Free check
Can ChatGPT cite your website? 50 checks to find out, in 30 seconds
The check reads your site the way Google and AI assistants read it: fifty checks in 30 seconds, and tells you what is missing in order of importance. Enter your website address below.
Free · 30 seconds · no sign-up · no card
- How Google finds youSEO score · Good78
- How AI reads youGEO score · Critical40
Overall: 59 Needs work
What it measures
50 checks, answering six questions
Can Google find you?
Title, description, indexing, canonical address, HTTPS and structured data: what decides whether and how you show up in results.
title element, meta description, robots directives (noindex), rel=canonical, HTTPS and schema.org structured data in JSON-LD: the signals that decide indexing and the preview in the results.
Fix readyDo AI assistants understand you?
Who you are, where you work, who signs the pages, whether you answer customers' questions and whether the bots of ChatGPT and the others can get in.
Entity declared in JSON-LD (Organization, Person, LocalBusiness), author and E-E-A-T, FAQPage, machine-readable contact details and robots.txt rules for AI crawlers (GPTBot, ClaudeBot, PerplexityBot).
Fix readyIs the text useful?
How much text there is, how the headings are organised, whether images have a description and whether the page declares its language.
Word count (threshold at 300), h1 and h2 hierarchy, alt attribute on images and lang attribute on html.
We tell you what to doDoes it read well on a phone?
Fit on a small screen, response speed, colour contrast, social preview and site icon.
Meta viewport, server response time (threshold at 2.5 seconds), WCAG contrast between the text and background colors declared in the CSS, Open Graph tags and favicon.
We tell you what to doAre you ready to be cited?
Citation readiness: a band from A+ to F and the AI bot matrix, that is, which bots can read your site and which cannot.
Citation readiness score: structural signals of the page (question headings, short sections, direct answer, sources, page JSON-LD) summed up in a band from A+ to F, plus the AI bot matrix read from robots.txt.
Band and matrix in the reportAre you easy to find nearby?
For trades that serve a local area, extra local SEO checks: city, address, phone and opening hours.
For local businesses: consistency of name, address and phone number, LocalBusiness and subtypes in JSON-LD, local references in the text and opening hours.
Only where it applies
See the list of all the checks →
Where to check the rules behind the checks: structured data according to Google · the schema.org guide · the llms.txt file proposal · OpenAI's crawlers
The full list
50 checks, one by one
This list is not written by hand: the analysis engine generates it, so if a check is added or removed it changes here together with the number. For each one we say whether it counts toward the SEO score, the GEO score or the citability signals.
SEO checks 18
They count toward the SEO score: they say how ready the page is for Google.
- Page titleSEO
That the page has a title of the right length to show in full in Google's results.
Presence of a non-empty title between 10 and 60 characters: above 60 Google truncates the title in the results, below 10 it is considered too short.
- Google summarySEO
That there is a description sentence under the title, neither too short nor too long.
Presence of meta name="description" between 70 and 160 characters, within which the sentence is not truncated in the snippet; its absence is the most serious flag.
- Main visible headingSEO
That the page has a visible main heading, without repeats and without too many of them.
A single non-empty h1 (the text may come from an image's alt or from aria-label). Zero h1, duplicate h1, more than five, or several distinct h1 on a well-structured page are flagged.
- Heading structureSEO
That the text is divided into sections with subheadings, so you can see how it is organised.
At least two h2 on pages of 400 words or more (nav, header and footer excluded); below 400 it is not judged. Subheadings marked as h1 instead of h2 fail with low severity.
- The site can be indexedSEO
That the page does not tell Google not to show it in the results.
No noindex in either the meta robots or the X-Robots-Tag HTTP header; its presence in one of the two makes the check fail with high severity.
- How much text there isSEO
That the page has enough text to say something useful, at least 300 words, except pages that are short by nature such as contacts or privacy.
At least 300 body words, excluding nav, header and footer; URLs whose path points to contact, privacy, cookie, terms, login, checkout or thank-you pages are exempt.
