DOCS 01 ANALYSIS METHODOLOGY

How TextRank analyzes content for AI answers

TextRank analyzes competitor content and other available AI-referenced sources for a selected buyer question. It compares them with up to three of your pages and prepares a specific edit plan: passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk), gap, reasoning and priority.

Important: Content analysis supports an AI visibility audit. Its deliverable is an implementable edit plan that complements broader brand-presence assessment, technical review and business measurement.

A separate guide explains report scope, full 7-credit analyses, the shared Gemini / Google AI Overviews result, pilot budgeting and measuring results.

Read the analysis scope and costs guide

Choosing an analysis for yourself, your team or a client? Find practical answers about prompts, implementation, evidence, budgets and agency work.

Read questions from founders, teams and agencies

Content matters, and so does whether customers and AI understand it

TextRank analyzes the presence of important concepts and the meaning of the complete message. A strong passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) clearly names the product, action and conditions. This helps customers and AI systems understand it even outside the context of the full page.

GENERAL CLAIM

“TextRank provides comprehensive content analysis for AI visibility.”

EVIDENCE-BACKED CLAIM

“TextRank compares a page with ChatGPT, Gemini, Claude and Grok answers for a selected promptA user question or instruction for which we analyze the answer and brand visibility.Prompt, then identifies passagesA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) that need data, sources or context.”

SEOWork that helps a search engine find, understand and display a page in results. In GEO, SEO remains the foundation that gives AI systems access to useful content.SEO helps a page get found. GEOWork that helps content answer natural questions clearly and become usable in a generative answer. It combines SEO foundations with readability, credibility and information structure.GEO / AEO helps its content get used.

SEOWork that helps a search engine find, understand and display a page in results. In GEO, SEO remains the foundation that gives AI systems access to useful content.SEO helps make a page accessible, understandable to search engines and aligned with query intentThe goal behind a prompt: definition, problem solving, comparison, recommendation, purchase or risk assessment.Intent. Technically accessible and readable content can enter the set of sources analyzed by an AI system.

GEOWork that helps content answer natural questions clearly and become usable in a generative answer. It combines SEO foundations with readability, credibility and information structure.GEO / AEO and AEOWork that helps content answer natural questions clearly and become usable in a generative answer. It combines SEO foundations with readability, credibility and information structure.GEO / AEO go a step further. They assess whether content answers a natural question, presents credible evidence and makes a specific passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) easy to use in an AI-generated answer.

In short: SEOWork that helps a search engine find, understand and display a page in results. In GEO, SEO remains the foundation that gives AI systems access to useful content.SEO helps a page enter the pool of possible sources. GEOWork that helps content answer natural questions clearly and become usable in a generative answer. It combines SEO foundations with readability, credibility and information structure.GEO / AEO helps prove that your source deserves to be selected.

Analysis step by step

  1. Choose the question and intentThe goal behind a prompt: definition, problem solving, comparison, recommendation, purchase or risk assessment.Intent.Determine what the user is really looking for and which conditions matter, such as budget, industry or market. Select one page or up to three page URLs to analysis for that promptA user question or instruction for which we analyze the answer and brand visibility.Prompt, such as a product page, pricing page and documentation page.
  2. Retrieve and split the content.Read the page structure, check accessibility and divide the text into self-contained passagesA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk).
  3. Identify entitiesAn unambiguously named thing or concept, such as a brand, product, service, audience, method, competitor or outcome.Entity and claimsA single statement that can be checked and supported by a source, such as “the product integrates with Shopify” or “the Pro plan costs X”.Claim.Find products, problems, features, outcomes and individual statements that should be supported.
  4. Calculate RetrievalThe stage where we assess whether a specific URL can become a candidate source for a given prompt.Retrieval.Assess whether a specific URL fits the question and can enter the candidate source pool.
  5. Collect competing sources.Check which pages answer the same question and are surfaced by AI models.
  6. Calculate SelectionThe stage where found sources are compared to identify which one supports a specific claim best and most safely.Selection.Compare source passagesA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) and identify which ones best support each important claimA single statement that can be checked and supported by a source, such as “the product integrates with Shopify” or “the Pro plan costs X”.Claim.
  7. Choose a passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) to improve.Define the additions needed: specificity, evidence, context, freshness or a clearer format.
  8. Run a re-test.Compare the result before and after the change, then monitor mentions and citations over time.

How we select passages from your page and other sources

TextRank compares competitor content and other available AI-referenced sources with your page for a selected buyer question. You can select up to three of your URLs. That limit concerns your pages, not the model’s sources. The result identifies a passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk), gap, reasoning and edit plan.

