AEO NIUEXA

Content Quality & Authority

25% Weight (Highest) Check reference Priorities vary by result

Content checks look at what your page says and how: whether it answers questions clearly, shows who stands behind it, cites its sources and is easy to quote. This dimension has the highest weight (25%) in the AEO Score. Checks marked Unproven are shown in your report but not scored.

Show who is behind the page: Experience, Expertise, Authoritativeness and Trustworthiness are the quality criteria in Google's search quality guidelines. Showing who wrote a page, their credentials and your contact details helps readers and search engines judge it.
Seven checks are AI reviews: Content Quality, E-E-A-T Signals, AI Readiness, Semantic Richness, Direct Answer Provision, Topical Coverage and Sentiment & Tone are scored by an OpenAI model that reads part of the page: on long pages, the part that matters for each check. The model is told the page's language and judges it in that language, so an Italian page is not marked down for not being in English. These reviews are our model's opinion, not a search engine's. When the model cannot run, the check is "not measured" and left out of the score, and if none of the seven could run, the scan of that page is not charged. The other checks are rules computed from the page itself, with Italian wording added on Italian pages.
How these checks count: A check counts by its importance in this dimension multiplied by how well the evidence supports it: Official and Corroborated checks count in full, Practitioner advice counts at half weight, and checks marked Unproven are shown in your report but not scored. A check that is N/A or not measured is left out of the score rather than counted as a pass or a failure; if nothing in this dimension can be weighed, the dimension is shown as not measured and the other weights are scaled up. The badge next to each check shows the usual priority of a failure; your report shows the priority of your own result. See the scoring system for the details.

Quality & Authority Checks

Signals that show the content is useful and who stands behind it

Content Quality

High
What it checks
An OpenAI model reads the beginning and end of the page (on longer pages, the first 1,800 and the last 1,200 characters) and rates its overall quality (poor, fair, good or excellent) and its clarity from 1 to 10, noting readability, tone, grammar and depth.
Why it matters
Search engines say they reward helpful, reliable content written for people, beyond facts readers could find anywhere. This review is our model's opinion, not a search engine's score.
How it is scored
Official evidence, counts in full. The overall rating sets the base (poor 25, fair 50, good 75, excellent 100) and each clarity point above or below 5 adds or removes 5, up to 100. It passes at 60 (High below 60).
Pass: 60 or more Fail: under 60 Not measured: the AI review could not run
How to fix
Answer the reader's question fully and accurately. Add what only you can say (your experience, data or examples), back claims with sources and cut filler.

E-E-A-T Signals

High
What it checks
An OpenAI model reads the beginning and end of the page (on long pages, the first 2,000 and the last 3,000 characters, where bylines and author boxes usually are) and looks for trust signals: a named author and their credentials, cited sources, original research or data, expert language, first-hand experience and trust indicators such as contact details or policies. It rates the page from 1 to 10 and lists what is missing.
Why it matters
Google's quality guidelines look for experience, expertise, authoritativeness and trust. They are a quality lens, not a ranking formula, and AI tools are not proven to apply them the same way.
How it is scored
Corroborated evidence, counts in full. The model's score from 1 to 10, times 10. It passes at 60 (High below 60).
Pass: 60 or more Fail: under 60 Not measured: the AI review could not run
How to fix
Name the author with a short bio and real credentials. Show first-hand experience, cite sources for claims, and make About and Contact information easy to find.
E-E-A-T Components
  • Experience: First-hand experience with the topic (case studies, personal use, real examples)
  • Expertise: Credentials, qualifications, depth of knowledge demonstrated
  • Authoritativeness: Industry recognition, citations from others, brand reputation
  • Trustworthiness: Transparency, accuracy, security, contact information

Author Credentials

High
What it checks
Looks for a named author in the page's HTML: an author meta tag, or a byline or author box (an element whose class or id says author or byline, rel="author" or itemprop="author"; on Italian pages also autore, autrice or firma). Inside it, it looks for a bio (more than 50 characters of text), credentials (such as PhD, Dr., CEO or "10 years of experience"; on Italian pages also titles such as Dott., Avv. or Ing.) and links to LinkedIn, X/Twitter or GitHub profiles.
Why it matters
A named author with relevant experience tells readers who stands behind the content, a trust signal Google's quality guidelines describe.
How it is scored
Corroborated evidence, counts in full. 100 with a bio, credentials and profile links; 70 with a bio or credentials; 50 for a name only (all three pass). 25 (fail) when author elements are found but no name, 0 (fail, High) when there is no author information.
Pass: a named author (100 with bio, credentials and profile links) Fail: no author, or an author box without a name
How to fix
Show the author's name on the page with a short bio, relevant qualifications and links to professional profiles, in an element marked as the author (for example class="author" or rel="author"). Use the same name as the author in your Article markup.