- Image descriptionsSEO
That almost all images have a written description, useful for people who cannot see them and for search engines.
Share of img without an alt attribute equal to or below 20%. alt="" counts as a decorative image and is not penalized; img with src, data-src or srcset are counted.
- Canonical addressSEO
That the page declares which address is its main one, so it is not mistaken for a duplicate.
Presence of link rel="canonical" with a filled-in href; the check only verifies that the tag exists, not that the URL it points to matches the page's own.
- Readable on mobileSEO
That the page adapts to a smartphone screen.
Presence of meta name="viewport" with a non-empty content; the check only reads the tag and does not measure the actual responsive layout.
- Internal linksSEO
That the page links to the other pages of the site, with at least three internal links.
At least 3 unique internal links (same origin, deduplicated by URL), with menu and footer included; # anchors, mailto:, tel: and javascript: are excluded.
- Secure connectionSEO
That the site opens over a secure connection, that is, HTTPS.
The page's final URL, after redirects, uses the https:// scheme. The check does not assess the validity or expiry of the certificate.
- Declared languageSEO
That the page declares which language it is written in.
Presence of a non-empty lang attribute on the html element; the language code is not compared with the actual language of the text.
- Social previewSEO
That the WhatsApp and social preview has a title, a description and an image that really exists.
Presence of og:title, og:description and og:image; if og:image is declared, a HEAD request must not return 404 or 410. Other network errors are not counted.
- Structured data for GoogleSEO
That the page carries schema.org structured data written without errors, the kind Google and AI assistants read.
At least one JSON-LD block, all valid: correct syntax and required properties present for Organization, LocalBusiness, Article, FAQPage, Product and BreadcrumbList. A single invalid node is enough to fail.
- Site iconSEO
That the site has its small icon, the one shown in the browser tab.
Presence in the head of a link with rel icon, shortcut icon or apple-touch-icon and a filled-in href; the file is not downloaded and its format is not verified.
- Response speedSEO
That the server responds in under two and a half seconds.
Response time to the HTML request equal to or below 2500 ms; it does not measure LCP, CLS or page weight.
- Colour contrastSEO
That the text can be read against the background, with enough contrast between the colours written in the CSS.
WCAG contrast ratio between text color and background, read from inline styles and from the body rule in style. The worst pair is used: below 4.5:1 is a warning, below 3:1 is serious.
Not always possible: the page does not declare text and background colours
- Colours written by nameSEO
That the CSS colours are written in hexadecimal and not with a name such as “navy”.
Use of CSS color names (black, white, navy and the like) instead of hexadecimal values in inline styles or in the body rule; only the names on an internal list are recognized.
Not always possible: the page does not declare colours in its CSS
GEO checks 15
They count toward the GEO score: they say how readable the page is for AI assistants.
- Who you are, statedGEO
That the site says who it is in its structured data: an organisation or a person, with its name.
Presence in JSON-LD of a node with @type Organization, Person or a LocalBusiness subtype (Attorney, Physician, Restaurant and the like), read at the top level or inside @graph.
- Business type declared in the dataGEO
That the structured data declares your type of business (practice, shop, professional), even with little data: the type counts, not how complete the profile is.
Presence in JSON-LD of a business @type: ProfessionalService, LocalBusiness, Service, Person or a subtype such as Attorney, Physician, Dentist, Restaurant. The type counts, not how complete the fields are.
- Questions and answersGEO
That the frequently asked questions and answers are on the page and also declared in the structured data.
Presence in JSON-LD of a FAQPage, HowTo or QAPage node; if it is missing, the message tells apart FAQs already visible on the page (details, faq classes, headings with "?") from absent ones.
- Readable contact detailsGEO
That phone, email and address are written so that a machine can read and verify them.
Presence of a phone number (tel: link or a number next to words like phone) or of an address starting with a street word (via, piazza, street, calle, rue); the footer is included.
