The Two-Tier Access Doctrine — why pages that load still fail in AI answers
Information access (reaching a page) is solved. Cognitive access (extracting an answer from it) is not. Corrao + Fulantelli 1998 is the doctrine that closes the gap — 25 years early.
By Jonathan Landman · Published · 12 min read
The 60-second answer
A page can load for a human and load for an AI engine and still fail both. The HTTP request succeeds; the render completes; the crawler indexes the URL — and the answer inside the page never gets extracted. That is the silent failure mode of the 2026 web. The discipline that closes it is cognitive access: designing the page so a reader, human or model, can pull a self-contained answer out of it in seconds.
The doctrine has a name and a lineage. In the late 1990s, two researchers at the Italian National Research Council in Palermo — Corrao and Fulantelli — drew the line between information access (reaching a page) and cognitive access (cognitively elaborating its content). They were designing a web-based instruction system for architecture students. They were 25 years early to the problem that now decides which brand the AI engine cites.
The two-tier failure mode.
A premium agency ships a beautifully designed site. Core Web Vitals are green. Lighthouse scores in the 90s. Schema is present. The page is indexed in Google within hours of publication. By every conventional measure, the page has arrived.
Then the AI Overview answers the prospect's buying question — and cites the competitor instead. The engine visited the page. The crawler retrieved the HTML. The model parsed the DOM. And the answer it composed did not pull from that page, because that page never made its answer extractable in the first place. The page loaded for the AI. It was not read by the AI.
This is the failure mode 2026 web optimisation does not see. Lighthouse audits whether the page loads. The Wiele Cognitive Access Audit (£1,950) audits whether the page is cognitively accessible. They measure different tiers. Most pages pass the first. Few pass the second.
Where the doctrine comes from — Palermo, 1998.
The philosophical origin sits in a short paper from the Italian National Research Council (Consiglio Nazionale delle Ricerche, CNR). R. Corrao and G. Fulantelli, at the Institute for Education and Training Technologies in Palermo, were building a web-based instruction system for University of Palermo architecture students. Their paper — Cognitive accessibility to information on the Web — drew a careful distinction the wider web-design field never fully internalised.
Reaching a page, they argued, is not the same as cognitively elaborating its content. A learner can request a resource, render it, scroll it, and walk away without having learned anything. The HTTP transaction succeeded; the cognitive transaction did not. They called the first information access and the second cognitive access, and they argued that web design that ignored the second tier was design that worked only on paper.
Their answer was a set of on-page scaffolds they called the Working Tools Bar — post-its, markers, page-marks, a kit-bag for saved fragments, an operation-aware history they named Iter. The tools were architectural; what matters today is the doctrine underneath them. Corrao and Fulantelli named the wedge before LLMs existed, before the iPhone existed, before mobile-first design existed. The architecture of the problem outlasted their architecture of the solution.
The AEO era brings the wedge back, harder. AI answer engines do not retrieve passages — they cognitively elaborate them into answers. A page that does not present an extractable, self-contained answer fails AI cognitive access for exactly the reason Corrao identified in 1998. The doctrine is not new. The cost of ignoring it is.
Why pages that load still fail.
Four recurring failure modes turn a fast, indexed page into one the engine cannot extract from. Each is invisible to Lighthouse. Each is fixable inside a sprint.
01
Anaphora across paragraph boundaries
Pronouns and demonstratives — this, that, it, they — that refer back across a paragraph break lose their referent the moment the engine chunks the passage. The human reader inferred the antecedent; the extractor sees a passage with a dangling pronoun and skips it. Each H2 block must be its own closed referent system.
02
Heading hierarchy that does not segment the answer
Marketing-style headings (The Future of Search, Why It Matters) read well to humans and tell the extractor nothing. A heading should name the answer it introduces. Declarative headings outperform rhetorical headings across every engine class because the heading tree IS the engine's passage-retrieval index.
03
Visual emphasis without semantic weight
A phrase styled large and bold in the design system but
wrapped in a plain <span> reads as
important to the eye and as body prose to the parser.
Importance has to be expressed in HTML semantics
(<strong>, <h2>,
<dt>) before any CSS treatment lands
on top. Pretty markup with semantic body text is the
commonest agency-built failure pattern.
04
Single-source self-reference
The page asserts facts and cites nothing. No Schema.org markup, no named author, no inbound entity links, no sourced statistic, no dated revision. The engine has no signal to weigh the claim and no fragment of citation graph to attach the page to. Authority is something the page either participates in or sits outside of. Self- reference alone keeps the page outside.
The three tests Wiele uses.
Cognitive access is not a vibe. It is auditable against three named tests from the cognitive-psychology and usability canon. Wiele applies all three on every page that needs to be cited.
