AX is the New AEO was submitted to arXiv on 2026-09-28 by Ido Finder, Assaf Elovic, Gad Shalev, Liad Yosef. Read it on arXiv or as a PDF.
Why it matters for getting cited
- AI models draw only 7-10% of their final answers from training knowledge, relying heavily on live web retrieval.
- Agent-ready websites have answers built from their own pages 78% of the time, compared to 56% for sites that agents cannot easily read.
- Businesses with readable sites are clearly recommended 1.9x more often by autonomous evaluation agents across four testing setups.
- Generating a grounded answer about a site that is not agent-ready costs the agent 64% more on average.
- Holding the business, harness, and question fixed, answers drawn directly from the business website are 41% more accurate.
- When answers rely on third-party web content, they are 3.7x more likely to omit all facts requested by the buyer.
The main takeaway. Optimizing for AI engines requires making your own website readable to automated agents during deep research, because being readable drives recommendations far more than being mentioned across the web.
The abstract, as written
In 2023, AI models answered from training data and hallucinated when it ran out, and businesses were told to seed that knowledge. Models’ training knowledge has since given way to live web search, and the advice followed it there: answer-engine optimization, or AEO, now tells businesses to scatter breadcrumbs across forum threads, listicles, and off-site citations, so AI engines are likelier to surface and recommend them. But being surfaced is no longer enough: an agent opens the results and reads them before deciding, and one buyer question sends it through several rounds of search and fetch. What decides the outcome at this drill-down step is whether the agent can fetch and read the business’s own site: agent experience (AX). We argue that AX is the new AEO. We run 37,927 agent journeys, each a buyer question about a business, across four independent harnesses over 1,056 real businesses, matched on fame, prior model knowledge, and two AEO proxies, then split based on their AX level. Only 7-10% of the finished answer comes from the model’s training knowledge, whether or not the site is readable. Agent-ready businesses have answers built from their own pages 78% of the time against 56% and are clearly recommended 1.9x more often, while a grounded answer about a not-agent-ready business costs the agent 64% more on average. Holding business, harness, and question fixed, answers built from the site are 41% more accurate on average. The dominant failure is not fabrication but omission: web-built answers are 3.7x more likely to contain none of the facts the buyer asked for. Baselines differ sharply across the four harnesses, with clear-recommendation rates varying sevenfold from stack to stack, yet the recommendation gap holds in every one. In the agentic web era, being readable beats being talked about, and improving a site’s AX is the strongest lever a business has.
Key terms
- Agent Experience (AX): The technical ease with which an autonomous software agent can request, parse, and comprehend the content of a website.
- Answer Engine Optimization (AEO): The practice of structuring and distributing content across the web to increase the likelihood of being cited by conversational AI systems.
- Drill-down Step: The secondary phase of an agent journey where the system opens and reads specific retrieved links rather than scanning surface search snippets.
- First-party Grounding: Building a factual response using verifiable information taken directly from the official domain of the subject business.
- Agent Readiness: A measure of how well a digital property serves automated bots that do not execute client-side scripts or solve interactive challenges.
From the paper
Most public web properties remain unsuited for modern automated agent access.
Across nearly 100,000 sites, fewer than 1% earn its top grade of A or above: most of the web is not ready for the consumers it now has, and a standards effort is under way to define what readiness requires ( Anthropic, 2024 , Microsoft, 2025 , agentready.org, 2026 ) .
Surfacing in search results is separate from providing the actual substance of an answer.
A single buyer question sends an agent through several rounds of search and fetch, and what it reads in those rounds is what the answer gets built from. SEO and current AEO both govern the surfacing step; the drill-down step decides what the answer is made of.
When a primary site fails to load for an agent, the retrieval falls back onto secondary sources.
If the agent can read you, it grounds its answer in your first-party facts. If it cannot, because the page is JavaScript-only, because the site’s bot controls block the agent ( Prince, 2025 ) , or because access is otherwise gated ( Longpre et al., 2024 ) , the agent does not stop. It has the open web to fall back on: third-party pages, competitors, listicles, aggregators.
Direct site readability represents the primary asset an enterprise directly controls.
Being found still matters. Being read is what decides the answer, and it is the one input the business owns. AX is the new AEO, and it is where the investment belongs.
Figures from the paper
Agent-ready domains receive clear buyer recommendations nearly twice as often as non-ready domains across all evaluated systems.
Figure from the paper, arXiv:2609.34951, by its authors. Caption as written: Figure 4: One in five answers clearly recommends an agent-ready business; one in nine when it is not agent-ready. Bars are the share of answers that both blind judges rate a clear recommendation, all four stacks pooled; both judges must award the top grade for an answer to count (permutation p < 0.0001 p<0.0001 ).
What the authors say it means
- Traditional AEO strategies that plant mentions on third-party forums help agents discover a brand but fail to control the final narrative.
- Blocking autonomous agents with aggressive bot firewalls harms brand perception by forcing models to synthesize answers from low-quality external aggregators.
- Search engines and answer systems penalize unreadable sites through omitted information and hesitation hedges rather than explicit falsehoods.
- The engineering priority for digital marketing must shift toward server-rendered text and machine-friendly site architectures.
What to do with it
- Ensure key product information and pricing are accessible in plain text without client-side script execution. Supported by: Answers built directly from primary company pages are 41% more accurate than answers drawn from third-party sources.
- Audit web security rules to avoid inadvertently blocking legitimate autonomous research agents. Supported by: A grounded answer about a site that is not agent-ready costs the agent 64% more on average due to extra retrieval loops.
- Focus optimization budgets on primary domain accessibility over third-party link placement. Supported by: Agent-ready businesses secure clear recommendations 1.9x more often across independent testing harnesses.
Caveat. The study analyzed business fact lookups across software and commerce websites, without evaluating multi-turn transactional workflows or non-English inquiries.
In short
AI engines have moved away from memorized facts and now browse the live web before answering buyer queries. If an agent cannot read your primary pages, it reconstructs facts from third parties and drops its recommendation rate. Ensuring clean machine readability on your own domain provides far higher returns than scattering mentions across the web.
Written from the paper's own text and checked by a person before publishing. Quoted passages are verbatim; numbers are the authors' own. For the method and full results, read the paper.