Citation Quality

Low
What it checks
Looks at the page's links to other websites: how many different sites they point to and how many are on our list of recognised sources (government, universities and international bodies first; then encyclopedias, major news outlets, scientific publishers and official technical documentation; Italian pages also count Italian public bodies, universities and national news). With no links to other sites, it looks for citations written in the text instead, such as "according to …", "(Author, 2024)" or a quote followed by "said".
Why it matters
Links to credible sources let readers check your claims and show where your facts come from.
How it is scored
Corroborated evidence, counts in full. The score combines the average authority of the linked sites, how varied they are and how many are recognised sources, with 5 more each for an academic and a government source and 5 less for each problem (many links to few sites, more unknown than recognised sources, a low average authority). It passes when the linked sites average 40 or more on our authority scale (unknown sites count 30) and there are at least two of them. A page with no links to other sites is not failed: it scores 60 when it cites sources in the text and 50 otherwise.
Pass: links to two or more sites averaging 40 or more, or no outside links at all Fail: links to a single site, or mostly to sites not on our list
How to fix
Link each statistic or claim to its primary source (a study, official statistics, a standard or the original report). Use descriptive link text and replace dead or outdated references.

AI Readiness Checks

Whether the page states its answers and facts in a way that can be quoted correctly

AI Readiness

High
What it checks
An OpenAI model reads paragraphs taken evenly from the whole page and judges how easily its facts can be lifted out and quoted: how clearly they are stated, how well the text splits into self-contained chunks and how likely it is to be cited.
Why it matters
AI answers quote short passages. A fact stated plainly in one sentence that makes sense on its own is easier to quote correctly.
How it is scored
Corroborated evidence, counts in full. The average of the model's citation potential (1 to 10, times 10) and its rating of the chunks (poor 25, fair 50, good 75, excellent 100). It passes at 60 (High below 60).
Pass: 60 or more Fail: under 60 Not measured: the AI review could not run
How to fix
Start sections with the direct answer, then explain. State key facts in plain sentences that name what they are about, and keep numbers, names and dates explicit.

Semantic Richness

Unproven · not scored
What it checks
An OpenAI model reads samples from the beginning, middle and end of the page and rates vocabulary variety, related concepts and topic coherence.
Why it matters
Write naturally; varied wording is not a proven citation factor. Shown for information; it does not count in your score.
How it is scored
Unproven, not scored. The model's score from 1 to 10, times 10; it passes at 60.
How to fix
Nothing is required. Cover the subtopics your readers ask about, in the words they use.

Direct Answer Provision

High
What it checks
An OpenAI model reads the question-and-answer passages of the page (questions in the text; if there are none, FAQ markup, details and summary blocks or headings with the paragraphs under them; failing that, the start of the page) and judges whether the page answers questions directly and clearly, and how well those answers would work as a quoted snippet.
Why it matters
Answers that come early and stand on their own are easier for readers to find and for search and AI tools to quote.
How it is scored
Corroborated evidence, counts in full. The model's snippet potential from 1 to 10, times 10. Without direct answers it drops by 20, to no less than 30, and the check fails (High). It passes when the page has direct answers and a snippet potential of 60 or more.
Pass: direct answers with a snippet potential of 60 or more Fail: no direct answers, or answers that are hard to quote Not measured: the AI review could not run
How to fix
Put the direct answer in the first sentence or two of the page or section, for example "X is …", then add the supporting detail.
Example Format
<h2>What is AI Engine Optimization (AEO)?</h2>
<p><strong>AI Engine Optimization (AEO) is the practice of optimizing
web content to be better understood, indexed, and cited by AI systems
like ChatGPT, Claude, and Perplexity.</strong> Unlike traditional SEO
that focuses on search rankings, AEO focuses on how AI interprets and
presents your content in generated answers.</p>

