Liminal

AEO Readiness Scorecard

In 15 minutes and 40 questions, it measures how ready your company is to be cited by AI assistants (ChatGPT, Claude, Gemini and Perplexity), and tells you where to start.

A companion tool to the book Be the Answer, by Tiago J. C. Sousa.

Your progress is saved in this browser, so you can stop and continue later.

  • 40 questions across five dimensions
  • About 15 minutes
  • A score from 0 to 100, a level and your weakest dimension
  • Free, with no sign-up: the result appears on screen

What you get at the end

Example with fictional answers.

Result screen with the score, the four-level scale, the answers that count as zero and the bars for the five dimensions.
No sign-upYour score from 0 to 100, your level, how far you are from the next one and the five dimensions, with the weakest flagged.
Result screen with the first five actions and the 90-day plan in three blocks of weeks.
No sign-upYour first five actions, chosen with the rules in chapter 9, and the 90-day plan for your level.
Full report with the gaps in priority order, each with the answer given, how to fix it and the common mistake.
With emailThe full report: every gap, with how to fix it, the most common mistake and the evidence level.

Why do this now

B2B buyers have started building their vendor shortlists inside AI assistants. And the initial list decides almost everything.

In the first diagnostics run with the scorecard, with B2B companies in Portugal, all of them landed at the Invisible level (chapter 3).

What the scorecard measures

  1. TTechnical accessibilityCan engines read the website?8 questions · 15 points out of 100
  2. EEntity and brandDo engines know who we are and describe us well?8 questions · 20 points out of 100
  3. CCitable contentAre our pages easy to cite?10 questions · 25 points out of 100
  4. OOff-site presenceDo third parties talk about us where engines look?7 questions · 20 points out of 100
  5. MMeasurement and operationsDo we know what is happening and what it is worth?7 questions · 20 points out of 100

Before you start

  • Answer with the person who manages the website and the person who manages the CRM in the same room.
  • Check instead of assuming: every question has a 5-minute "how to check".
  • Have access to the website (robots.txt and sitemap), Google Search Console, GA4 and the CRM at hand: that is where each "how to check" happens.
  • Save the result and repeat it at 90 days.

Step 1 of 5

T Technical accessibility

Can engines read the website?

Weight: 15 points out of 100 · Chapter 4 of the book: Be readable

How to answer

0
Does not exist
1
Exists in part or without a routine
2
Exists, complete and with an owner
?
I don't know (counts as 0)

The rule "I don't know counts as zero" is deliberate. If nobody in the company knows whether the website blocks AI crawlers, in practice nobody is making sure it doesn't.

T1 AI crawlers can get in Does your site's robots.txt file let AI assistant crawlers (ChatGPT, Claude, Perplexity) in, as a deliberate decision?
How to check

Open yourdomain.com/robots.txt and look for "Disallow" tied to any of these names or to "User-agent: *".

Learn more

What it is. The robots.txt file states which crawlers may read the site. AI assistants use their own crawlers: GPTBot and OAI-SearchBot (OpenAI), ChatGPT-User (requests made during a conversation), PerplexityBot and Perplexity-User (Perplexity), ClaudeBot, Claude-SearchBot and Claude-User (Anthropic) and Google-Extended (a signal, not a crawler, that controls the use of content to train and ground Gemini; it does not affect Google Search or AI Overviews, according to Google (developers.google.com)). Anthropic separates the three: ClaudeBot collects content for training, Claude-SearchBot indexes for search and Claude-User reads pages at a user's request (Anthropic (support.claude.com)).

Why it matters. A blocked crawler does not read, and what is not read is not cited. Evidence: C.

Chapter 4 of the book: Be readable

T2 Site protection does not stop them Does the site's protection (firewall, CDN, anti-bot) let AI crawlers through without verification screens?
How to check

Review the bot rules in the CDN, hosting or security plugin. isitagentready.com (isitagentready.com) shows the rules declared for AI crawlers, but it is not designed to detect firewall challenges.

Learn more

What it is. Firewalls, CDNs, hosting and security plugins can show verification screens ("One moment, please...") to any automated request, including crawlers allowed in robots.txt.

Why it matters. robots.txt says the crawler may come in; the firewall decides whether it actually does. Evidence: A (Cloudflare (developers.cloudflare.com)).

Chapter 4 of the book: Be readable

T3 The text exists without JavaScript Does the page text appear even without JavaScript enabled (without depending on scripts loading)?
How to check

In the browser, use "View page source" and search for a sentence from the main text. If it is not there, it depends on JavaScript.

