The State of AI Search: H2 2026
What actually changed in AI search since March — a Core Update that separated sites with real E-E-A-T infrastructure from sites without it, and where the market is headed next.
// The Update Didn't Move Rankings. It Sorted Sites.
Most "state of AI search" roundups this year are the same four industry reports wearing different bylines. This one isn't. In April, our own average position on Google fell from roughly 22 to 44 in a single day — not because of a penalty, but because Google's March 2026 Core Update went looking for E-E-A-T signals most sites don't have. We watched it happen to our own numbers, fixed the specific things the update was checking for, and recovered. That's the data this post is built on.
This is for marketers and founders trying to get cited by ChatGPT, Perplexity, Gemini, and Claude — not just ranked by Google. Here's what actually changed in AI search since spring, what the Core Update really did, and what to fix first if you haven't already.
// How Much Has AI Search Grown Since Spring?
Growth hasn't plateaued since our last check-in — it's accelerated past most projections made earlier this year.
💡 Key Insight
The market shifted from rankings to citations this year. Founders are now competing to be cited inside a generated answer, not just to appear in a ranked list underneath one. Prompt coverage and citation rate are becoming the metrics that matter — traffic alone doesn't tell you whether AI systems trust you enough to name you.
// The March 2026 Core Update — What Actually Happened
Google's March 2026 Core Update completed rollout on April 8. It was heavily weighted toward E-E-A-T signals, and the volatility it caused was broad — not targeted. Sites without strong author, credential, and citation infrastructure lost ground fast. Sites with it held, or gained.
Why This Wasn't a Penalty
A penalty targets a specific site for a specific violation. What happened in April was different: average position across a wide swath of the web — including our own site — dropped within a matter of days, with no manual action notice, no policy violation, nothing tied to a single page or link. That pattern is the signature of a broad core update re-weighting its ranking factors, not a site-specific strike.
The E-E-A-T Signal Google Weighted Hardest
Verifiable authorship. Pages with a real, schema-backed Person entity behind them — credentials, a bio, an identifiable track record — held their positions far better than anonymous or thinly-attributed content. That's the single factor most sites still get wrong: publishing content with no verifiable entity behind it at all.
// Our Own Recovery, By the Numbers
Here's what our own Search Console data showed in the weeks after the crash and the E-E-A-T fixes that followed — author schema additions, external citations, and expanding thin pages past the citability threshold. (This snapshot is from our most recent full Search Console check; a further-updated figure will follow in a later post.)
The lesson isn't the specific numbers — it's the lever. We didn't change our content strategy or chase new keywords. We strengthened exactly the signals the Core Update was checking for: author schema, Person entities, and external citations to authoritative sources. That was enough to reverse the crash.
// What AI Traffic Is Actually Worth
AI summaries roughly halve raw link-click rates — from around 15% down to 8%, per Semrush's 2026 Trends report. That looks like a loss until you look at what actually converts.
| Metric | Value |
|---|---|
| Branded search traffic lift for AEO-optimized sites | +32% |
| Conversion rate, AI-referred vs. non-AI traffic | +42% better |
| Link-click rate when AI summaries appear | 15% → 8% |
Fewer clicks, but the ones that arrive convert significantly better — the visitor already had their question answered by an AI system that trusted your page enough to cite it before they ever landed.
💡 Key Insight
Perplexity is still the fastest AEO feedback loop. Its citations are visible inline and update near real-time, which makes it the best place to test a structural or schema change before rolling it out everywhere else. ChatGPT and Gemini cite less transparently, but reward the same underlying fundamentals — structured schema and clear authorship.
// AEO, GEO, AIO, LLMO — Why the Terminology Still Hasn't Settled
Query volume blending all four terms together — AEO, GEO, AIO, and LLMO — keeps growing. That's not noise. It's a market that hasn't agreed on a name for what it's doing yet, which means content that only targets one term is leaving intent on the table.
We cover the actual distinctions — and why treating them as interchangeable produces vague, unmeasurable guides — in a dedicated breakdown: AEO, GEO, AIO, and LLMO aren't interchangeable. Until the market consolidates on one term, covering all four remains the safer bet for visibility.
// What to Actually Change Right Now
In priority order, based on what actually moved our own recovery:
1. Author & Entity Schema
The fastest lever for E-E-A-T recovery and the cheapest to implement. Most sites still publish content with no verifiable Person or Organization entity behind it — that's the single most common gap.
2. External Citations
Cite authoritative external sources inside your own content. It signals to both AI systems and Google that your claims are checkable, not just asserted.
3. Expand Thin Pages
Pages under roughly 1,500 words on competitive topics sit below the AI citability threshold. Depth is a trust signal now, not just a ranking one.
// Frequently Asked Questions
What actually changed in AI search between March and September 2026?
Google shipped a March 2026 Core Update (completed April 8) that heavily weighted E-E-A-T signals, causing broad position volatility across the web. AI referral traffic kept growing through the period regardless, with Claude alone accounting for roughly 18.5% of AI referrals. The sites that recovered fastest treated E-E-A-T as infrastructure, not an afterthought.
Is AI search still growing in H2 2026?
Yes. ChatGPT Search now handles roughly 18% of all search queries, Perplexity has captured 12% share among professionals and researchers, and Google AI Overviews has passed 2 billion monthly users.
What is answer engine optimization (AEO)?
AEO is the practice of structuring content so AI systems — ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews — can extract, trust, and cite it directly in generated answers, rather than optimizing purely for a ranked list of blue links.
What triggered the April 2026 position crash, and was it a penalty?
No — it was diagnosed as broad Core Update impact, not a site-specific penalty. The March 2026 Core Update was E-E-A-T-focused; sites without strong credential/author signals saw average position drop sharply within days, with no manual action notice involved.
How do you recover from a Core Update E-E-A-T hit?
Strengthen author schema and Person entities across content pages, add external citations to authoritative sources, ensure NAP/entity consistency, and expand thin pages past the AI citability threshold — roughly 1,500+ words for competitive topics.
Which AI platforms cite sources most reliably?
Perplexity shows citations inline and near real-time, making it the fastest feedback loop for testing AEO changes. ChatGPT and Gemini cite less transparently but increasingly reward structured schema and clear author attribution.
Do I need a different content strategy for each AI platform?
Partially. Core E-E-A-T and schema fundamentals transfer across all platforms, but nuances matter — Perplexity rewards real-time freshness, ChatGPT rewards structured FAQ/HowTo schema, Claude rewards precise sourcing and citation density.
What's the single highest-leverage change to make right now?
Author and entity schema. It's the fastest lever for E-E-A-T recovery and the cheapest to implement — most sites are still shipping content with no verifiable Person or Organization entity behind it.
Why do AEO, GEO, AIO, and LLMO all show up as separate terms?
The market hasn't consolidated on one name for AI-search optimization yet. Query volume blending all four terms keeps growing, which means content targeting only one of them risks missing real search intent during the consolidation period.
Track Your Own Numbers
Recovery data is only useful if you can see it happening on your own site. AEOfix's AI Bot Tracker shows exactly which AI crawlers are visiting, which pages they're pulling from, and how that shifts as you make E-E-A-T changes.