Agent Experience Optimization Is Now a Product Category — Your Website Has Two Audiences
On October 28, 2026, two companies — one a startup, one an enterprise PIM vendor — independently launched products built on the same premise: your website is no longer just for humans.
Fibr AI released Agent Experience Optimization (AXO), a platform that detects whether a visitor is a human or an AI agent and reshapes the page content accordingly. The same day, Akeneo shipped its Autumn Release 2026 with Readiness Policies and AI discoverability scoring — features that evaluate how findable your products are by AI systems and feed those signals back into your data workflows.
Neither company coordinated with the other. They arrived at the same conclusion from different directions. And that convergence is the signal. Agent Experience Optimization is no longer a concept mentioned in conference talks. It is a product category with venture-backed vendors, enterprise customers, and measurable KPIs.
What AXO Actually Does
AXO — Agent Experience Optimization — is the practice of making your website's content readable, queryable, and actionable by AI agents. It sits alongside UX (user experience), CX (customer experience), and CRO (conversion rate optimization) as a distinct discipline with its own metrics and tooling.
The difference between AXO and traditional SEO is the audience. SEO optimizes content for search engine crawlers that index pages and rank them in results. AXO optimizes content for AI agents that read pages, extract information, and take action — purchasing products, filling forms, comparing options, and answering questions on behalf of their users.
This distinction matters because what makes a page rank well in Google and what makes a page useful to an AI agent are increasingly different things. A page stuffed with keywords in promotional prose might rank for a search query. An AI agent trying to extract your product's specifications from that same page will struggle to separate marketing copy from structured facts.
How Fibr AI Approaches It
Fibr AI's AXO platform does something that sounds simple but has significant implications: it identifies whether a page visitor is a human or an AI platform, and adapts the content for that visitor.
For humans, the page renders normally — marketing copy, visual hierarchy, promotional elements. For AI agents, the same page surfaces structured data, clear attribute-value pairs, and machine-readable content that the agent can parse without navigating a user interface designed for a different species of intelligence.
The platform also supports agent actions. AI agents can complete forms, submit applications, and make purchases on behalf of consumers — either directly on the website or through AI assistants via MCP integrations. This is the transactional layer that separates AXO from simple structured data markup.
Fibr reports testing AXO on its own website. Average daily AI citations rose from 41 during the initial launch week to 114 during October 1–7, nearly tripling. These are self-reported figures, and the methodology and timeframe are narrow, but the directional claim — that adapting content for AI visitors increases AI citations — aligns with what the broader data shows about agent behavior.
How Akeneo Approaches It from the Data Layer
Akeneo's Autumn Release takes a different angle. Instead of adapting the presentation layer for AI visitors, Akeneo focuses on the product data itself — making sure the data is agent-ready before it ever reaches a website or marketplace.
Readiness Policies let teams define what a product must have before it is considered ready for a specific channel, market, or use case. This goes beyond traditional completeness checks. A product might have all required fields filled in but still fail an AI readiness policy because its specifications are embedded in a marketing paragraph instead of structured as discrete attributes.
PX Insights, the analytics layer, now tracks how products are discovered by AI systems and scores their readiness for AI-mediated commerce. This creates a feedback loop: teams can see which products AI agents are finding, which they are skipping, and what data gaps cause the skips.
Akeneo also added transparent AI labeling, which marks content that was generated by AI agents versus validated by humans. As AI agents increasingly write product descriptions, knowing which copy is machine-generated and which is human-reviewed becomes a data quality signal in itself.
The Convergence Is Not a Coincidence
Fibr AI and Akeneo launched on the same day by coincidence. But the convergence of their product strategies is not coincidental. It reflects a structural shift that three data points make visible.
First: agent traffic is at infrastructure scale. DataDome's Q3 2026 report showed 25 billion AI agent requests per quarter across its network, with agentic browser traffic — agents that navigate websites like humans — jumping 4.7x in September alone. These agents are reading websites right now. The question is not whether to optimize for them. It is how badly your current site fails at serving them.
Second: commerce protocols assume agent-readable data. Google's Universal Commerce Protocol requires merchants to publish a discovery manifest at /.well-known/ucp so agents can find their capabilities. Walmart's Agent Gateway serves structured product data through MCP. OpenAI's Agentic Commerce Protocol powers ChatGPT Shopping. Every commerce protocol launched in 2026 assumes that product data exists in a machine-readable format that agents can query. None of them assume agents will scrape a marketing website.
Third: the two-audience problem has arrived. Every website now has two audiences: humans who browse visually and AI agents who read structurally. These audiences want the same information but in fundamentally different formats. A human wants a visually appealing product page with lifestyle imagery and persuasive copy. An AI agent wants structured attributes — material, dimensions, price, certifications, availability — that it can evaluate against alternatives in milliseconds.
Serving both audiences from a single page without compromising either is the core problem AXO solves. And it is now a problem that multiple vendors are building products to address.
Why Traditional Approaches Are Not Enough
You might think existing tools solve the two-audience problem. Schema markup, for example, adds structured data to pages. Meta tags provide machine-readable descriptions. Sitemaps tell crawlers what pages exist. These tools were designed for a world where machines indexed content and humans consumed it.
The AXO world is different. AI agents are both the indexers and the consumers. They do not just discover and rank your content — they read it, evaluate it, compare it, and act on it. And they do this at a scale and speed that reveals every gap in your data.
