Schema won't save you: what this week's AEO data actually says.

I've spent the past week going through every major AEO dataset that dropped or resurfaced, and one theme keeps showing up: most of what the industry sells as "AI optimization" is theater. The data is blunt about what AI engines actually read, and it's not what most agencies are billing for.

Here's the short version, then the receipts.

Now the detail, and what I'd do about each one.

1. Schema markup doesn't move AI citations. The data is in.

Ahrefs ran the study everyone in AEO should read this week. They tracked 1,885 pages that added JSON-LD schema between August 2025 and March 2026, matched against 4,000 control pages that didn't. After stripping out platform-wide growth, the citation lift was noise: +2.4% on Google AI Mode, +2.2% on ChatGPT, and treated pages actually dropped 4.6% versus controls on AI Overviews.

One important caveat, and it matters for how you apply this: every page in the dataset was already being cited heavily before schema was added. So the honest read is that schema won't push an already-visible page higher. It may still help a page that AI systems can't parse at all get crawled and understood in the first place.

My take: Schema is table stakes plumbing, not a growth lever. If an agency's AEO deliverable is "we added FAQ schema to 40 pages," you paid for a checkbox. Keep Organization and Article schema for entity clarity, then move your budget to the things below.

2. AI engines read what humans read: visible HTML

The companion finding comes from searchVIU's retrieval experiment. When AI engines fetch a page in real time to answer a query, every major engine extracted only the visible HTML content. JSON-LD was ignored entirely at retrieval time.

Sit with that for a second. The optimization layer half the industry obsesses over is invisible at the exact moment an engine decides whether to cite you.

My take: This is why we start every engagement with the words on the page, not the code behind it. If your answer isn't in the visible copy, structured clearly, stated plainly, it does not exist to an answer engine. This is the core of how we think about citability at SuperMarketers.

3. The door into AI citations is narrow, and it's getting narrower

BrightEdge's weekly AI search tracking published a number that should reframe how urgent this is: only 0.4% of tracked domains gained new citations this week. Citations are heavily concentrated among a small set of domains, and the trend is pruning, not expansion. Their data also surfaced a divergence worth watching: several categories saw mentions rise while citations fell. Engines will talk about your brand by name and still decline to link to you.

Being known to the model and earning the citation are becoming two different games.

My take: This is a compounding-advantage market. The brands getting cited today are training the retrieval patterns that keep them cited tomorrow. Waiting a quarter to "see how AI search shakes out" is how you end up on the wrong side of a 0.4% door. It's the whole reason our entry point is a 30-day sprint and not a 12-month roadmap.

4. Optimize the passage, not the page

Pillarbase spent a year analyzing 15.7 million Google AI Mode citations. The headline finding: 47.7% of citations were scroll-to-text highlights rather than plain links. Google isn't citing your page. It's citing a specific passage on your page, and deep-linking straight to it. One highlighted passage in the study appeared across 221 unique queries.

That's the leverage math of AEO in one stat. A single self-contained, well-structured paragraph can do the citation work of an entire content calendar.

My take: Audit your money pages at the paragraph level. Every key claim should survive being lifted out of context: direct answer first, 40 to 60 words, one idea, a number and a source attached. This is exactly what our Citability Score measures, and it's the first thing we run in a visibility audit.

5. Freshness is the strongest signal you actually control

AirOps' 2026 State of AI Search Report analyzed citation patterns across ChatGPT, Perplexity, AI Overviews, and Gemini. For commercial and evaluation-stage queries, the ones tied to pipeline, 83% of AI citations came from pages updated within the past 12 months. More than 60% came from pages refreshed within six.

Pair that with the macro backdrop: Chartbeat data in the Reuters Institute's 2026 report shows Google referral traffic fell 33% globally between November 2024 and November 2025. The clicks are leaving. The citations are what's left, and they go to whoever showed up recently with something worth extracting.

My take: A refresh cadence beats a publishing calendar. Your ten highest-intent pages, updated on a rolling 90-day loop with new data and tightened answer blocks, will out-earn thirty net-new posts. Volume was the old game. Recency plus structure is the new one.

What this means if you run marketing at a B2B SaaS

Strip the week's data down and the playbook is uncomfortable in its simplicity:

• Stop paying for markup theater. Schema is hygiene, not strategy.
• Put the answer in the visible copy, structured so a machine can lift it cleanly.
• Treat your best paragraphs as assets. One citable passage can cover hundreds of queries.
• Refresh relentlessly. Recency is a ranking factor you control every single week.
• Move now. The citation set is concentrating, and the annual gain rate is measured in fractions of a percent.

Become the answer. Then do the unglamorous work of staying it.

If you want to know where you stand today, that's what our free visibility snapshot is for. Twenty of your buyers' real queries, run across the engines that matter, with the gaps mapped.

FAQ

Does schema markup help with AI citations?

Not for pages already being cited. Ahrefs' controlled study of 1,885 pages found no statistically meaningful citation lift from adding JSON-LD. Schema may still help unparsed pages get understood initially, but it is not a growth lever.

What content gets cited most by AI engines?

Passages, not pages. 47.7% of Google AI Mode citations are scroll-to-text highlights of specific paragraphs. Self-contained answer blocks of 40 to 60 words with a clear claim, a statistic, and a source perform best.

How often should I update content for AI visibility?

For commercial-intent pages, at least every six months. 83% of AI citations for commercial queries come from pages updated within 12 months, and over 60% from pages refreshed within six.

Sources

• Ahrefs: We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved.
• searchVIU: real-time retrieval extraction experiment
• BrightEdge: Weekly AI Search Insights, week of August 15, 2026
• Pillarbase: year-long Google AI Mode citation study (15.7M citations)
• AirOps: 2026 State of AI Search Report
• Reuters Institute / Chartbeat: 2026 Journalism Trends Report