Free tool · Email required

Free AI visibility grader

Enter your site. One AI engine answers three questions generated from your homepage, and you get a score out of 100 with the answers and the checks behind it. A snapshot, not a tracked trend.

Required, because each run calls an AI model. Your address joins CiteCue’s waitlist and we may email you about access and product updates. We don’t sell it. Your grade shows on this page, not by email. Privacy Policy

One grade a day. A run takes about a minute.

What the grade coversReady
100points at mostA letter grade too, A to F. One grade a day.
  • Not checked yet: Site: not entered yet
  • Not checked yet: Email: needed to run
  • Note: 3 questions, generated from your homepage, put to one AI engine
  • Note: Scored on how often the answers name you and cite your pages
  • Note: With checks on AI readiness and on what an AI agent could do on your site
How it works

A first reading in three steps

  1. Step 01

    Enter your site and your email

    A domain is enough. The grader reads your homepage to work out what you sell and who else sells it; it has nothing else to go on.

  2. Step 02

    One AI engine answers three questions

    Two about your category and one about your brand, generated from your homepage. Meanwhile it checks how ready your site is for AI crawlers and agents.

  3. Step 03

    Read the grade and the evidence

    A score out of 100, each question with the start of its answer, the competitors named, and every check with what to fix. Publish it if you want a link to share.

Everything you need to know about an AI visibility grade

Four short chapters on what the grade measures, where its questions come from, why one reading is a snapshot, and what the checks behind the number look at.

Chapter 1: What the grade measures

Two things an answer can do for you, and one number over them.

An AI answer can do two different things for a business. It can name you, which puts you on the reader’s shortlist, and it can cite one of your pages as a source, which makes your page the evidence. The grade reports both: the share of the answers that name your brand, and the share of the sources cited in those answers that are your own pages. The lesson on why mentions and citations are two different wins covers what each one takes.

The score out of 100 and the letter are a summary over those answers, and a summary hides its parts. Before you act on it, open it up: which questions named you, which didn’t, and who was named instead. The lesson on reading a visibility score walks through that, and the tutorial on how CiteCue’s visibility score is built shows the same idea over a full set of tracked questions.

Chapter 2: Where the questions come from

Your homepage. Not a record of anyone’s searches.

The grader starts by reading your homepage and working out three things from it: what you are called, what you sell, and who else sells it. From that it fills in 3 questions: two that ask about your category without naming you, and one that asks about your brand by name. So the questions are generated. Nobody typed them, and neither the grader nor CiteCue has any record of what people ask an assistant.

That matters for how you read the result. A homepage that describes the business vaguely produces vague questions, and the question that names you will nearly always get an answer that names you, which says little about being found. A real measurement uses a larger, fixed list you have chosen with care. The lesson on what a prompt set is explains that list, and the free AI visibility prompt generator drafts one for a buying stage.

Chapter 3: Why one grade is a snapshot

The same question, asked again, names different brands.

AI answers are not stable from one run to the next. In a study SparkToro published in January 2026 (opens in a new tab), 600 volunteers ran 12 prompts through ChatGPT, Claude and Google’s AI Overviews or AI Mode a combined 2,961 times. For ChatGPT and Google’s AI, the write-up puts the chance of getting the same list of brands in any two responses at less than 1 in 100. Its author’s own conclusion is that to know an AI’s recommendations you need to ask over and over, usually at least 60 to 100 times, and average the results.

A grade built on 3 answers is far short of that, on purpose: it is a quick first reading, and it is labelled as one. It can show that you are absent, or that a crawler is blocked. It cannot show a trend. For that you need a dated starting point and the same questions asked again, which is the subject of why every fix starts with a baseline and of why the same question gets different answers.

Chapter 4: The checks behind the number

Whether a crawler can read you, and whether an agent could act.

Beside the answers, the grader looks at your site’s setup. The readiness checks ask whether your homepage returns real HTML to a plain request, whether it carries structured data, a meta description and Open Graph tags, whether your robots.txt lets AI crawlers in, and whether you publish an llms.txt. Each check says what it found and, when it didn’t pass, what to change. Three other free tools here go deeper on single checks: the AI crawler checker for your robots.txt, the schema markup generator for JSON-LD and the llms.txt generator built from your sitemap.