Krug's Trunk Test — 5-second self-evidence. Steve Krug's 2014 diagnostic: imagine being dumped on the page with no context. Within five seconds, can a reader identify what the page is, what it answers, where it sits in the site, and what to do next? If the page fails that test for a human reader, it fails it for an extractor running on a tighter time budget.
Whalen's Six Minds audit grid. John Whalen's 2019 framework: a customer experience is not one thing but six parallel cognitive processes — Vision and Attention, Wayfinding, Memory, Language, Decision-Making, Emotion. A page that lands on five of the six and leaks the sixth is a page that loads and fails. The grid converts cognitive access into a 6-row audit any operator can execute.
Jones et al. 50-millisecond first-impression test. Jones, Fitzpatrick and Chassy (Liverpool Hope University, 2015): aesthetic judgement on a web page locks in at 50 milliseconds and persists. The three engineerable variables are prototypicality (does the page match what the reader already expects an authoritative page to look like?), moderate visual complexity (rich enough to signal authority, not so dense it overflows working memory), and processing fluency (clean typography, predictable layout, no cognitive friction in the first viewport).
What cognitive access requires — the seven signals.
Seven engineering decisions, applied together, move a page from information-accessible to cognitively-accessible. Each is auditable. Each is shippable inside a two-to-four-week sprint.
01 · Entity-rich first paragraph.
The opener names the entities the page is about — brand, framework, category, named author — in the first 50 to 100 words. Entity prominence on first appearance lifts both human salience and crawler resolution.
02 · Self-contained H2 answer blocks.
Every H2 stands alone. A reader landing on the section via anchor link, or an extractor pulling that block in isolation, gets a complete answer without needing the sections above or below.
03 · Working-memory-safe chunking.
Cowan's 4-chunk cap on working memory: visible vertical lists, callouts, and option sets stay at four or fewer items. Longer enumerations get sub-grouped. Compression beats truncation.
04 · JSON-LD structured data, inlined.
Article, FAQPage, HowTo, Person, BreadcrumbList — server-rendered into the HTML response, not injected after hydration. Non-JS crawlers see the schema; non-JS extractors use it.
05 · Internal entity links, not topical anchors.
Anchor text is the entity name (Citation Score™) not a generic phrase (learn more). Internal links form an entity graph the engine can traverse; topical anchors form a dead-letter office.
06 · Explicit re-elaboration of the question inside the answer.
A passage that answers how do agencies get cited in AI answers repeats the question's entities and verbs inside the answer body, not just in the heading. Re-elaboration is what makes the passage extractable on its own.
07 · Versioned date stamps and authority signals.
Visible datePublished and dateModified, named author, named methodology source, dated revision history. The engine's recency filter and authority weighting both reward the visible trail.
Information Access vs Cognitive Access — the side-by-side.
Most pages pass the first column. Few pass the second. The gap between the two is the citation gap.
| Dimension | Information Access | Cognitive Access |
|---|---|---|
| What it measures | The page is reachable, parseable, indexable. | A reader (human or model) can extract a self-contained answer. |
| Signals | HTTP 200 · Core Web Vitals green · valid HTML · indexed in Google. | Prototype match · processing fluency ≥ threshold · self-contained passages · entity prominence. |
| Tooling | Lighthouse · PageSpeed Insights · Search Console. | Trunk Test · Six Minds grid · 50ms first-impression test · extractability rubric. |
| Failure mode | Page does not load or does not index. | Page loads, indexes, ranks — and is silently passed over by the AI Overview. |
| Wiele instrument | Site Hygiene baseline (precondition). | Citation Score™ §2 · Cognitive Access Audit. |
Why Lighthouse will not tell you this.
Lighthouse, PageSpeed Insights, Search Console, the entire classical web-perf tool-chain — they all audit the first tier. They ask whether the page loaded, how fast it loaded, whether the HTML validated, whether the schema parsed. They cannot ask whether the answer was extractable, because that question requires a reader. The reader is the missing component in classical web auditing.
The Wiele Cognitive Access Audit is the second-tier audit. It executes the three named tests (Trunk Test, Six Minds, 50ms first-impression) against a client's top buyer-decision pages, scores each page on the seven cognitive-access signals, and outputs a prioritised remediation roadmap. One fixed fee. £1,950. Productised methodology, not consultative hand-waving.
The audit is the front-door product. The instrumentation behind it is the Citation Score™ subscription — Starter (£2k/mo) tracks the citation lift over time; Pro (£4k/mo) tracks the lift against a named competitor set; Authority (£6k/mo) tracks the lift, the competitor set, and the Six Minds grid as a monthly diagnostic on the client's top 10 pages.
Read next · Brief #003
AEO is the 2026 WCAG — accessibility for machine readers.