Answer Readiness

Unproven · not scored
What it checks
One AI reading of the page, part of the scan at no extra cost. It names the main question the page sets out to answer, says whether the first 40-80 words answer it (yes, partly or no, with the opening quoted), lists sections that would not make sense if quoted on their own, and lists up to 8 claims with no visible source. Every quote is checked against the page text; anything not found on the page is left out.
Why it matters
Answer engines often quote one passage without the rest of the page. An opening that answers the question, sections that name their subject and keep their conditions, and claims with a source are easier to quote correctly.
How it is scored
It is not scored. It is advice shown in its own "Answer readiness" panel and is not ranked in "What to do next" until we have measured how often the AI judgement is right. When the AI could not run, the panel says "Not measured"; a page with almost no text (under 200 characters) gets no advice. Direct Answer Provision above still scores the answers to the questions on the page.
How to fix
Start the page with a one to three sentence answer to its main question. Open each section listed by naming its subject instead of "it" or "this". Link a source next to each claim listed, or remove the claim.

Topical Coverage

Medium
What it checks
An OpenAI model reads an outline of the page (its headings with the opening and closing text) and judges whether it covers the main topic and the follow-up questions a reader is likely to have, listing the gaps it sees.
Why it matters
Readers usually have follow-up questions. A page that answers them is more useful and covers more of what people search for.
How it is scored
Corroborated evidence, counts in full. The model's completeness score from 1 to 10, times 10. It passes at 60 (Medium below 60).
Pass: 60 or more Fail: under 60 Not measured: the AI review could not run
How to fix
List the questions your customers ask about the topic and answer the important ones on the page, with examples where they help.

Structure & Format Checks

How the text is organised, how hard it is to read, how long it is and whether it repeats itself

Content Structure

Medium
What it checks
Computed from the page text, not by the AI: the average paragraph length, whether there are lists and whether there are section headings. On pages longer than about 2,000 characters it reads an outline (the headings and the first sentence under each), so the paragraph-length points mostly apply to shorter pages.
Why it matters
Short paragraphs, lists and headings make text easier to scan and to quote. This is common practice.
How it is scored
Practitioner advice, counts at half weight. 50 to start, 25 more when the average paragraph is 50 to 200 characters long, 15 for lists and 10 for headings. It passes at 70 (Medium below).
Pass: 70 or more Fail: under 70
How to fix
Break long paragraphs up, add descriptive subheadings and use lists for steps and options.

Reading Level

Unproven · not scored
What it checks
Estimates how hard the text is to read, leaving out menus, headers, footers and sidebars. English pages use the Flesch-Kincaid grade, sentence length and long words; Italian pages use the Gulpease index, the readability formula built for Italian; other languages use sentence length and word complexity.
Why it matters
Write for your audience: no reading grade is proven to win AI citations. Shown for information; it does not count in your score.
How it is scored
Unproven, not scored. English: 100 from grade 6 to 10, with points off for harder text, very simple text, long sentences and many long words; it passes up to grade 12. Italian: 100 for a Gulpease index of 60 or more, 85 from 50, 70 from 40, 40 below that; it passes at 40. Other languages: points off for average sentences over 25 words and many long words; it passes at 60.
Pass: grade 12 or lower (English), a Gulpease index of 40 or more (Italian) N/A: fewer than 100 words
How to fix
Prefer short sentences and common words where your audience expects them; keep the technical terms your readers use.

Content Depth

Unproven · not scored
What it checks
Counts the words in the main content (menus, headers, footers and sidebars removed) and looks at sections, paragraph length, vocabulary variety and the kinds of content used: headings, paragraphs, lists, images, tables, code and quotes.
Why it matters
Length should follow the task: no word count is proven to help with search or AI. Thin pages lack depth and very long pages may lack focus. Shown for information; it does not count in your score.
How it is scored
Unproven, not scored. 30 to start, plus 10, 20, 30 or 40 at 300, 800, 1,500 and 3,000 words, up to 10 for H2 and H3 sections, 2 for each kind of content used and up to 10 for varied vocabulary, minus 5 for each problem (under 300 words, a long text with no section headings, paragraphs averaging over 150 words, repetitive vocabulary). It passes at 300 words.
How to fix
Cover what the reader needs to finish the task, then stop. Add depth (examples, data, steps) where the topic needs it, not to reach a number, and split very long content into focused pages.