  1. Keep the answer in context.The analysis records the promptA user question or instruction for which we analyze the answer and brand visibility.Prompt, selected model, date, answer and sources. ChatGPT, Claude, Gemini and Grok results stay separate because they may surface different material.
  2. Split the selected pages and sources into passagesA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk).We examine text from each selected page and available sources cited by the model. We look for passagesA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) relevant to the question and the answer's claimsA single statement that can be checked and supported by a source, such as “the product integrates with Shopify” or “the Pro plan costs X”.Claim. For a long document, candidate selectionThe stage where found sources are compared to identify which one supports a specific claim best and most safely.Selection may cover only part of the text; interpret the result with that scope in mind.
  3. Compare answers, not just matching words.We assess whether a passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) answers directly, names the product and conditions, gives a verifiable fact and makes sense away from the full page. Your passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) and the cited source are assessed in the same promptA user question or instruction for which we analyze the answer and brand visibility.Prompt and model context.
  4. Turn the difference into a decision.The result may point to a passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) to keep, clarify or support with evidence. If a fact is missing from your materials, the recommendation should call for verification before publication.
Example: A buyer asks about CRM integrations. A cited source names supported systems and sync conditions. Your page says only “easy integration.” The finding identifies information a buyer needs to choose; it does not prove why the model cited the other page.

What the score means

RetrievalThe stage where we assess whether a specific URL can become a candidate source for a given prompt.Retrieval, SelectionThe stage where found sources are compared to identify which one supports a specific claim best and most safely.Selection and Citation OpportunityA relative indicator of a passage's potential for use after retrieval, selection and passage assessment. It sets the priority for content work.Citation Opportunity are TextRank assessments that help prioritize content work. They combine model-based assessment with stated criteria and product rules. They are not a reading of ChatGPT's internal ranking or a promise of a citation. The useful output is the passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk), the reason and the next step you can check.

What the analysis cannot determine

The report shows an observed answer and differences between available pages. It cannot see a provider's full internal source-selectionThe stage where found sources are compared to identify which one supports a specific claim best and most safely.Selection process. A source may be unavailable, and model answers can change over time. After an edit, repeat the analysis in comparable conditions and distinguish a better page from a changed model answer.

See what data a TextRank analysis uses

From question to citation in three stages

This is TextRank's assessment model: we examine how well a page fits the question, how strongly it supports a claimA single statement that can be checked and supported by a source, such as “the product integrates with Shopify” or “the Pro plan costs X”.Claim and how useful a passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) is on its own. Each stage helps prioritize an edit. It is not a description of every AI provider's internal algorithm.

RetrievalThe stage where we assess whether a specific URL can become a candidate source for a given prompt.RetrievalSelectionThe stage where found sources are compared to identify which one supports a specific claim best and most safely.SelectionCitation OpportunityA relative indicator of a passage's potential for use after retrieval, selection and passage assessment. It sets the priority for content work.Citation Opportunity
01

RetrievalThe stage where we assess whether a specific URL can become a candidate source for a given prompt.Retrieval

Will the page enter the source pool?

For a given promptA user question or instruction for which we analyze the answer and brand visibility.Prompt, we assess each selected URL, with up to three of your pages per promptA user question or instruction for which we analyze the answer and brand visibility.Prompt. We check topic and intentThe goal behind a prompt: definition, problem solving, comparison, recommendation, purchase or risk assessment.Intent fit, important concepts, freshness, page depth and technical accessibility. The result describes each URL's strength as an answer candidate and identifies areas that need development.

02

SelectionThe stage where found sources are compared to identify which one supports a specific claim best and most safely.Selection

Why should AI choose this source?

We break the answer into individual claimsA single statement that can be checked and supported by a source, such as “the product integrates with Shopify” or “the Pro plan costs X”.Claim, such as a feature, price or implementation time, and compare sources for each one. We check whether a passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) supports the claimA single statement that can be checked and supported by a source, such as “the product integrates with Shopify” or “the Pro plan costs X”.Claim directly, contains concrete data and comes from an appropriate source. The result identifies the strongest evidence and the elements that would improve your content against competitors.

GENERAL CLAIM

“Advanced Shopify integration.”

EVIDENCE-BACKED CLAIM

“The Shopify integration synchronizes orders, customers and products every five minutes; it is available from the Pro plan.”

The second passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) states the scope, frequency and condition. The model gets a ready-to-use fact with the context required for an answer.

