Your questions, our answers

An SEO audit primarily answers this question: why is a site not performing correctly in search engines? It examines points such as indexing, technical factors, content, links, tags, the structure of certain pages and the site’s ability to rank.

The digital readability diagnostic adds a broader layer. It answers a different question: is the organization’s digital presence correctly understood, connected and usable by humans, search engines and AI systems? This of course includes some SEO concerns, but it is not limited to them.

Concretely, a digital readability diagnostic can reveal:

• a poorly prioritized offer;
• an ambiguous brand;
• an overly scattered editorial corpus;
• insufficient evidence;
• missing or inconsistent machine surfaces;
• inadequate governance.

An SEO audit remains useful. In some contexts, it is sufficient. But when the problem extends beyond simple organic visibility and touches the overall understanding of the site, brand coherence or machine reading, the digital readability diagnostic gives a more accurate picture of the workstream to launch.

No. One of the diagnostic’s roles is precisely to avoid a complete rebuild when it is not necessary.

Some organizations do need a deep overhaul. This is often the case when the architecture is inherited (what we describe as a fragmented or legacy site), when content has become inconsistent, when several brands or offers are entangled, or when technical and editorial signals contradict each other. In other cases, a more targeted intervention is enough: reorganizing offer pages, clarifying vocabulary, restructuring the corpus, adding governance surfaces or strengthening evidence.

The bad scenario is rebuilding everything because you do not yet know what is actually breaking. The good scenario is starting by measuring where the most costly fault lies, then deciding whether the intervention should be light, structural or deep.

The right starting point is therefore not a reflexive rebuild, but a qualification of the problem.

In many cases, preserving what works and correcting what is missing produces a better outcome than a total reconstruction.

No. LLMs are currently the most visible part of the problem, but not the only one.

When we speak of AI governance and machine reading, we are speaking more broadly about how different systems read, discover, link, rephrase or exploit a digital presence. This can include:

• search engines;
• generative systems;
• assistants;
• recommendation tools;
• agents or discovery layers;
• internal systems that reuse public content or data.

Reducing governance to LLMs alone would be a mistake, because the real problem concerns the coherence of surfaces, signals and reading rules. A site can produce poor answers in an assistant because the structure is weak, but also because the corpus is vague, the entities are unstable or the artefacts of the governance layer are missing.

The right angle is therefore not: “How do we make the site readable for several systems?” The right angle is: “How do we make our digital presence more readable, more stable and more controlled for every machine reader that matters?”

Yes, but WordPress does not define the approach.

We know this environment well, particularly through publishing and maintaining plugins. This knowledge is useful because it provides a concrete understanding of the integration, structure, content, performance and operational constraints in real-world contexts.

That said, the offer does not depend on any particular CMS. The core of the work concerns digital readability, architecture, evidence, entity coherence, machine signals and governance. These challenges exist within WordPress, but also elsewhere.

In other words: if your site runs on WordPress, that is not a problem. If your site does not run on WordPress, that is not a limitation either.

This nuance matters: knowing WordPress is field evidence, not a positioning constraint. What is structural is not the CMS, but the way the digital presence is designed, published and governed.

No. To our knowledge, Lighthouse Agentic Browsing is not a Google Search ranking factor.

Two layers should be separated. Google Search says sites do not need to add llms.txt, Markdown or special machine files to be eligible for AI features in Search. Lighthouse, on the other hand, adds experimental audits that check selected machine-interaction signals.

Those audits can become very useful for web teams, agencies, developers and organizations preparing their sites for agents. But they should not be presented as a ranking promise.

At Pagup, the right use is this: Lighthouse is an input signal. The agentic readiness audit then verifies real journeys, accessibility, forms, CTAs, stability, proof and governance.

In other words: it is not a magic SEO factor, but it is a serious maturity indicator for the agentic web.

Lighthouse Agentic Browsing measures technical signals related to a page’s ability to be understood or used by agents. Public signals include llms.txt, WebMCP, accessibility, programmatic names of interactive elements and layout stability.