Learn more

What it is. Some pages only show their text after the browser runs scripts. Many AI crawlers read only the initial HTML.

Why it matters. If the text only appears with JavaScript, the crawler sees an empty page. Evidence: C.

Chapter 4 of the book: Be readable

T4 Pages are indexed, once each Are the main pages indexed in Google and Bing, without duplicates?
How to check

Google Search Console (pages report) and Bing Webmaster Tools. Search "site:yourdomain.com" and see whether both versions of the domain appear.

Learn more

What it is. The main pages are in the Google and Bing indexes, with no duplicates (with and without www, with and without a trailing slash).

Why it matters. AI Overviews draw on Google's index and Copilot on Microsoft's. Duplicates split authority between two versions of the same page. Evidence: C.

Chapter 4 of the book: Be readable

T5 The sitemap is alive Does the site have an up-to-date XML sitemap submitted to Google and Bing?
How to check

Open yourdomain.com/sitemap.xml; confirm the dates are recent; confirm it has been submitted in Search Console and Bing Webmaster Tools.

Learn more

What it is. An XML file that lists the site's pages and the date each one was last changed.

Why it matters. It tells the engines what exists and what has changed, which supports content freshness (criterion C8). Evidence: C.

Chapter 4 of the book: Be readable

T6 Pages are fast on mobile Do the main pages pass Core Web Vitals on mobile (PageSpeed Insights)?
How to check

PageSpeed Insights (pagespeed.web.dev) in mobile mode, for the homepage and the three main service pages.

Learn more

What it is. Core Web Vitals measure loading (LCP), responsiveness to interactions (INP) and visual stability (CLS). The thresholds considered good are LCP up to 2.5 seconds, INP up to 200 milliseconds and CLS up to 0.1 (web.dev (web.dev)).

Why it matters. They are a quality signal for Google, and slow pages give a worse experience to people arriving from an answer. Evidence: A.

Chapter 4 of the book: Be readable

T7 Structured data is correct Does the site have valid structured data (schema): organization, authors, articles, FAQs?
How to check

Rich Results Test and Schema Markup Validator on the homepage, a service page and an article.

Learn more

What it is. Code (schema.org, in JSON-LD) that describes the page for machines: the organization, the authors, the articles, the services, the frequently asked questions.

Why it matters. It helps the engines understand who the company is and what each page contains, without guessing. Evidence: A (Google Search Central (developers.google.com)).

Chapter 4 of the book: Be readable

T8 Made-up URLs do not return errors Do you detect and redirect URLs on your site that AI assistants invent (404 errors coming from AI)?
How to check

In GA4, cross pages with an error title against the "AI Assistant" channel. In the server logs, filter 404s by assistant referrer.

Learn more

What it is. AI assistants sometimes cite site addresses that do not exist, built from the names of the services. Whoever clicks lands on a 404 error.

Why it matters. You lose a visit that arrived already recommended. Evidence: C.

Chapter 4 of the book: Be readable

Step 2 of 5

E Entity and brand

Do engines know who we are and describe us well?

Weight: 20 points out of 100 · Chapter 5 of the book: Be recognized

E1 The same name everywhere Are the company's name, description and category the same on the site, LinkedIn, Google and directories?
How to check

Open the four main profiles side by side and compare the name, the description sentence, the founding year and the key figures.

Learn more

What it is. The same name, short description and category on the website, LinkedIn, Google Business Profile, partner directories and review platforms.

Why it matters. The model pulls information together from several sources. If each one says something different, it either fails to connect them or connects them wrongly. Evidence: C.

Chapter 5 of the book: Be recognized

E2 An About page with facts Does the About page state, with verifiable facts, what you do, for whom, where and since when?
How to check

Does the page exist, is it in the menu, and can every sentence be confirmed (year, number of clients, team, sectors, address)?

Learn more

What it is. A page that states, in verifiable sentences, what the company does, for whom, where and since when.

Why it matters. It is the official source the assistant looks for when asked "who is company X". Evidence: C.

Chapter 5 of the book: Be recognized

E3 A complete Google Business Profile Is the Google Business Profile complete, with recent reviews?
How to check

Search the company name on Google Maps and see whether the profile exists, is verified and has reviews from the last six months.

Learn more

What it is. The company's profile on Google, with category, description, hours, website and reviews.

Why it matters. It is one of the structured sources Google uses about the company, including in AI-generated answers. Evidence: C.

Chapter 5 of the book: Be recognized

E4 A Wikidata entry Does the company have an entry on Wikidata (or Wikipedia)?
How to check

Search the name on wikidata.org.