Schema markup is necessary but insufficient. As we covered in our analysis of schema markup and AI recommendations, structured data helps agents parse your pages but does not determine whether agents recommend your products. Authority, consistency, and content clarity matter more than the presence of JSON-LD alone.
SEO optimization targets the wrong intermediary. SEO optimizes for Google's ranking algorithm. AXO optimizes for the agent that reads your page after finding it through Google, or through a commerce protocol, or through a direct API query. The ranking gets the agent to your page. The page content determines whether the agent trusts, understands, and acts on what it finds.
Content management systems serve human workflows. Most CMSes are built around the workflow of creating content for human consumption — WYSIWYG editors, preview modes, responsive layouts. None of them natively support the dual-output requirement of AXO: human-optimized rendering and agent-optimized data extraction from the same source content.
What the AXO Stack Looks Like
The emerging AXO stack has four layers, and different vendors are addressing different layers:
1. Data Quality Layer
This is where Akeneo operates. Product data must be structured as discrete, queryable attributes — not embedded in marketing prose. Every specification, certification, dimension, and category must exist as a standalone data point that an agent can evaluate without parsing natural language.
2. Presentation Adaptation Layer
This is where Fibr AI operates. The same underlying data must render differently depending on the visitor. Human visitors get visual hierarchy, imagery, and persuasive copy. AI visitors get structured attributes, clean data formats, and machine-actionable endpoints.
3. Protocol Compliance Layer
This is where Google UCP, Walmart's Agent Gateway, and OpenAI's commerce protocol operate. Your data must be discoverable through standardized interfaces — discovery manifests, MCP servers, API endpoints — that agents know how to query without reverse-engineering your website.
4. Verification Layer
This is where brand data consistency matters most. AI agents cross-reference your product data against third-party sources — review platforms, directories, business registries, social profiles. Inconsistencies between your website and these verification sources reduce the trust score that agents assign to your brand.
Fetching Company operates primarily in this fourth layer. Our API extracts structured brand data — logos, colors, fonts, social profiles, contact information — and provides a machine-readable brand fingerprint that agents can use to verify brand identity across sources.
The Metrics That Matter
AXO introduces metrics that traditional web analytics do not track:
AI citation count. How many times AI systems cite or reference your content when answering user queries. Fibr reports tripling this metric through AXO optimization.
Agent action completion rate. What percentage of AI agent visits result in a completed action — a product added to cart, a form submitted, a purchase completed. Traditional conversion rate measures human visitors. Agent action rate measures agent visitors.
Data extraction accuracy. When an AI agent reads your page, how accurately does it extract the structured data you intended? Ambiguous content, marketing prose masquerading as specifications, and inconsistent formatting all reduce extraction accuracy.
Cross-source consistency score. How consistent is your brand data across every surface an agent checks? Your website, directory listings, review platforms, social profiles, and product feeds should all present the same information. Every inconsistency is a trust penalty in agent-mediated evaluation.
What This Means for Your Brand
The launch of AXO as a product category is an inflection point. It means that the infrastructure to serve two audiences from a single website is becoming productized. Early adopters will gain a structural advantage in agent-mediated discovery and commerce. Late adopters will find their websites increasingly invisible to the fastest-growing traffic segment on the internet.
Here is what to do now:
Audit your website for agent readability. Load your key pages and ask: if an AI agent read this page, could it extract structured product data without parsing marketing copy? If the answer is no, your content needs restructuring.
Separate structured data from narrative content. Every product specification, price point, certification, and category should exist as a discrete data point — in schema markup, in API responses, and in your CMS — not only inside a paragraph of marketing prose.
Ensure brand consistency across verification sources. AI agents cross-reference your brand data. Use a tool like Fetching Company to extract your current brand fingerprint and compare it against what appears in directories, review platforms, and social profiles. Fix every inconsistency.
Monitor AI agent traffic today. Check your server logs for AI crawler user agents. The 25 billion quarterly requests mean agents are already visiting your website. Understanding which agents visit, what pages they read, and how long they spend tells you what your AXO baseline looks like.
Evaluate the emerging AXO tools. Fibr AI, Akeneo, and the commerce protocols are the first wave. More tools will follow. Understanding the AXO stack now — data quality, presentation adaptation, protocol compliance, verification — helps you evaluate which tools address the gaps in your current setup.
The Two-Audience Reality
For fifteen years, websites have been optimized for one audience with two variations: humans who browse and search engines that index. SEO bridged the gap between these variations because search engines are fundamentally intermediaries — they rank content for humans to consume.
AI agents are not intermediaries. They are consumers. They read your content, evaluate it, compare it, and act on it — sometimes without a human ever seeing your page. This makes them a genuinely new audience, not a new variation of the old one.
The brands that recognize this distinction earliest will build the data infrastructure, presentation layers, and protocol integrations to serve both audiences effectively. The brands that treat AI agents as just another search engine to optimize for will find their websites increasingly bypassed by agents that can get better data faster from competitors who took the two-audience problem seriously.
AXO is now a product category. The tooling exists. The protocols are live. The agents are already visiting. The only question left is whether your brand data is ready for its second audience.
Make your brand agent-ready. Extract your brand fingerprint with Fetching Company and verify that your brand data is structured, consistent, and queryable — for humans and AI agents alike.