It does not measure full agentic readiness. It does not always know whether the offer is clear, whether proof is well connected, whether the right CTA is selected, whether an action consequence is sufficiently explained or whether the site properly governs its limits.

At Pagup, Lighthouse is therefore an entry signal. The Lighthouse Agentic Browsing Audit interprets those signals, then the Agentic Readiness Audit verifies real journeys.

No. A good Lighthouse result is useful, but it is not enough to say a site is truly agent-ready.

Lighthouse can flag important issues: visual stability, accessibility, programmatic names, machine surfaces or WebMCP. These signals should be taken seriously. But they do not always measure offer clarity, proof quality, internal-link coherence, commercial journey precision or the site’s ability to maintain context after interaction.

A site can therefore pass several audits and remain fragile: poorly contextualized form, generic CTAs, services that are too similar, isolated proof, contradictory pages or important content missing from the journey.

At Pagup, a good Lighthouse result is a starting point. Agentic readiness is then verified in journeys: understand, choose, act, confirm and return to the right context.

No. WebMCP should not be mechanically added to every form.

The right first step is to identify actions that truly deserve structured exposure: diagnostic request, booking, add to cart, quote request, registration or another repeatable and well-bounded action.

Before WebMCP, the basics must already be fixed: form title, labels, name attributes, error messages, confirmation, CTA and consequence. If those elements are unclear, WebMCP may only expose a poorly defined action.

The WebMCP and agent-friendly forms audit exists to decide where WebMCP is useful, premature or unnecessary.

The three layers do not play the same role.

llms.txt is a context surface. It helps LLMs and agents find important pages, files, policies and resources on a site. It orients reading.

WebMCP is closer to structured action exposure. When relevant, it can help agents identify functions, capabilities or interactions more explicitly than a classic interface.

An agentic audit is a reliability verification. It checks whether the real site can be understood, navigated and operated correctly: HTML, DOM, visual rendering, accessibility tree, forms, CTAs, proof, internal links, governance and journeys.

The right hierarchy is therefore: llms.txt helps decide what to read, WebMCP can help decide what to execute, and the agentic audit verifies whether the whole system holds up in a real context.

An agent-friendly form is a form whose role, fields, constraints, errors, confirmation and consequence are explicit enough to be understood by an AI agent.

It is not only a visually clean form. It must expose clear intent, labels connected to fields, coherent name attributes, associated error messages, a precise CTA and a confirmation that explains what happens next.

The test is simple: can a system understand what it must fill, why, in which context and what the submission will produce? If the answer depends on human intuition, the form is not agent-friendly yet.

Yes. In many cases, it is in fact the best entry point.

The diagnostic exists precisely to avoid two costly mistakes:

• launching a poorly framed engagement too quickly;
• locking in a hypothesis about the nature of the problem too early.

Some organizations arrive with a very clear and already mature need. Others simply sense that something is wrong: unstable visibility, misunderstood messaging, a site that underperforms despite effort, imprecise AI responses, a confusing structure. In those contexts, jumping straight into a redesign, a content workstream or a governance project would be premature.

The diagnostic clarifies:

• the real problem;
• the starting level of maturity;
• the assets already in place;
• the depth of intervention required;
• the best engagement trajectory.

It can then lead to a targeted workstream, a broader one, or sometimes show that a lighter intervention is sufficient. To learn more, visit the strategic digital readability diagnostic page.

No. SEO is not disappearing. It simply ceases to be the only useful framework for reading the problem.

GEO and AEO remind us that enriched engines, generative systems and other response layers now read sites through uses far broader than classic indexing. This does not make SEO useless. It rather forces us to widen the lens toward an interpretive SEO that also covers AI-based reading.

When the site is structurally weak, neither SEO, nor GEO, nor AEO alone is sufficient. You first need to clarify the offer, strengthen the corpus, surface the evidence and stabilize interpretive governance.