Learn more

What it is. Wikidata is the open database that feeds Wikipedia and many knowledge systems. An entry links the company name to structured facts and to its official profiles.

Why it matters. It helps separate the company from other entities with the same name. Evidence: C.

Chapter 5 of the book: Be recognized

E5 Authors with names and credentials Do articles have an identified author, with their own page, credentials and a link to LinkedIn?
How to check

Open three articles and follow the author's name.

Learn more

What it is. Every article has an identified author, with their own page, role, experience and a link to LinkedIn.

Why it matters. Google steers toward content with demonstrated experience, expertise, authoritativeness and trust, and the author is part of that proof (Google Search Central (developers.google.com)). Evidence: A.

Chapter 5 of the book: Be recognized

E6 Service pages with facts Do service pages give concrete facts (industries, integrations, timelines, prices) instead of adjectives?
How to check

On a service page, count the verifiable facts and the adjectives ("innovative", "fast", "leading"). If there are more adjectives, it fails.

Learn more

What it is. Each service page states which sectors, which platforms, what timeline and what price range.

Why it matters. The assistant compares vendors on the facts it finds. Adjectives cannot be compared. Evidence: C.

Chapter 5 of the book: Be recognized

E7 An honest comparison Do you have an honest comparison page with the alternatives or competitors?
How to check

Search the site for "vs", "alternatives" or "comparison".

Learn more

What it is. A page that compares the company with the alternatives the client is considering, including when the company is not the best choice.

Why it matters. Many buying questions are comparisons ("X or Y", "alternatives to X"). Whoever publishes the most useful comparison has a better chance of being the source. Evidence: C.

Chapter 5 of the book: Be recognized

E8 AI assistants describe the company correctly When you ask ChatGPT, Claude, Gemini and Perplexity about the company, is the description correct?
How to check

Ask five questions in each assistant, in a session with no history: "What does company X do?", "Where is it based?", "What services does it offer?", "Who are X's competitors?", "Is X a good option for [need]?". Record every error.

Learn more

What it is. What ChatGPT, Claude, Gemini and Perplexity answer when asked about the company.

Why it matters. A wrong description is repeated in every conversation. It is the only criterion that triggers an alert in the scorecard. Evidence: C.

Chapter 5 of the book: Be recognized

Step 3 of 5

C Citable content

Are our pages easy to cite?

Weight: 25 points out of 100 · Chapter 6 of the book: Be citable

C1 The answer comes first Do the main pages answer the central question within the first two sentences?
How to check

Read the opening of five main pages. If the first sentence is a slogan, a notice or a generic introduction, it fails.

Learn more

What it is. The page's central question is answered in the first two sentences. Context comes after.

Why it matters. According to an analysis by Kevin Indig (Growth Memo), 44.2% of ChatGPT citations come from the first 30% of the page (Kevin Indig, via Nectiv (nectivdigital.com)). Evidence: B.

Chapter 6 of the book: Be citable

C2 Sections that answer on their own Are section headings questions, and does the first paragraph of each one answer on its own?
How to check

Read each heading and only the paragraph that follows. Does it make sense in isolation?

Learn more

What it is. Section headings are the client's questions, and the first paragraph of each section answers without depending on the rest of the page.

Why it matters. Engines extract pieces, not whole pages. HubSpot recommends that the first paragraph of each section answer the question implied by the heading in one or two self-contained sentences (HubSpot (blog.hubspot.es)). Evidence: B.

Chapter 6 of the book: Be citable

C3 Numbers with sources Do important claims have numbers and cite their source?
How to check

On a service page, count the numbers with a source and the numbers without one.

Learn more

What it is. Important claims carry a number and the source of that number.

Why it matters. In the paper that gave GEO its name, adding statistics and citing sources increased visibility by up to 40% in controlled experiments (Aggarwal et al. (arxiv.org)). Pages with numbers had 62% more average influence on answers (arXiv 2604.25707 (arxiv.org)). Evidence: A.

Chapter 6 of the book: Be citable

C4 Comparisons in tables Are comparisons and specifications in tables?
How to check

Look for HTML tables on the service and comparison pages. Images of tables do not count.

Learn more

What it is. Specifications, plans, timelines and comparisons appear in tables.

Why it matters. ChatGPT citations are 2.3 times more likely to include a table than an individual Google result (Nectiv (nectivdigital.com)). Comparative content had 55% more average influence (arXiv 2604.25707 (arxiv.org)). Evidence: B.

Chapter 6 of the book: Be citable

C5 Specific language Is the text specific and assertive, without vague expressions like 'innovative solutions' or 'many tools'?
How to check

Underline the vague expressions on three pages. More than one per paragraph is too many.

