The State of AEO: Key Insights from Vercel's 2026 Report
Vercel published "The State of AEO", a report on how AI-powered search changes the way people find websites. It includes input from Vercel CTO Malte Ubl and from experts at Andreessen Horowitz, LTIMindtree, and Growth Plays. Its main argument is that search is moving from an attention economy to an answer economy.
When someone searches on Google, ChatGPT, or Perplexity, the answer often appears on the results page itself, so fewer people need to visit your website to get it. The goal changes from earning clicks to becoming the source that AI systems cite.
The Numbers in the Report
The Vercel report opens with measurements of how search works now, not projections:
The last figure shows the effect on behavior. 26% of users end their session after reading an AI answer, versus 16% without an AI summary. Those users stop searching, so a brand that the answer does not cite does not reach them in that session.
ChatGPT alone drives 5.72 billion monthly visits. The highest AI summary rate appears for informational queries (88.1%), which are the queries where brands build awareness and authority.
The Platform Landscape
The report covers four major platforms. They approach AI search differently, but they draw on the same web content and reward the same structural signals:
Google AI Overviews
The mainstream option. It combines multiple sources, appears above organic results, and includes experimental ad placements. The report calls it the biggest SERP change since featured snippets.
ChatGPT-As-Search
The all-in-one challenger. OpenAI's integrated model benefits from keeping users on the platform, and its recent transactional features change how content is monetized.
Perplexity
The citation-focused option. It offers inline citations for every sentence, has 22M active users and 50% YoY growth, and draws on trusted sources instead of the open web.
Bing / CoPilot
Microsoft's integrated option. It shows 1.8x higher ad CTR than traditional search and is built into Windows 11 and Edge, which positions it to gain share through OS defaults.
I covered how three of these platforms select sources in separate articles: How ChatGPT Selects Sources , How Google AI Overviews Selects Sources , and How Claude Selects Sources .
Why AEO Is Winning the Naming Battle
The report treats AEO (Answer Engine Optimization), GEO (Generative Engine Optimization), and AIO (Artificial Intelligence Optimization) as names for the same shift. All three optimize for machine understanding through structured data, clear formatting, and entity authority.
I covered the terminology debate in the SEO vs AEO comparison . The short version is that the label matters less than the shift from rankings to citations.
SEO Is Not Dead, but It Is Not Enough
The Vercel report does not frame SEO and AEO as a choice between two options. AI crawlers access and parse content using signals similar to those of traditional search engines. Site speed, mobile responsiveness, clean HTML, and other technical SEO basics still matter for visibility in AI-powered search.
The report also says that E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) matters more than before. AI-powered search engines prefer E-E-A-T content more strongly than traditional search does, which fits the way AI models synthesize information in detail instead of only ranking it.
Malte Ubl, CTO at Vercel, describes it this way: "Like how you had public perception, there's now LLM perception... the tactics are almost similar [to SEO]."
What Is Fundamentally Different
The foundations carry over, but the Vercel report names several areas where AI search differs from traditional SEO:
The Atomic Unit of Optimization Has Changed
AI-powered search breaks content into paragraph-level chunks for retrieval and synthesis. A single page can contribute to several answers, or to none. Optimization moves from the page level to the snippet level.
Entity Authority Outweighs PageRank
In SEO, backlinks are the proof of reputation. In AI-powered search, branded web mentions take that role. The report states that branded mentions matter 3x more than backlinks, because AI assesses entity authority by how often the web discusses your brand.
Natural Language Queries
The average ChatGPT query is 86 words and 8 messages long. Longer contextual questions replace short keywords, and users ask follow-up questions to an AI more often than they explore search results. Answering them requires an understanding of intent and context, which differs from the single matching keywords that traditional SEO favors.
Success Metrics Look Different
AEO success metrics move from site traffic, the main SEO metric, to branded mentions in LLM answers. Users rely on "zero-click" results at least 40% of the time, and web traffic drops by up to 25%. Visibility in AI answers still builds brand authority, which traditional rankings do not measure.
Content Architecture for Machines
The report outlines three content architecture principles for staying visible to AI systems:
Inverted Pyramid
Answer first, explain later. AI systems are more likely to pull from top-of-page content, so answer the question upfront, as early as the title and metadata.
FAQ-Formatted Content
Q&A formats dominate AI answers and overviews. A Q&A structure matches how users query and how AI systems structure responses.
Self-Contained Paragraphs
Paragraphs of 2-5 sentences with subject, context, and conclusion. Self-contained chunks get extracted more often than longer passages that need surrounding context.
Our scanner measures all three. Question-based H2 headings (checkpoint 3.8) capture the FAQ format signal. According to Kevin Indig / Growth Memo , among ChatGPT citations tied to a question, 78.4% came from a heading, so question-led citations tend to come from headings. FAQPage schema (checkpoint 2.7) covers structured data. Cited sites carry FAQPage schema far more often than non-cited ones (6.2% versus 0.8% in Insightland's data). That is an 8x correlation, not proof that schema makes a page 8x more likely to appear. A 2026 Ahrefs study of 1,885 pages found that adding schema did not measurably change AI-search citations, so treat this as a correlation: cited pages tend to be well-structured, and schema itself does not directly drive citations. The third check is server-side rendering (checkpoint 3.1), which checks that the content is accessible to AI crawlers in the first place. I explain why in how AI agents see your website .