Yes, provided there is a genuine issue of structure, readability or coherence.

The offer is not designed for micro-tactical needs such as “fix two tags” or “quickly gain a few positions.” That said, a specialized SME can be an excellent fit if it already has:

• an offer that deserves better structure;
• a site or content corpus that is becoming an important asset;
• a brand or expertise that needs to be better understood;
• a need to clarify before investing further.

This often applies to specialized B2B SMEs, niche consultancies, advisory firms, technical players or niche businesses that have crossed an initial threshold of digital maturity.

The right criterion is therefore not size alone. It is the nature of the problem, the potential value of the digital asset and the willingness to treat the subject in depth.

An SME looking only for a one-off tactic is not always a good fit. An SME that wants to clarify an important digital asset may very well be one.

Interpretive governance is the discipline that maintains a coherent reading of your organization over time.

Concretely, it concerns the way your pages, your evidence, your brand, your documentation and your machine surfaces say the same thing without contradicting one another.

Without interpretive governance, a site can very well be relaunched, enriched and then become vague again a few months later. With it, new pages, new content and new signals are added to an already framed system.

Interpretive governance is formalized as a versioned doctrine on interpretive-governance.org, with verifiable machine-first artefacts: interpretation policy, output constraints, doctrinal index and AI policies.

Yes, but only if the measurement relies on a baseline. One generated answer is not enough to conclude that drift exists.

The right method begins by defining what the organisation should be able to make understandable: identity, category, offer, audiences, proof and limits. Several systems are then queried with comparable prompts, and the gaps are observed.

The important question is not only whether AI mentions the organisation. It is whether it describes it in the right category, with the right differentiators, the right proof and coherent recommendability.

Once the gap is confirmed, Pagup can determine whether it comes from missing content, brand ambiguity, old remanence, category drift or a deeper governance problem.

Yes, but only if the measurement relies on a baseline. One generated answer is not enough to conclude that drift exists.

The right method begins by defining what the organisation should be able to make understandable: identity, category, offer, audiences, proof and limits. Several systems are then queried with comparable prompts, and the gaps are observed.

The important question is not only whether AI mentions the organisation. It is whether it describes it in the right category, with the right differentiators, the right proof and coherent recommendability.

Once the gap is confirmed, Pagup can determine whether it comes from missing content, brand ambiguity, old remanence, category drift or a deeper governance problem.

Yes, but only if the measurement relies on a baseline. One generated answer is not enough to conclude that drift exists.

The right method begins by defining what the organisation should be able to make understandable: identity, category, offer, audiences, proof and limits. Several systems are then queried with comparable prompts, and the gaps are observed.

The important question is not only whether AI mentions the organisation. It is whether it describes it in the right category, with the right differentiators, the right proof and coherent recommendability.

Once the gap is confirmed, Pagup can determine whether it comes from missing content, brand ambiguity, old remanence, category drift or a deeper governance problem.

Yes, but only if the measurement relies on a baseline. One generated answer is not enough to conclude that drift exists.

The right method begins by defining what the organisation should be able to make understandable: identity, category, offer, audiences, proof and limits. Several systems are then queried with comparable prompts, and the gaps are observed.

The important question is not only whether AI mentions the organisation. It is whether it describes it in the right category, with the right differentiators, the right proof and coherent recommendability.

Once the gap is confirmed, Pagup can determine whether it comes from missing content, brand ambiguity, old remanence, category drift or a deeper governance problem.

The right angle is to optimize the reading terrain, not to chase a specific interface.

If the site is clearer, better structured, better evidenced and better governed through improved digital readability, multiple systems benefit simultaneously: search engines, generative systems, recommendation layers and agents.

Conversely, working to correct a single channel often produces fragile adjustments. The site remains confusing, and the problem reappears elsewhere.

The cost depends primarily on three factors:

1. the depth of the problem;
2. the nature of existing assets;
3. the type of intervention required.