- About us pageGEO
That the site has a page or a section that says who is behind it.
Presence in the text or in CSS classes of signals of an about page (the wordings in Italian, English, Spanish and French; classes about, bio) or of a link to an about page.
- Publication or modification date presentGEO
That somewhere, on the page or in the data, there is a publication or modification date, even an old one: what counts is that there is one.
Presence of a date: time datetime, itemprop datePublished or dateModified, the same properties in JSON-LD, or text such as "aggiornato il" or a date with the month in Italian. Outside articles, a missing date is not judged.
- Recognisable authorGEO
That there is at least one clue about who writes: a byline, a professional title or an author in the data. One clue is enough.
Presence of rel="author", author, class .author or .byline, of a professional title (Dott., Avv., Dr., Prof.) with a surname in the first 500 words, or of a JSON-LD Person node or one with author.
- Navigation pathGEO
That the navigation path, the “breadcrumbs”, is declared in the structured data without errors.
Presence of a valid JSON-LD BreadcrumbList: schema.org @context, non-empty itemListElement, entries with a sequential numeric position from 1 and with item or name. On a home page without breadcrumbs it is not judged.
- Where you are, for AIGEO
That the page gives a phone number, an address or a VAT number, the details that say where and who you are.
Presence of a tel: link, of telephone or address in JSON-LD, or in the text of a phone number, an address (via, rue, street + number) or a VAT number. Not judged on software; not run on e-commerce.
- Page structureGEO
That the page separates the real content from the surroundings using the elements made for it.
Presence of main or article, or of at least two section, on pages of at least 150 words; below 150 words the check passes. Heuristic signal.
- Conversational languageGEO
That the text uses the words people use when they ask an assistant, not only technical terms.
Density of interrogative words (who, how, when and the like, depending on lang) of at least 3 per 500 body words; heuristic signal, not judged for unsupported languages.
- Words from your tradeGEO
That the page contains the phrases a client would use to look for your trade.
Presence in the text of at least two typical question phrases for the sector (how much does it cost, how it works), of one phrase plus a question heading or an FAQ, or of three question headings; heuristic signal.
- No AI bot blocked, training includedGEO
That the robots.txt file does not bar any of the AI bots we know, including those that train the models: a single ban is enough to trigger the warning.
In robots.txt none of the nine AI user agents (GPTBot, ChatGPT-User, ClaudeBot, Claude-Web, PerplexityBot, Google-Extended, CCBot, anthropic-ai, cohere-ai) has Disallow: /, either by name or through User-agent: *.
Not always possible: the robots.txt file could not be read
- llms.txt fileGEO
That the llms.txt file exists, which briefly explains to AI assistants what the site does.
Positive response to the request for /llms.txt at the domain root, with content that is not an HTML page; the content of the file is not evaluated.
Not always possible: the llms.txt file could not be read
- Paragraphs that open with a statementGEO
That the paragraphs of the text start with a statement and not with a preamble: it looks at how every paragraph opens, not only at the opening of the page.
Over the first twelve paragraphs of 8 words or more (at least three are needed), at least 50% open without an Italian subordinating conjunction, without a question and with a first sentence of 30 words or fewer. Heuristic.
Not always possible: the page has no paragraphs long enough to assess
Citability signals 17
They are GEO signals: they weigh 30% of the GEO score and feed the A+ to F band. If the page has too little text or is in a language we cannot read, they are not judged and do not enter the score.
Text that is easy to extract
- Headings phrased as questionsGEO · citability
That some headings are written like a client's question, with the answer right below.
h2-h4, summary or dt headings that end with "?" or start with "¿": at least 2 are needed under 400 body words, at least 3 above. Not judged under 150 words.
- Short sectionsGEO · citability
That the text is in short sections, each short enough to be cited on its own.
Sections delimited by h1-h3: passes if less than 40% of the words sit in sections over 300 words and the median is 220 or less. Not judged under 200 words. Heuristic signal.