Learn more

What it is. Affirmative, concrete sentences, without vague expressions like "innovative solutions", "many tools" or "the platform".

Why it matters. 36.2% of cited text uses definitive language, compared with 20.2% of uncited text (Kevin Indig, via Nectiv (nectivdigital.com)). Evidence: B.

Chapter 6 of the book: Be citable

C6 Proprietary data Do you publish at least one proprietary study, survey or dataset per year?
How to check

Look on the site for a study, survey or report with proprietary data from the last 12 months.

Learn more

What it is. At least once a year, the company publishes data that only it has: a survey, a benchmark, the average of its projects.

Why it matters. Original data is the strongest reason for an engine to cite one company instead of another, because it does not exist anywhere else. Evidence: C.

Chapter 6 of the book: Be citable

C7 The follow-up questions are answered Does the content also answer the sub-questions a customer would ask after the main question?
How to check

Run three main questions through a fan-out mapper, such as the one from Brandleap (brandleap.ai), and see how many sub-questions the site answers.

Learn more

What it is. The content also answers the sub-questions the engine generates from the main question (fan-out, chapter 2).

Why it matters. Citations often happen in the sub-questions. Evidence: C.

Chapter 6 of the book: Be citable

C8 Visible freshness Do the main pages show an update date and are they reviewed at least every quarter?
How to check

See whether the ten main pages show a date and whether that date is less than 90 days old.

Learn more

What it is. The main pages show the date of their last update and are reviewed at least once a quarter.

Why it matters. According to data from AirOps and Kevin Indig, pages that are not updated quarterly are three times more likely to lose citations (Erlin (erlin.ai)). Evidence: B.

Chapter 6 of the book: Be citable

C9 Real questions, taken from the CRM Does each service have real FAQs, drawn from sales, support or the CRM?
How to check

Compare the site's FAQ with the last 20 questions recorded by the sales team.

Learn more

What it is. The frequently asked questions for each service come from what clients actually ask: meeting notes, emails, support requests, CRM records.

Why it matters. The questions clients ask the sales team are the same ones they ask AI assistants. Evidence: C.

Chapter 6 of the book: Be citable

C10 Core topics in video too Do the core topics exist as YouTube videos, with a descriptive title and a transcript?
How to check

Does the company channel have videos on the three main topics, with question-style titles and a reviewed transcript?

Learn more

What it is. The main topics exist as YouTube videos, with a descriptive title, a full description and a transcript.

Why it matters. According to a Semrush report on 126 million prompts, AI Overviews cite YouTube 2.3 times more than the next source (Semrush, via Artificial Studio (artificialstudio.ai)). Evidence: B.

Chapter 6 of the book: Be citable

Step 4 of 5

O Off-site presence

Do third parties talk about us where engines look?

Weight: 20 points out of 100 · Chapter 7 of the book: Be talked about

O1 Press and industry publications Has the company been mentioned in the press or industry publications in the last 12 months?
How to check

Search for the company name in Google News, limited to the last 12 months.

Learn more

What it is. Mentions of the company in newspapers, magazines and specialized industry sites over the last 12 months.

Why it matters. News and industry sites are among the sources the engines choose most often (arXiv 2604.25707 (arxiv.org)). Evidence: B.

Chapter 7 of the book: Be talked about

O2 Reviews and directories Does it have complete profiles and reviews on review platforms and partner directories?
How to check

Look up the company on each platform relevant to the industry and count the reviews from the last 12 months.

Learn more

What it is. Complete, rated profiles on review platforms (G2, Capterra, Clutch, Google) and in technology partner directories.

Why it matters. According to Erlin's citation analysis, review platforms such as G2 and Capterra account for 17% of the sources where AI platforms find information about a brand (Erlin (erlin.ai)). Evidence: B.

Chapter 7 of the book: Be talked about

O3 Communities where customers ask questions Does it take part in communities where your customers ask questions (forums, Reddit, groups)?
How to check

Search for the company name and its main topics in those communities.

Learn more

What it is. Authentic participation in forums, groups and communities where your audience asks questions: Reddit, LinkedIn groups, software user communities, industry forums.

Why it matters. Some platforms cite communities very frequently: according to Frase.io data, Perplexity cites Reddit in 46.5% of its responses (Erlin (erlin.ai)). Evidence: B.

Chapter 7 of the book: Be talked about

O4 Third-party lists Does it appear in third-party comparison lists (for example, 'best X in [your country]')?
How to check

Search for five lists in the category in your market's language and in English and see how many include the company.