A diagnostic does not carry the same scope as a partial architecture rebuild, a brand disambiguation project or a multi-month strategic advisory engagement. Two organizations that “want to be more visible” may in fact have very different needs.

Rather than quoting a single misleading rate, we prefer to frame the problem first, then propose the appropriate form of intervention. This avoids having an organization pay for a workstream that is too narrow or too broad.

When the need is still unclear, a diagnostic is generally the best way to arrive at a fairer budget recommendation.

In practice, the right question is not just “how much?” but “for what level of transformation?” Sober upstream framing often prevents unnecessary spending later on. Once the context has been qualified, the budget conversation becomes far more useful, because it rests on an actual trajectory rather than a generic estimate.

In practice, the most common entry point, the structured diagnostic, generally starts at a few thousand dollars. Architecture, governance or advisory engagements then scale according to the depth of the problem, the number of assets involved and the expected level of transformation.

The duration depends on the breadth of the context to be read. A diagnostic on a simple site with a single brand and a single offer does not require the same effort as an environment with multiple product lines, several layers of legacy, a large editorial corpus or significant brand ambiguity.

In practice, expect a multi-phase process:

• context gathering;
• reading of existing assets;
• structural analysis;
• synthesis;
• delivery and trajectory recommendation.

The goal is not to produce a decorative document, but a reading that can be acted upon. That is why we favour a diagnostic that illuminates the next steps rather than a simple report that accumulates observations.

The diagnostic must remain proportionate, yet deep enough to prevent a false start. Its duration should therefore be judged by its function: avoiding several weeks or months of poorly oriented execution.

Yes. The challenges of digital readability, structure, disambiguation, documentation and machine reading are not specific to a single market.

The core of the work takes place on the digital assets themselves: the coherence of the offer, the quality of the corpus, the governance surfaces and the intervention trajectory. This can therefore concern organizations located outside Quebec, as long as the working context, language and engagement coordination are compatible.

French is the root language of the site for the current phase, but this does not limit the scope of work to local organizations only.

What matters first is alignment on the way of working and on the depth of the problem to be addressed. Geographic distance changes the nature of the work far less than the quality of the framing and the digital assets to be treated.

SEO remains an important component of digital visibility. It is not set aside; it is placed within a broader framework.

An SEO agency primarily works to improve discoverability in search engines. We work to make a digital presence more readable, more coherent and better governed across multiple reading layers. This can include SEO, but also:

• offer structure;
• entity clarification;
• evidence corpus;
• documentation;
• machine surfaces;
• reading governance.

In some contexts, the right intervention does indeed resemble an expanded SEO engagement. In others, it calls for an architecture rebuild, a brand disambiguation effort or a deeper layer of governance.

The point is therefore not to oppose SEO against everything else. The point is to place it within a broader framework.

Yes, but we favour inspectable evidence over vague promises.

Depending on the nature of the context, results can be shown through:

• published surfaces;
• captures of structures or artefacts;
• maintained tools or plugins;
• typical transformation trajectories;
• observable data;
• before/after comparisons where possible.

The goal is not to produce “client cases” turned into advertising. The goal is to show what was built, what changed and what an outside person can actually verify.

This is the function of the Evidence hub, which should progressively become one of the most important pages on the site.

This logic also protects against client cases turned into opaque sales arguments. A verifiable proof is worth more than a polished story that cannot be examined closely. For an organization seriously evaluating an engagement, this transparency is worth more than a superficial portfolio.

Yes. A recent site is not automatically a readable site.

Many new sites are visually clean, fast and technically acceptable, yet they remain weak on deeper dimensions:

• poorly prioritized offer;
• insufficient evidence;
• unclear vocabulary;
• thin documentation;
• ambiguous brand;
• missing machine surfaces;
• too-weak editorial linking.

In other words, a site’s freshness guarantees neither its coherence nor its ability to be correctly interpreted. In some cases, it is even more cost-effective to intervene early, when the structure can still be corrected without a heavy workstream.

