.png)
What Does PageLens.ai Actually Measure and Fix? Our PageLens.ai Methodology
TL;DR
Our PageLens.ai Methodology measures how an approved brand appears in defined AI answers, then connects evidence to content and technical opportunities. We show the metric formulas, evidence controls, workflow, implementation boundaries, and enterprise questions teams should resolve before treating an AI-visibility signal as a fix.
What Does PageLens.ai Actually Measure and Fix? Our PageLens.ai Methodology
AI answers can vary even when the prompt stays the same. A 2026 study notes that identical queries can return different responses and cited sources, which is why a useful visibility report needs a documented measurement boundary.
Our PageLens.ai Methodology measures how an approved brand appears in defined AI answers: whether it is mentioned, recommended, cited, described positively or negatively, and represented in the category conversation. We record the evidence behind each result, turn recurring gaps into content or technical recommendations, and separate work we can deliver from changes that need approval or implementation.
This page explains our identity, inputs, metrics, evidence controls, improvement loop, technical scope, and enterprise decision criteria.
What Does PageLens.ai Measure, and What Does It Not Measure?
We are PageLens.ai, published at pagelens.ai. We measure selected AI answers about a brand and its category. We are not the similarly named .com website-audit product, which focuses on browser-rendered site checks and reports an 11-category inspection framework.
Our work starts with what a buyer asks an AI engine and what that engine returns. We measure answer-level evidence, not every private AI conversation, total market demand, organic rankings, referral traffic, or revenue. A visibility change can identify a useful opportunity, but it does not prove a commercial outcome on its own.
We use the result to make a practical next decision. That may be to investigate a missing entity signal, improve a source page, create a content brief, publish an approved page, audit a technical issue, or leave the finding alone because the evidence is too weak. Teams that need a stronger prompt set can start with a buyer prompt dataset.
Which AI Answers Does Our PageLens.ai Methodology Track?
A measurement begins with an approved buyer prompt, an answer engine, and a scheduled run. We preserve the prompt text, engine, timestamp, and returned answer so the number in a dashboard remains tied to inspectable evidence.
Our public plans currently describe a 50-prompt daily ChatGPT baseline, a 100-prompt daily plan across ChatGPT, Google AI, and Perplexity, plus a 100-prompt daily managed plan across seven listed answer engines. That broader set includes ChatGPT, Claude, Gemini, Perplexity, Grok, Copilot, and Google AI. Enterprise coverage, prompt volume, refresh cadence, and engine mix are custom.
We record locale, language, and geographic context whenever they are set for a run. We do not treat a localized answer as a universal result. Public plan pages do not state default geography, default language, or a fixed answer-history retention period, so those settings belong in the customer’s written scope. For the mechanics behind cross-model collection, see how we track brands across AI engines.
A missing or unusable answer is not silently counted as a missed mention. We label it as missing, show the count, and exclude it from the eligible-answer denominator for percentage metrics.
How Does the PageLens.ai Methodology Calculate Each Metric?
We define metrics before interpreting them. That discipline prevents a broad label such as “visibility” from hiding different inputs, denominators, or comparison sets. Our methodology follows the same general principle in the NIST framework: document the methods, metrics, limitations, and conditions behind a measurement.
| Metric | What We Count | Formula or Treatment |
|---|---|---|
| Visibility | Eligible answers with an approved brand mention | Mentioned eligible answers ÷ all eligible answers × 100 |
| Recommendation Position | Explicitly ordered recommended brands | Ordinal position in the answer, or “not ranked” when no order exists |
| Citation Rate | Eligible answers citing an approved brand property | Citing eligible answers ÷ all eligible answers × 100 |
| Sentiment | The exact extracted language about the approved brand | Positive, neutral, negative, mixed, or insufficient evidence |
| Share Of Voice | Counted brand mentions or recommendations in a defined set | Brand events ÷ all included brand events × 100 |
| Category Average | Visibility rates for included entities | Arithmetic mean, with prompt and entity counts displayed |
What Counts as a Brand Mention?
A mention is a matched reference to an approved brand entity, domain, product name, or documented alias. We separate a mention from a citation because an answer can name a brand without citing its website. We also separate an explicit recommendation from a mere list appearance.
When a name is shared, abbreviated, or used by a subsidiary, we do not assume a match. We use approved entity rules, flag ambiguous references, and send uncertain results to review. This is why a raw answer matters more than a single score.
What Does Citation Rate Show?
Citation rate shows whether an answer visibly cites an approved brand property. It does not measure every source the model may have learned from, and it does not prove that a citation caused a recommendation.
We retain the cited URL or source domain where the answer exposes one, then compare it with the brand’s approved properties. Teams can use that evidence alongside citation tracking to distinguish a source gap from a mention gap.
How Do We Treat Sentiment and Share of Voice?
Sentiment is a label for the language the answer uses about a brand, not a claim about customer satisfaction. We preserve the phrase that produced the label so a team can see whether “expensive,” “easy to use,” or another description is accurate, outdated, or context-dependent.
Share of voice is comparative. We divide a brand’s counted mentions or recommendations by all counted events in the defined prompt, engine, date, and entity set. We show the comparison set because a percentage without its denominator can create false confidence.
How Do We Turn Measurement into Content and Technical Work?