Duplicate Content

Medium
What it checks
Compares the text blocks of this page (paragraphs, list items, quotes, table cells and captions of 50 characters or more, outside menus, headers, footers, sidebars and review sections) and finds blocks repeated word for word, five-word phrases repeated three or more times, and stock phrases such as "all rights reserved" or "privacy policy" (with Italian equivalents on Italian pages). It looks within this page only: it does not compare your pages with each other or with other sites.
Why it matters
Repeated blocks add length without adding information and make the page harder to read and quote.
How it is scored
Official evidence, counts in full. Starts at 100 and loses one point for each percent of repeated blocks, 5 for each repeated phrase listed (up to five), 10 when three or more stock phrases appear, and 10 more for each problem (more than 30% of the blocks repeated, more than three repeated phrases). It fails (Medium) when more than 30% of the blocks are repeats.
Pass: 30% of the blocks or fewer are repeats Fail: more than 30% of the blocks are repeats N/A: fewer than three text blocks
How to fix
Remove repeated paragraphs and boilerplate so each thing is said once. If whole pages duplicate each other, merge them or point a canonical at the preferred one.

Trust Signal Checks

Dates, ways to reach you and tone

Content Freshness

Low
What it checks
Looks for the dates the page declares: datePublished and dateModified anywhere in its JSON-LD, and date meta tags (article:published_time, article:modified_time, og:updated_time, date, pubdate and similar). Italian dates such as "15 marzo 2024" are read on Italian pages. It then judges how recent the update date is (or the publication date, when there is no update date). A date shown only in the text, in a time element or in a copyright line is not enough on its own.
Why it matters
A visible, accurate date tells readers and search engines how current the information is. Update the facts when they change, not just the date.
How it is scored
Official evidence, counts in full. 50 to start, 15 for a publication date, 15 for an update date and 10 when they come from JSON-LD. Then 20 more when that date is up to 90 days old, 10 when it is up to a year old, 5 less when it is up to two years old and 20 less when it is older, and 10 less for each problem (no date, older than two years, an update date before the publication date). It passes when a publication or update date is found. Its priority is Medium for content over two years old and Low otherwise.
Pass: a publication or update date in the markup or meta tags Fail: no date found
How to fix
Show the publication and last-updated dates on the page, and add datePublished and dateModified to your Article or WebPage markup. Change dateModified only when the content really changes.

Contact Information

High
What it checks
Looks for ways to reach you on the page: an email address (in JSON-LD, a mailto link or the text), a phone number (JSON-LD, a tel link or the text), a postal address (JSON-LD, itemprop="address" or a street address in the text; Italian street addresses on Italian pages), a contact form and a link to a contact page.
Why it matters
Visible ways to reach you show there is a real business behind the site, which matters to readers and to Google's quality guidelines.
How it is scored
Corroborated evidence, counts in full. Counting email, phone, address and contact form: 100 with three or more, 70 with two, 50 with an email or a form alone, 40 with a phone alone (all pass), and 10 more, up to 100, when they come from JSON-LD. An address alone (30, or 40 from JSON-LD) or only a link to a contact page (30) fails; nothing at all scores 0 (High).
Pass: an email address, a phone number or a contact form on the page Fail: only an address or a contact-page link, or nothing
How to fix
Show at least an email address or a contact form, ideally with a phone number and a postal address, on the page or in the footer, and add them to your Organization or LocalBusiness markup (ContactPoint and PostalAddress).

Sentiment & Tone

Unproven · not scored
What it checks
An OpenAI model reads the first 1,000 characters and labels the sentiment (positive, neutral or negative) and the tone (professional, casual, academic or promotional).
Why it matters
A plain, professional voice reads well, but AI tools are not proven to prefer positive wording or to skip promotional pages. Shown for information; it does not count in your score.
How it is scored
Unproven, not scored. 100 for a professional tone with positive or neutral sentiment, 75 for other professional or academic text, 50 for casual or promotional text; it passes at 75.
How to fix
Match the tone to your audience and topic: factual for information, confident but not hyped for offers.