A recent site can therefore absolutely justify a diagnostic if it still gives an impression of vagueness, fragmentation or unmet promise. A site’s newness does not protect against a digital readability problem.

A classic rebuild starts from the visual: new branding, new mockups, new CMS. The result is often more attractive, but the reading structure stays the same. Search engines continue to index poorly connected pages, AI systems continue to infer vague conclusions, and the interpretive debt remains intact beneath the new surface.

A machine-first rebuild takes the problem from the other direction. Before touching the visual layer, you reconstruct what needs to be understood: the site tree, the internal linking, the templates, the structured data, the machine surfaces and the evidence logic. The goal is for the site to be readable across all reading layers simultaneously: human visitors, search engines, generative AI systems and autonomous agents.

The term “AI-first” completes the idea: it is no longer enough to satisfy classic search engines. Generative systems, conversational assistants and comparison agents read your public assets to infer what you are. If the structure is vague, the inference will be too.

In practice, this type of rebuild produces a site that is more stable, easier to maintain and more resilient to changes in reading systems. It is aimed at organizations whose current site has accumulated too many layers, contradictions or noise for a simple refresh to suffice.

To learn more about this workstream, visit the machine-first / AI-first web architecture page.

SEO (Search Engine Optimization) aims to make a site more visible in classic search engine results. It works on keywords, tags, links, speed and perceived authority.

GEO (Generative Engine Optimization) is more recent. It aims to influence the way generative AI systems (ChatGPT, Perplexity, Gemini and others) cite or represent an organization in their responses.

Both disciplines are useful, but each operates on a single reading layer. SEO does not guarantee that AI systems understand you. GEO does not guarantee that your pages support one another or that your brand remains coherent over time.

Interpretive governance replaces neither SEO nor GEO. It sits above them: it is the discipline that maintains the coherence of what all systems (engines, AI, agents) understand from your public assets. It governs structure, evidence, vocabulary, machine surfaces and brand so that the whole remains readable and stable over time.

In summary: SEO and GEO are tactical levers. Interpretive governance is the framework that prevents those levers from contradicting each other. To go further, visit the AI governance and machine readability page.

An agentic navigability audit verifies whether an AI agent can use your site as an action environment. It does not only check whether pages are indexable or fast. It checks whether actions are understandable, stable and properly exposed.

In practice, the audit compares visual rendering, HTML, hydrated DOM and the accessibility tree. It looks for divergences: fake buttons, poorly named fields, unstable menus, layout shifts, ambiguous CTAs, hidden actions and poorly connected forms.

The outcome is a prioritized list of technical corrections. The goal is not to create a site “for robots”, but a site that is more reliable for all its readers: humans, search engines, assistive technologies and AI agents.

The service page describes the full scope: agentic readiness audit.

Not always. WebMCP is a serious direction for exposing structured actions to AI agents, but it remains an emerging layer. The right reflex is to verify the foundations first: semantic HTML, named forms, accessibility, visual stability, internal linking, structured data and governance.

A site with poorly connected forms, ambiguous buttons or a DOM that changes after hydration does not become reliable simply because it adds WebMCP. The tool can reduce ambiguity, but it does not replace a clean interface.

For a SaaS, transactional site, marketplace or website with complex journeys, WebMCP may belong on the roadmap. In that case, the agentic readiness audit helps determine whether the site is ready for that layer or must first correct its action signals.

Interpretive debt is the accumulated gap between what your organization truly is and what reading systems (search engines, generative AI, agents) infer from your public assets.

It accumulates silently. Every vague page, every contradiction between two properties, every missing or outdated machine signal adds a layer of ambiguity. Systems do not ask you for clarification: they infer with what they find. If what they find is incomplete or contradictory, the representation they build of you is too.

Interpretive debt manifests in several ways: AI systems say false things about your organization, your brand is misunderstood, your competitors are cited in your place, or your own properties contradict one another.