Brief #003 frames AEO as the structural accessibility discipline of the AI era — an architectural commitment, not a content tactic. Cognitive access is the underlying doctrine; AEO-as-WCAG is the policy framing every agency operator needs to adopt before the next planning cycle.
Read Brief #003 →
What it means for agency owners.
Agency operators recognise this pattern immediately. A client's site is polished. Performance scores are strong. Rankings on the head terms are healthy. The traffic graph looks fine. And the AI Overview, the Perplexity answer, the ChatGPT response — they cite a different brand. The client asks why, and the conventional toolchain has no answer for them, because the conventional toolchain measures information access and the failure is at the cognitive access tier.
This is the diagnostic agencies need to be able to deliver in 2026, and the fix is teachable. The doctrine has a 25-year lineage and a peer-reviewed cognitive-psychology defence. The audit is one fixed fee. The remediation is a schema bundle, a heading rewrite, an entity-rich opener, a re-chunked layout, and a date-stamped versioning discipline. None of it is mystery. All of it is measurable.
The agencies that productise this discipline first will own the diagnosis in their market. The agencies that wait will find their clients hiring the agencies that did not.
The Citation Score™ tie-in.
The Two-Tier Access doctrine is not a positioning idea pulled from nowhere. It is the operating substrate of Citation Score™ §2 — the on-page extractability dimension of the Wiele subscription product. Every Citation Score delivery measures the client's pages against the seven cognitive-access signals; every monthly run scores the drift; every named-competitor comparison surfaces which competitors are out-engineering the client on cognitive access specifically.
The Cognitive Access Audit (£1,950) is the front-door product — fixed-fee, one-off, ICP-friendly. The Citation Score™ subscription is the recurring instrument — £2-6k per month, depending on tier, tracking the lift over time. Both productise the same doctrine. Both are buildable in a sprint. Both are owned positioning that no Wiele competitor is currently shipping.
The compounding loop sits between the two. Audit identifies the gaps; subscription tracks the close; the close becomes the case study; the case study compounds the authority that makes the next audit easier to sell. Information access was the discipline of the 1998-2024 web. Cognitive access is the discipline of the AI-mediated web from 2025 forward.
Methodology & sources.
The Two-Tier Access doctrine combines a primary academic source from 1998 with three secondary sources from the usability and cognitive-engineering canon, instrumented against the Wiele AI Citation Tracker dataset:
Corrao, R. & Fulantelli, G. (c.1998). Cognitive accessibility to information on the Web: insights from a system for teaching and learning Architecture through the Net. Italian National Research Council, Institute for Education and Training Technologies, Palermo — the primary source for the Information Access vs Cognitive Access distinction.
Krug, S. (2014). Don't Make Me Think, Revisited (3rd ed.). New Riders / Peachpit — source for the Trunk Test and the three-laws-of-usability operational doctrine.
Whalen, J. (2019). Design for How People Think. O'Reilly Media — source for the Six Minds audit grid (Vision/Attention, Wayfinding, Memory, Language, Decision- Making, Emotion).
Jones, J., Fitzpatrick, J. & Chassy, P. (2015). The Cognitive Engineering of Memory in Educational Website Design. Advances in Social Sciences Research Journal 2(5), pp.115-121. DOI 10.14738/assrj.25.1147 — peer-reviewed source for the 50ms first-impression window, prototypicality, and the inverted-U complexity curve.
Schema.org — Article, FAQPage, HowTo, Person, BreadcrumbList type specifications.
Google Search Central — Structured-data guidelines, Rich Results eligibility, AI Overview opt-in mechanics.
Wiele AI Citation Tracker dataset — private, anonymised, longitudinal engine-run data feeding the Citation Score™ subscription rubric.
Companion briefs: the Five-Stage Citation Hierarchy (Brief #001) and Stage 3 — Structured Extractability (Brief #002).
The Corrao & Fulantelli paper is a Wiele owned-lineage asset. To the best of the Wiele Citation Tracker's knowledge, no current AEO/GEO agency has yet cited this source in their public positioning. Engagement clients receive the full cognitive-access scorecard against their top buyer-decision pages inside the Citation Score™ dashboard.
Information access is the discipline of the web that loads. Cognitive access is the discipline of the web that gets cited. If you want the cognitive-access scorecard against your live site — plus the seven-signal remediation roadmap — start with the Cognitive Access Audit, or instrument the lift over time with the Citation Score™ subscription. Brief #001 covers the Five-Stage Citation Hierarchy; Brief #002 deep-dives Stage 3 Structured Extractability; Brief #003 frames AEO as the 2026 WCAG.
The next step
Start with a Signal Audit.
A diagnostic that maps your citation graph, entity baseline, and authority gaps — plus a 30-day implementation roadmap. The fastest way to know where you stand inside the answer economy.