- Lists and tablesGEO · citability
That where needed there are bullet lists or tables, which an assistant extracts whole.
Presence in the body of ul or ol with at least three li (menus made only of links excluded) or of a table with at least three tr: 1 is needed under 600 words, 2 above. Not judged under 250 words.
- Direct answer at the top of the pageGEO · citability
That the first sentence under the page heading says right away what it is about: it looks only at that one, the opening of the whole page.
First sentence of the first paragraph after the h1: 8 to 30 words, neither a question nor a greeting, with a word of at least 4 letters in common with the h1 or title. Under 5 or over 45 words it counts as absent.
- Sentence lengthGEO · citability
That the sentences are short enough to be cited whole.
Average sentence length in paragraphs of 24 words or fewer, with no more than 15% of sentences over 35 words; up to 30 on average is only improvable. Not judged under 5 sentences or 80 words.
Evidence and sources
- Figures and concrete dataGEO · citability
That the text contains figures and concrete data: years in business, prices, percentages.
Distinct figures in the body: percentages, amounts in € or $ and numbers with units (years, clients, projects, hours, sq m). At least 2 are needed under 500 words, at least 4 above. Not judged under 200 words.
- Sources citedGEO · citability
That the text links to the sources of what it states: a body, a law, a study.
Links to different external sites inside paragraphs, lists or quotes, excluding social networks, maps, marketplaces and rel sponsored or ugc: 2 or more passes, 1 is improvable. Not judged under 300 words.
- Named testimonialsGEO · citability
That there are testimonials or quotes with the name of the person speaking.
Presence of a blockquote or q with cite, signature or figcaption, of a JSON-LD Review or AggregateRating node, or of an element with class testimonial. Not judged under 200 words.
- Trust pages linkedGEO · citability
That the pages that build trust can be reached from the page: about, contact, privacy and terms.
Links, recognized from the text or the path, to at least two of the three trust pages: about, contact, privacy or terms. Only one is improvable.
Data for machines
- Page described in structured dataGEO · citability
That the page is described in the structured data and linked to the organisation.
A page-type JSON-LD node (WebPage, Article, Service, Product, FAQPage and the like) with name or headline and with description, image or url, linked to the organization through isPartOf, publisher, author or an @id.
- Complete organisation profileGEO · citability
That the organisation profile has all its details: name, website, logo, contact and official profiles. Declaring the type of business is not enough.
A JSON-LD Organization node or subtype with sameAs of at least two https profiles and at least three other data points among name, url, logo or image, contact (telephone, email, address, contactPoint).
- Consistent name everywhereGEO · citability
That the business name is the same in the title, the structured data and the footer.
The JSON-LD name (or og:site_name) matches, ignoring accents and company forms, title, og:site_name and the © line of the footer; at least two possible comparisons are needed, otherwise it is not judged.
Crawler access
- AI search and citation crawlers allowedGEO · citability
That the crawlers assistants use to search and cite, such as OAI-SearchBot and PerplexityBot, can read the site, even if you ban the ones that train models. It reads the whole robots.txt and tells blocked, partial and undeclared apart.
None of the five search and citation crawlers (OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot) has Disallow: / in robots.txt, by name or through *; three blocked zero the signal.
- Text readable without JavaScriptGEO · citability
That the main text is already in the page a crawler receives, without needing to run JavaScript.
Body words already present in the HTML downloaded without running scripts: passes with at least 120 words, or 60 if there is no empty root (#root, #app, #__next) or "enable JavaScript" notice.
- Snippets allowedGEO · citability
That the page does not forbid quoting excerpts from it, with nosnippet or max-snippet.
No nosnippet or max-snippet:0 in meta robots, meta googlebot and the X-Robots-Tag header; a max-snippet between 1 and 49 characters is only improvable.
Freshness
- Recent update visible on the pageGEO · citability
That the update date is visible on the page and recent, within a year: a date two years old does not pass.