03

Citation OpportunityA relative indicator of a passage's potential for use after retrieval, selection and passage assessment. It sets the priority for content work.Citation Opportunity

Which passage can be used?

We select the passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) that best supports the answer after RetrievalThe stage where we assess whether a specific URL can become a candidate source for a given prompt.Retrieval and SelectionThe stage where found sources are compared to identify which one supports a specific claim best and most safely.Selection. We check directness, complete context, evidence and whether the passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) keeps its meaning outside the page. The result sets the order of changes and shows the passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk)'s potential for use in an AI answer.

The full model at a glance

StageQuestion we answerWhat can be improved
RetrievalThe stage where we assess whether a specific URL can become a candidate source for a given prompt.RetrievalIs this page a strong candidate for the promptA user question or instruction for which we analyze the answer and brand visibility.Prompt?URL topic, intentThe goal behind a prompt: definition, problem solving, comparison, recommendation, purchase or risk assessment.Intent, entitiesAn unambiguously named thing or concept, such as a brand, product, service, audience, method, competitor or outcome.Entity, structure, freshness and technical accessibility
SelectionThe stage where found sources are compared to identify which one supports a specific claim best and most safely.SelectionWhy should this source support the claimA single statement that can be checked and supported by a source, such as “the product integrates with Shopify” or “the Pro plan costs X”.Claim?Directness, specificity, evidence, method, first-party data and source credibility
Citation OpportunityA relative indicator of a passage's potential for use after retrieval, selection and passage assessment. It sets the priority for content work.Citation OpportunityWhich passageA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) is best suited for use?Answer-first structure, context, conditions, list/table/FAQ format and readable HTML

From customer questions to a ready-to-use action plan

01

Customer prompt discovery

What are potential customers asking AI systems?

We analyse your brand, product, audience and market language. From that context, we discover and generate natural questions about problems, choices, pricing, implementation, risk, comparisons and alternatives, then prioritise them by intentThe goal behind a prompt: definition, problem solving, comparison, recommendation, purchase or risk assessment.Intent and business value.

02

AI competitor analysis

Which sources do Claude, ChatGPT, Gemini and Grok surface and reference?

We review model responses for selected promptsA user question or instruction for which we analyze the answer and brand visibility.Prompt, identify the companies, domains and pages they mention, and show what distinguishes those sources: clearer answers, stronger data, better examples, structure, entitiesAn unambiguously named thing or concept, such as a brand, product, service, audience, method, competitor or outcome.Entity or credibility.

03

Content gap verification

What is your page missing compared with referenced sources?

We compare your page sentence by sentence with competing content. We identify missing questions, keywords, entitiesAn unambiguously named thing or concept, such as a brand, product, service, audience, method, competitor or outcome.Entity, facts, evidence and sections, and show which vague or incomplete answers make the content harder to use for a specific promptA user question or instruction for which we analyze the answer and brand visibility.Prompt and model.

04

Content and change plan

What should you keep, rewrite or add to make the answer stronger?

We create a prioritised implementation plan covering passagesA self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.Passage (chunk) to improve, new content opportunities, answer-first responses, and places for data, sources, examples, tables and FAQs. You get specific tasks, ready to execute or delegate.

Definitions used in the report

SEO
Work that helps a search engine find, understand and display a page in results. In GEO, SEO remains the foundation that gives AI systems access to useful content.
GEO / AEO
Work that helps content answer natural questions clearly and become usable in a generative answer. It combines SEO foundations with readability, credibility and information structure.
Prompt
A user question or instruction for which we analyze the answer and brand visibility.
Intent
The goal behind a prompt: definition, problem solving, comparison, recommendation, purchase or risk assessment.
Entity
An unambiguously named thing or concept, such as a brand, product, service, audience, method, competitor or outcome.
Claim
A single statement that can be checked and supported by a source, such as “the product integrates with Shopify” or “the Pro plan costs X”.
Passage (chunk)
A self-contained content unit: an answer-led heading, paragraph, list item, table row, FAQ or case-study excerpt.
Retrieval
The stage where we assess whether a specific URL can become a candidate source for a given prompt.
Selection
The stage where found sources are compared to identify which one supports a specific claim best and most safely.
Citation Opportunity
A relative indicator of a passage's potential for use after retrieval, selection and passage assessment. It sets the priority for content work.
Extractability
How easily a passage can be separated without losing meaning, which is important when creating concise answers and citations.
Semantic gap
The difference between analyzed content and cited competitors across important information, entities, relationships, questions or formats.