The False Dichotomies
The report disputes two common misconceptions:
The Magic AEO Tag Is a Myth
No schema markup or meta tag guarantees visibility in AI search results, whatever a vendor claims. Visibility depends on content quality, structure, and authority together. On claims of guaranteed inclusion, the report is blunt: "they're lies."
Blocking AI Crawlers Is Usually Self-Defeating
Publishers have blocked OpenAI's and Google's AI crawlers to protect their content. The report argues that this costs the content its AI visibility, because a page that no crawler fetches cannot appear in an answer. Emerging standards like WebMCP and MCP let agents interact with your content on your terms. Our scanner checks the crawler side of this. AI crawler directives (checkpoint 1.2) evaluates whether you have explicit Allow/Disallow rules for 13 different AI user-agents.
The Revenue Question: Who Pays When No One Clicks?
One section of the report covers the economic tension in AEO. If AI provides answers without generating traffic or clicks, the advertising revenue that supports free content disappears. The report calls this the "garbage in, garbage out" problem for AI:
- If content creators stop producing new information, AI quality will degrade
- If AI quality degrades, content creators won't want to produce for degraded AI models
- The result is a negative feedback loop that could undermine the AI search ecosystem
Some companies already use solutions. OpenAI has signed content licensing agreements with The Financial Times and News Corp: upfront payments, ongoing royalties, and mandatory attribution. NerdWallet is a counterexample. Despite a 20% decline in site traffic, the company generated 35% revenue growth by optimizing for AI-driven brand recognition that converts through direct visits later.
The report also points out the opaqueness of LLM algorithms. With SEO, backlinks and PageRank gave transparent signals. With AI search, you cannot trace why a given answer appeared or link an optimization action to an output. Identical prompts can yield different results because of stochastic sampling, context window shifts, or training data noise. Optimization therefore moves from exploiting explicit signals to probabilistic influence: raising the likelihood of inclusion through structured, authoritative content.
How Our Scanner Maps to These Recommendations
I built the IsAgentReady scanner to measure signals like the ones the Vercel report names. This table maps the report's recommendations to our 77 checkpoints across 5 categories:
| Vercel Recommendation | Our Checkpoint | Category |
|---|---|---|
| Optimize for AI crawlers | AI crawler directives (1.2), robots.txt (1.1), HTTP bot access (1.8) | AI Content Discovery |
| Use SSR/SSG for crawlers | Server-side rendered content (3.1) | Content & Semantics |
| Structured data for AI parsing | JSON-LD (2.1), Schema types (2.2, 2.3), Schema validation (2.6) | AI Search Signals |
| FAQ-formatted content | FAQPage schema (2.7), Question-based H2s (3.8) | AI Search Signals + Semantics |
| Content freshness signals | Content freshness (1.6): dateModified, article:modified_time | AI Content Discovery |
| E-E-A-T signals | Author attribution (2.8), Organization schema (2.2) | AI Search Signals |
| Entity authority / structured markup | Entity linking @id (2.4), BreadcrumbList (2.5) | AI Search Signals |
| Clean, semantic HTML | Heading hierarchy (3.2), Semantic HTML (3.3), ARIA (3.4) | Content & Semantics |
| llms.txt for AI discovery | llms.txt (1.4) | AI Content Discovery |
| Security & trust signals | HTTPS (5.1), HSTS (5.3), CSP (5.4), CORS (5.7) | Security & Trust |
Adapting to the Answer Economy
The Vercel report closes with this line: "Shifting your tactics to cater to an answer economy is not optional - it's existential." The report's argument is that brands cited as "the answer" to user queries gain visibility and the others lose it. Search is not the only channel that changes. Agents also affect distribution, as I wrote in why CLIs are the new distribution channel .
Most of the foundational work overlaps with good SEO practice. The additions are structured data, AI crawler access, content chunking, and FAQ formatting, and our scanner can measure each of them. More on what AI agent readiness means in practice.
Sources
- Vercel: The State of AEO (2026). Gated report by Vercel with input from Malte Ubl (CTO), Nick Lafferty (Profound), Zach Cohen & Seema Amble (a16z), and market data. Distributed via LinkedIn.
- How Vercel's Adapting SEO for LLMs and AI Search (Vercel): Vercel's public strategy for AI search visibility and brand mentions
- How We Built AEO Tracking for Coding Agents (Vercel): How Vercel built their AEO measurement system
- How Generative Engine Optimization (GEO) Rewrites the Rules of Search (Andreessen Horowitz): Zach Cohen & Seema Amble on visibility in AI-generated answers
- AEO: How AI and LLMs Are Disrupting Search (Airtree Ventures): AEO guide featuring Malte Ubl (Vercel CTO) on content strategy
- The State of AEO / GEO in 2026: CMO Investment Report (Conductor): Survey of 250+ digital leaders on AEO investment and impact
- IsAgentReady: SEO vs AEO, How Traditional Search Differs from AI Optimization
- IsAgentReady: How ChatGPT Search Chooses Which Websites to Cite
- IsAgentReady: How Google AI Overviews Selects Sources to Cite
- IsAgentReady: How Claude Search Selects Sources to Cite