The second group asks a newer question: could an AI agent, sent to your site by a person, get something done there? It looks for product or service details an agent can parse, a declared way to take an action, and pages that load without a bot check in the way. The lesson on how AI agents browse and act on a site explains what stops them. None of these checks is a ranking factor with a known weight, and passing all of them promises nothing; the AI readiness checklist puts them in the order they depend on each other.

Best practices

How to read a grade well

  • Do: Read the answers before the numberThe score sums up a handful of answers. The start of each answer is on the page: see who was named and in what light before you quote the grade.
  • Do: Write the date beside the gradeA grade describes one day. Note when you took it, so the next reading has something to stand against.
  • Do: Fix a failed check firstA check that fails names its own fix, and a blocked crawler or missing description is cheaper to put right than a missing reputation.
  • Do: Look at who was named insteadThe competitors in the answers are the ones an assistant reached for in your category. Their pages show what it found worth repeating.
  • Do: Keep mentions and citations apartBeing named and having your page cited are separate results, from separate work. The grade shows both; act on whichever is missing.
  • Do: Move to repeated runs for decisionsOne reading can start a conversation. A budget or a roadmap should rest on the same questions asked many times, over weeks.
Common mistakes

What a grade can’t tell you

  • Avoid: Reading one grade as a trendAsk the same questions tomorrow and the brands named can change. A single grade cannot tell you whether you are gaining or losing.
  • Avoid: Taking the questions for demandThe questions are generated from your homepage. They show what could be asked about a business like yours, not what anyone has asked or how often.
  • Avoid: Treating one engine as all of themThe grade asks one AI engine. Another assistant, reading other sources, can answer the same question with other names.
  • Avoid: Expecting a grade to earn citationsA better grade is a better reading, not a cause. Nothing here promises that a higher score brings mentions, traffic or sales.
  • Avoid: Chasing every check to a passThe checks are signals, not ranking factors with a known weight. Fix what is plainly broken; a pass on all of them guarantees nothing.
  • Avoid: Publishing without reading itA published grade shows the questions, the answers and the competitors named. Read the whole result first; unpublishing is final.
FAQ

Questions people ask

What is an AI visibility grader?

It takes a first reading of how an AI assistant answers questions about your kind of business. You enter your site; the grader reads your homepage, asks one AI engine three questions generated from it, and scores the answers out of 100 by how often they name your brand and cite your pages. The result also lists checks on how readable your site is to AI crawlers and agents.

Are the questions ones real people asked?

No. They are generated from what your homepage says you sell: two ask about your category and one asks about your brand by name. The grader has no record of what anyone has typed into an AI assistant, and neither does CiteCue. Read them as plausible questions for a business like yours.

Which AI engine does it ask?

One engine per grade, so the result says nothing about the others. ChatGPT, Claude, Gemini and Perplexity can each answer the same question with different brands and different sources. Tracking all four over time is what the CiteCue product does; this grader takes one reading from one of them.

How reliable is one grade?

It is an honest snapshot and no more than that. AI answers vary from one run to the next, so a brand named today can be missing tomorrow with nothing having changed on your side. Use the grade to see where you stand roughly and what is broken, and rely on repeated runs of the same questions before you draw a conclusion.

Why does this tool ask for my email?

Because every grade calls an AI model several times, which costs us money, so we ask for an address in return. It joins CiteCue’s waitlist and we may email you about access and product updates. We don’t sell it, and you can ask us to delete it. Your grade appears on this page; it isn’t emailed to you.

How many sites can I grade?

One a day. If a grade fails because the grader couldn’t read your site, that attempt isn’t counted and you can correct the address. The grader also has a daily ceiling across all visitors; if it has been reached, the page says so and asks you to come back the next day.

What happens when I publish a grade?

Nothing is public until you choose it. Publishing creates a page anyone with the link can open, showing the grade as you see it here, and a small badge you can embed that links to it. Both are hosted on CiteCue’s app. The browser you publish from keeps the key that unpublishes it, when it is allowed to store one, so unpublish from this page in that browser. Unpublishing takes the page down for good, and the same grade can’t be published again.

Keep going

More free tools

All free tools

Track what AI says about your brand, not just one answer