Date read from time datetime, itemprop or text near words like "updated" or "published": the most recent must be at most 12 months old, between 13 and 24 is improvable. Without dates, outside articles, it is not judged.
- Modification date in the structured dataGEO · citability
That the modification date is written in the data for machines (dateModified) and matches the one shown on the page.
dateModified in JSON-LD (or article:modified_time, og:updated_time), within at most 31 days of the date visible on the page; datePublished alone is improvable.
How to read the score
Three numbers and four bands
SEO score
How ready the page is for Google: title, description, headings, images, speed and structured data.
GEO score
How readable and quotable you are for ChatGPT, Claude, Perplexity and Gemini.
Overall
The summary of the two, weighting GEO more when it is low: a weak GEO does not hide behind a good SEO.
- Excellent85 to 100The site is well prepared: only touch-ups remain.
- Good70 to 84A solid base, a few signals are missing.
- Needs work50 to 69There are problems that cost visibility.
- Critical0 to 49Google and AI assistants struggle to understand the site.
Citation readiness
Being there is not enough: you must be citable
Besides the two scores, the check gives a band from A+ to F to citation readiness: how well the page is written and declared so that an AI assistant can use it as a source.
Next to the band there is the AI bot matrix: for each bot you see whether the site lets it read the pages or blocks it. It is the quickest way to find a block that nobody put there on purpose.
What happens next
Other tools tell you what's wrong. Trovavio fixes it.
- 1 · Measures
In 30 seconds you know where you lose customers
We read the site the way Google and ChatGPT do, and rank the problems by importance.
- 2 · Writes
The fix arrives already written
Using words already on your site: made-up sentences are thrown out before they reach you.
- 3 · Publishes
You approve and the fix goes on your site
With the WordPress plugin or, for Wix, Shopify and other sites, with the connector: from the Solo plan. Or you copy it and send it to whoever manages your site.
Who it is for
For anyone who manages a site, their own or their clients'
If you manage your clients' sites, the check is the quickest way to start a conversation: a score, the problems in order and the fix ready to show.
What it costs
Try it for free. Pay when you see it works.
Solo
€49 a month, VAT included
- 1 site, 10 pages per scan
- 100 AI credits per month
- Weekly report by email
- Apply on WordPress in one click
With annual billing you pay ten months out of twelve. Cancel whenever you like.
Compare the plans →First, for free
Checking your home page needs neither sign-up nor card. You see the score, the problems in order and the first fix already written.
Questions
What people ask us before trying
How do I know whether ChatGPT cites my website?
Type your address above. The GEO score measures how readable and citable your site is for AI assistants, and the visibility test asks one assistant that searches the web which sources it cites. We query one assistant, not every engine one by one: the methodology page has the details.
Is the check really free?
Yes. Checking your home page needs no sign-up and no card: you see the scores, the problems in order of importance and the first fix already written. Paid features are for applying the fixes to your site and checking more pages.
How long does it take and what do you look at?
About 30 seconds. We download the page the way a search engine would, without running the site's JavaScript, and we also read the domain's robots.txt and llms.txt.
What does the A+ to F band show?
It shows how ready the page is to be cited by an AI assistant: A+ is the highest band, F the lowest. It does not replace the SEO and GEO scores: it sits next to them, because it looks at one specific thing, citability.
Does it work on any website?
On any public website, built with WordPress, Wix, Shopify or by hand. On sites built entirely in JavaScript the results can be less precise, and in that case we tell you.
Do you guarantee that ChatGPT will cite me?
No, and be wary of anyone who promises it. Nobody controls what an assistant answers: we can tell you what stops your site from being read and cited, and give you the fixes already written.
Is it useful for an agency too?
Yes: it is the quickest way to show a client where they lose visibility. With the Agency plan you manage up to 15 sites and hand over the report with your own logo.
Find out whether your site can be cited
Free · 30 seconds · no sign-up · no card
Paid plans start at 49 euros a month and can be tried for 14 days, no card.