There is a number buried in Adobe’s Q1 2026 retail data that should stop every marketing director mid-sentence: US retail product pages average 66% LLM readability. One-third of the content on the most commercially important pages on most retail websites is structurally invisible to AI systems — before anyone has asked a single question about content quality, keyword strategy, or brand authority.

That number comes from Adobe’s AI Content Visibility Checker, which analysed over one trillion visits to US retail sites and scored pages by what proportion of their content large language models can actually parse. Not what ranks well. Not what converts. What AI can read at all. (source: ai-traffic-retail-adobe.md)

Most GEO and AEO advice skips straight past this. It assumes AI systems can access your content and proceeds to debate tone of voice, FAQ structure, and entity salience. That assumption is wrong for a material proportion of most websites, and the consequences are not marginal — they are total. A page that scores 66% on machine readability has already lost a third of its content to the AI before it competes on anything else.

What Machine Readability Actually Measures

Machine readability is not about whether your content is good. It is about whether AI systems can retrieve it at all.

The score is simple in concept: 100% means every element of your page is accessible to an LLM crawler; 66% means one-third of it might as well not exist. It is the technical prerequisite that all other AI visibility work depends on. You cannot optimise content that an AI cannot see. You cannot build entity salience from material an AI cannot access. Brand authority is irrelevant on a page that is structurally inaccessible. (source: machine-readability.md)

This is distinct from content parsability — which concerns how well AI systems extract and quote the content they can reach. Machine readability is the prior question: whether AI can reach it at all. The order of operations matters: machine readability first, then content parsability, then entity recognition, then citation selection. Most brands are working on steps three and four while step one is broken. (source: content-parsability.md)

The Common Blockers

The causes of low machine readability are well-documented and, in most cases, fixable. They are not exotic technical problems. They are configuration decisions that were made without AI systems in mind, because when those decisions were made, AI systems were not a factor.

JavaScript rendering. Most AI crawlers cannot execute JavaScript. If your product descriptions, specifications, pricing, or navigation load via JS rather than appearing in the raw HTML response, that content does not exist for the crawler. Many modern e-commerce platforms default to JS-rendered product content. The result is precisely the 66% product page readability floor that Adobe’s data reveals. (source: geo-strategy-4-pillar-lumar.md)

robots.txt bot blocking. AI crawlers — GPTBot, PerplexityBot, Google-Extended, BingBot, CCBot — must be explicitly allowed. Many sites that updated their robots.txt during the early 2023–2024 AI scraping panic are still blocking bots they may now want crawling their content. This is not a theoretical risk; a blanket disallow on AI crawlers means zero AI visibility, regardless of every other optimisation effort. (source: geo-strategy-4-pillar-lumar.md)

nosnippet directives. Google’s AI Optimization Guide (2026) makes explicit that pages must be both indexed and snippet-eligible to appear in AI Overviews and AI Mode. A nosnippet directive removes a page from AI-generated responses even if that page is fully indexed, well-ranked, and technically excellent in every other respect. Many sites apply nosnippet directives to page types — campaign landing pages, legal notices, certain product variants — without realising they are simultaneously deregistering those pages from AI citation. (source: machine-readability.md)

Interactive content patterns. Accordions, dynamic tabs, sliders, and other interaction-dependent components present a particular problem: the content exists in the HTML but only renders when a user clicks. An AI crawler sees the shell, not the content. FAQ sections built as accordions — arguably the format most optimised for human convenience — are systematically disadvantaged in machine readability terms. (source: content-parsability.md)

Step Zero in Any AI Visibility Audit

The GEO funnel has three stages: AI Discovery (can AI find and access your content?), AI Understanding (can AI correctly interpret it?), and AI Inclusion (will AI cite it?). Machine readability sits at the very base of stage one. A brand that fails at Discovery cannot benefit from strong signals at any later stage. (source: geo-funnel.md)

The practical implication is that machine readability is not one consideration among many in an AI visibility strategy. It is step zero — the technical audit that must be completed before any content or authority work can be evaluated. Lumar’s 4-pillar GEO framework makes this explicit: Technical GEO is the first pillar precisely because it gates everything that follows. “Entity GEO must come first. Without it, Brand Authority GEO has nothing to attach to” — but before entity work, there must be technical access. (source: geo-strategy-4-pillar-lumar.md)

Most brands have never run a machine readability audit. They have run keyword audits, content audits, technical SEO audits, and CRO audits. They have not specifically asked: what percentage of my content can AI systems actually read? For many, the honest answer is somewhere below 75%, and for product pages specifically, the sector average says 66%.

The Opportunity Cost of Skipping Step Zero

Adobe’s data shows a 52-percentage-point readability gap between the best-performing and worst-performing US retailers — 82.5% versus 54.2% at homepage level. That gap is not primarily explained by content strategy. It is a technical configuration gap. The brands at the top have cleared the access layer; the brands at the bottom are investing in content quality that AI systems cannot retrieve. (source: ai-traffic-retail-adobe.md)

This matters more as AI traffic becomes commercially meaningful. AI-referred visitors to retail sites now convert 42% better than non-AI traffic, spend 48% longer on site, and browse 13% more pages per visit. The visitors AI sends are higher quality than the visitors traditional search sends. The question is not whether you want them — it is whether your site is technically accessible enough to receive them. (source: ai-traffic-retail-adobe.md)

What to Measure First

The diagnostic starting point is a page-level machine readability score across your key page types, with specific attention to product pages, category pages, and any FAQ or help content. Adobe’s AI Content Visibility Checker provides sector benchmarks. Google Search Console confirms indexing status. A robots.txt review across all AI bot user agents takes minutes. A nosnippet audit can be run with any standard crawl tool.

None of this is complicated. The gap exists not because the problem is hard but because the question has not been asked. Brands have not been measuring machine readability because the metric did not exist in any meaningful form until AI visibility became commercially urgent — and now that it has, most are still not asking.

The content strategy conversation is premature for any brand that has not completed this audit. You are debating the quality of words that one-third of your audience — the AI systems increasingly deciding which brands get recommended — may never see.

Fix the access layer first. Everything else is optimising content that AI cannot read.