The concept is analogous to technical debt in software development: the longer you wait, the more expensive the correction becomes. But unlike technical debt, interpretive debt directly affects how the market perceives you through the systems that mediate it.

To explore the concept further, consult the full definition in the glossary or the issue interpretive debt and the cost of inaction. Gautier Dorval also offers a detailed definition on his personal site.

Yes. A correctly structured site does not stay that way indefinitely. The corpus evolves, AI systems change their reading mechanisms, new surfaces emerge. Without follow-up, the gains from a rebuild or diagnostic degrade over time, and interpretive debt begins to accumulate again.

This is why strategic advisory and transformation includes an ongoing governance dimension. This follow-up consists of:

• maintaining coherence across your digital assets;
• monitoring what AI systems infer from your public surfaces;
• adjusting machine signals when the context changes;
• updating governance artefacts (llms.txt, entity-graph, JSON-LD).

This ongoing steering role aims at interpretive sustainability and prevents the classic scenario where an organization invests in a serious rebuild, then watches the benefits erode for lack of structural maintenance. AI governance and machine readability covers the technical side of this continuity.

The diagnostic is not a technical audit or an automated report. It is a structured framing session, conducted live, whose objective is to produce a usable reading of your situation.

Format and duration

The session lasts between 30 and 60 minutes, depending on the complexity of the context. It takes place over video call. You then receive a written summary with findings, identified priorities and a trajectory recommendation.

Who should participate

Ideally, the person who makes decisions about the digital presence. If you work with a marketing team, a technical lead or an agency partner, their presence is useful but not required.

What you leave with

• a clear reading of your existing assets and their condition;
• identification of structural problems (not just visible symptoms);
• an intervention hierarchy adapted to your context;
• a recommendation on the type of engagement to consider, if appropriate.

The diagnostic may be sufficient on its own or lead to a broader workstream. In all cases, it serves to prevent a false start.

To request a diagnostic, visit the diagnostic page or learn about the strategic digital readability diagnostic.

There is no guaranteed timeline, because the speed of integration depends on several factors you do not fully control: the crawl frequency of models, the depth of your corpus, the coherence of your existing signals and the competition on your topics.

What is observable

As a general rule, initial effects appear between 4 and 12 weeks after the structural interventions are implemented. This includes:

• the appearance or correction of mentions in LLM-generated responses;
• improved fidelity in summaries produced by AI agents;
• more coherent alignment between what you publish and what models return.

Why this timeline

The work of interpretive governance acts on structure, markup and corpus coherence. Language models do not index in real time: they progressively integrate available signals. This is what is called interpretive lag: the delay between what you publish and what systems return.

What accelerates results

• an already structured and coherent corpus;
• existing authority signals (recognized brand, citations, backlinks);
• clear governance of machine surfaces (llms.txt, entity graph, JSON-LD).

To go further, consult the AI governance and machine readability service.

Not everything needs to happen at once. That is precisely the diagnostic’s role: to distinguish what is urgent, what is structural and what can wait.

The diagnostic as a prioritization tool

When the budget is limited, the main risk is dispersing effort across interventions that are visible but not structurally meaningful. The diagnostic produces an intervention hierarchy adapted to your actual context:

• which problems create the most friction today;
• which existing assets can be consolidated without a heavy rebuild;
• which interventions will produce the most lasting effects.

Sequence rather than launch everything

It is entirely possible to work in phases. For example:

1. a diagnostic to frame the situation;
2. an initial targeted intervention on the most structural problem;
3. progressive support for subsequent workstreams.

This approach allows you to validate results at each stage before committing to the next. It also prevents global rebuilds whose scope inflates mid-course.

What this changes concretely

With clear framing, you invest in what actually matters for your situation, not in what a generic model recommends. You keep control of the pace and the budget.

To get started, you can request a diagnostic or consult the strategic advisory and transformation service.