Learn more

What it is. The company appears in comparison lists made by others: "best X consultancies in [your country]", "top Y vendors".

Why it matters. According to GenOptima data, 74.2% of AI citations come from content in a "Top N" format (Erlin (erlin.ai)). Evidence: B, with an undisclosed sample.

Chapter 7 of the book: Be talked about

O5 Spokespeople on third-party channels Do the company's spokespeople appear on third-party podcasts, webinars or videos?
How to check

Search for the spokespeople's names on YouTube and on podcast platforms.

Learn more

What it is. The company's people appear on other people's podcasts, webinars, conferences and videos.

Why it matters. Each appearance generates mentions, transcripts and third-party pages that associate the name of the person and the company with the topic. Evidence: C.

Chapter 7 of the book: Be talked about

O6 LinkedIn on a regular basis Do the company and its spokespeople post regularly on LinkedIn about the core topics?
How to check

Count the posts from the last 90 days on the company page and on the spokespeople's profiles, and see whether the topics match the services.

Learn more

What it is. The company page and the spokespeople post every week on the core topics.

Why it matters. LinkedIn is the main public space for B2B conversations and it associates the names of people and companies with topics. Evidence: C.

Chapter 7 of the book: Be talked about

O7 Ecosystems and associations Is it listed in associations, ecosystems and directories in your industry?
How to check

List five relevant directories and associations and confirm presence and profile in each.

Learn more

What it is. The company is listed in the partner directories of the vendors it works with, in industry associations and in regional ecosystems.

Why it matters. These are authoritative sources that confirm the company's category and specialty. Evidence: C.

Chapter 7 of the book: Be talked about

Step 5 of 5

M Measurement and operations

Do we know what is happening and what it is worth?

Weight: 20 points out of 100 · Chapter 8 of the book: From citation to pipeline

M1 Question library and share of voice Do you have a list of 20 to 50 questions your customers ask AI assistants, reviewed every quarter?
How to check

Is there a document or tool holding the questions?

Learn more

Chapter 8 of the book: From citation to pipeline

M2 Question library and share of voice Do you measure every month whether you appear in the answers, compared with 3 competitors?
How to check

Is there a monthly report, manual or in a tool (HubSpot AEO, Semrush, Ahrefs Brand Radar)?

Learn more

Chapter 8 of the book: From citation to pipeline

M3 AI traffic in GA4 Does GA4 separate traffic coming from AI assistants, and does someone analyze it every month?
How to check

GA4, Reports, Traffic acquisition: does the AI assistant channel appear?

Learn more

Chapter 8 of the book: From citation to pipeline

M4 From traffic to CRM and revenue Does the CRM record when a contact or deal originated from an AI assistant?
How to check

Does the CRM show contacts and deals with original source "AI assistant"?

Learn more

Chapter 8 of the book: From citation to pipeline

M5 From traffic to CRM and revenue Do forms and sales meetings ask 'How did you hear about us?' with an 'AI assistant' option?
How to check

Check the website forms and the discovery script.

Learn more

Chapter 8 of the book: From citation to pipeline

M6 The monthly routine and automation Is there an owner, a monthly meeting and a budget for AI visibility?
How to check

Ask who is accountable and when the last meeting took place.

Learn more

Chapter 8 of the book: From citation to pipeline

M7 The monthly routine and automation Does every piece of content go through a checklist before publishing, and are there alerts to review old pages?
How to check

Check the editorial process and the review workflows.

Learn more

Chapter 8 of the book: From citation to pipeline

Frequently asked questions

What is AEO?

AEO (Answer Engine Optimization) means preparing the company to be the answer chosen by AI assistants. GEO (Generative Engine Optimization) is the same goal, under the term used in academic research.

How long does it take?

About 15 minutes for the 40 questions. Your progress is saved in this browser, so you can stop and come back later.

Is it free?

Yes. The score, the weakest dimension, the first five actions and the 90-day plan appear on screen, with no sign-up. The full report, with what to fix for each criterion, is sent by email to anyone who requests it.

How is the score calculated?

Each criterion is worth 0, 1 or 2. For each dimension, add up the points, divide by the maximum and multiply by the weight (15, 20, 25, 20 and 20 points). The sum of the five dimensions is rounded to the nearest whole number: Invisible (0 to 39), Present (40 to 59), Citable (60 to 79) and Recommended (80 to 100).

What happens to my answers?

They are stored only in this browser. Nothing is sent to Liminal unless you request the full report. Usage is measured in aggregate, without your answers, and only if you accept analytics cookies.

· AEO Readiness Scorecard v1.0