We use a seven-stage loop: prompt selection, scheduled runs, raw-answer storage, extraction, scoring, recommendations, managed delivery, and follow-up measurement. The point is not to manufacture a score. It is to create a repeatable path from observed answer language to a justified next action.
How Do We Store and Extract Evidence?
We capture the returned answer before extracting entities, recommendation positions, sentiment, and citations. Each result carries its run timestamp and available engine context. If extraction confidence is low, or a name could refer to more than one entity, we flag rather than overstate the result.
This record lets a customer correct an entity mapping, challenge a classification, and see which answer created the finding. It also gives our team a clearer starting point for B2B buyer-prompt research.
What Do We Recommend?
A recommendation is not a fix. It is a proposed action tied to observed evidence, such as a missing page type, an incomplete explanation, weak entity consistency, an unaddressed objection, or a technical issue found during audit work.
Our managed content workflow can take an approved brief through draft, review, and publication on the customer’s domain. We then remeasure the defined prompt set, rather than declaring success because a page was published.
What Can We Diagnose, Manage, and Implement?
We distinguish monitoring and diagnosis from customer-approved implementation. This is especially important when a team asks us to resolve technical SEO work without repeated handoffs.
| Work Type | Our Methodology Treats It As | Required Evidence or Handoff |
|---|---|---|
| AI-answer monitoring | Measured evidence | Stored prompt runs and extracted results |
| Diagnosis | Analysis | Source, content, entity, or technical finding |
| Recommended action | Proposed work | Prioritized brief or remediation plan |
| Managed content production | Deliverable | Customer review and publication approval |
| Technical audit | Deliverable | Site evidence and agreed audit scope |
| Implemented technical fix | Confirmed outcome | Access, approval, deployment owner, and verification |
Our public managed scope lists technical audits and fixes, but implementation ownership, access, remediation catalogue, and service expectations must be agreed in writing. Technical findings also need their own evidence. Google’s published thresholds define good Core Web Vitals as LCP within 2.5 seconds, INP below 200 milliseconds, and CLS below 0.1, as explained in Google’s thresholds.
AI visibility monitoring and technical SEO are related, but they are not interchangeable. A crawl issue, a slow template, and a missing buyer answer can each need different owners. See AI visibility versus SEO monitoring for the operating distinction.
How Do Enterprise Teams Apply This Methodology Across Brands?
Enterprise teams need evidence that remains useful after the first dashboard review. For a multi-brand program, each brand needs its own approved entity map, prompt set, comparison set, markets, and permissions. Otherwise a parent company, subsidiary, and regional product can be incorrectly combined into one result.
We support custom enterprise coverage, multi-site rollout support, dedicated onboarding, and white-labeling as part of our published enterprise offering. We do not represent regional views, role-based permissions, data retention, audit logs, SSO, or support terms as standard capabilities unless they are explicitly included in the agreed scope.
| Evaluation Question | What To Confirm With Us | Why It Matters |
|---|---|---|
| Brand Separation | Canonical domains, subsidiaries, aliases, and exclusions | Prevents cross-brand misattribution |
| Regional Measurement | Locale, language, geography, and engine availability | Makes market comparisons reproducible |
| Evidence Access | Raw answers, timestamps, confidence flags, and corrections | Lets teams audit a reported change |
| Technical Ownership | Access, approvals, deployment owner, and verification | Separates diagnosis from a completed fix |
| Retention And Support | Storage period, permissions, support route, and escalation | Establishes operational accountability |
Before rollout, ask to see the methodology version, correction process, change log, and historical-data policy. Our enterprise deployment checklist helps teams establish those controls before rollout. Our change log should state the date, version, affected metric, rationale, and whether historical results were recalculated.
Put PageLens.ai to Work
PageLens.ai is for marketing, growth, SEO, and content leaders who need a practical view of what AI buyers are being told, plus a disciplined way to act on it. We start with the prompts and engines that matter to your category, preserve the answers behind every result, and make the next decision clear: investigate, publish, approve, or implement. For lean teams, that means fewer speculative content requests. For larger teams, it means a shared evidence trail across brands, markets, and stakeholders. We can show you how the workflow fits your operating model, what scope is included, and which technical changes need access or a separate implementation agreement. Bring a real buyer prompt, a priority market, and the site you want to improve. We will use the conversation with our team to define the measurement boundary before promising a result. Book a demo
FAQs on PageLens.ai Methodology
What Does PageLens.ai Measure?
We measure approved brand appearances in defined AI answers, including mentions, recommendation position, citations, sentiment, competing entities, visibility, share of voice, and category comparison metrics.
Does PageLens.ai Fix Technical SEO Issues?
Our managed scope includes content production and technical audits and fixes, but implementation ownership, access, approvals, and exact remediation scope must be agreed in writing.
How Is Share of Voice Calculated?
We calculate share of voice by dividing your counted mentions or recommendations by all counted brand mentions or recommendations in the defined comparison set, then multiplying by 100.
Why Can AI Answers Change for the Same Prompt?
AI answer engines can return different results for identical prompts over time. We preserve prompt, engine, timestamp, answer context, and confidence flags so changes remain interpretable.
How Do Enterprise Teams Prepare a Rollout?
Enterprise teams should define brand entities, markets, prompts, evidence access, approvals, retention, and support terms before rollout. Our enterprise monitoring guide helps structure that conversation early.
