.

Multiple independent analyses in 2026 agree with me: pages with FAQPage schema paired with real Q&A content see meaningfully higher citation rates across ChatGPT, Perplexity, and Google AI Overviews than pages without (Geol.ai, Odyssée agency). Exact multipliers vary by methodology, but the direction is consistent.

Voice and Structure Are Not a Tradeoff

You don’t have to choose between sounding good or ranking in AI. False binary. Total Montague-and-Capulet energy.

I wrote about this over a year ago. We use a two-track framework (Voice Engine and Discovery Engine) specifically because they solve different problems and both have to work. Your headers and FAQ blocks need to be written as natural-language queries because that’s what LLMs match against. Your body copy needs to sound like your brand because that’s what humans read and trust. They coexist on the same page. One doesn’t cancel the other.

The fix for “this sounds robotic” is a voice pass on the prose. The fix for “this won’t get cited by AI” is structural: FAQ blocks, schema, query-matched headers, standalone answers in the 40-80 word range (Averi AI has a good breakdown of why that length matters), and dated, sourced claims. If someone’s giving you one fix and calling it both, they’re either confused about the difference or hoping you are.

Here’s a Super Meta Thing I Never Would’ve Done Before

You might’ve noticed something weird about this article. I’m citing sources differently than I used to. Instead of hyperlinking a phrase like “recent research confirms” (the way every marketer’s been trained to do for 15 years), I’m writing out the full attribution in the text itself. “Ahrefs ran a causal study on 1,885 pages matched against 4,000 controls.” “Semrush analyzed 304,805 cited URLs.” Named institution, methodology, sample size, visible in the prose.

That’s not an accident. It’s ALSO AEO.

LLMs can’t click your links. They read text. When ChatGPT or Perplexity is deciding whether to cite your page, it’s scanning for named sources, specific figures, and attribution patterns it can extract and repeat. A hyperlink on the words “studies confirm” gives the model nothing. “The Princeton/Georgia Tech/Allen Institute GEO study found X” gives it a named, extractable citation with institutional credibility attached.

LLMs can’t click your links. If the source isn’t named in the sentence itself, it doesn’t exist to the model.

And this is where it gets beautifully circular: the GEO study I keep referencing found that adding visible citations and attributions to content raised AI visibility by up to 41% (Aggarwal et al., 2024). The technique I’m using to cite the study… is the technique the study says works. It’s very Inception, very meta, very NOW. We’re three levels deep and the top is still spinning.

The old way of citing (bury the source in a hyperlink, keep the prose clean) was built for human reading flow. The new way (name the source in the text, link it for verification) works for both humans and machines. The link is still there for anyone who wants to click through. But the attribution does the persuasion work before the click, and it does the AI citation work whether anyone clicks or not.

If the firm telling you your content “isn’t AEO-worthy” is still citing sources the old way… that tells you something about how current their methodology is. 👀

What I’d Check If Someone Tells You This

Pull up both versions. Open the source. Look for:

The structural AEO checklist (these are the things that drive citation):

• H2s have the terms and H3s are more marketing-y. PLUS, written as natural-language search queries (how buyers phrase questions, not how marketers label sections)
• A standalone FAQ section with 5-8 questions in the buyer’s words
• FAQ answers between 40-80 words each (complete enough to be useful, short enough for an LLM to snatch and cite)
• FAQPage schema and Article schema in the page markup
• Dated, attributed statistics with the source named in the text (not “studies show,” not a hyperlink on “research confirms,” but “Semrush’s analysis of 304,805 URLs found X”)
• Implementation notes for the dev team

If the version being pitched as “AEO-worthy” doesn’t have these, it’s not an AEO argument. It’s a preference argument wearing a lab coat.

We’ve Been Building This Way for Over a Year

This isn’t a position I cooked up last Tuesday to defend a single page. We’ve been publishing on it, building client pages around it, and measuring results. Here’s the longer version if you want receipts, these are few of MANY:

• AI Search for HR Tech Brands: How to Get Cited by AI Tools covers the exact method: standalone answers, headers as buyer questions, Article + FAQPage schema, dated claims.
• LLM Visibility Playbook for B2B Brands draws the explicit distinction between SEO, AEO, GEO, and “LLM visibility” as separate layers (which matters because people are using these interchangeably and they’re not).
• 5 Critical Thought Leadership Flaws in HR Tech Marketing cites that 32% of buyers now discover content through GenAI tools (ChatGPT, Perplexity, Claude), and frames “write for both humans and algorithms” as the fix.

“AEO” has a definition, and it’s a structural one. If the critique is about how something reads, say that. If the critique is about whether it’ll get cited by AI, open the code and show me.

I’m not a betting woman, but I bet I’m write LOL