Run a scan
Stay up. Get found.

Your site works. Does ChatGPT know you exist?

UpAndFound measures whether AI assistants find, name and cite your brand, then measures again after your change and says whether the difference is real or noise. It starts with 30 technical checks, run live on this page.

  1. 01 Scan Free

    We read your page the way a machine does and name what stops it. Thirty checks, running on this page, no account.

  2. 02 Monitor From EUR 29

    We ask assistants the questions your buyers ask, every day, and record who gets named — you, or somebody else.

  3. 03 Verify From EUR 99

    You change something. We measure the two weeks before and the two weeks after, and say whether the difference is real or noise.

A readable site is not automatically a cited one. Those are two measurements, and this page starts with the first. See the second one

Free. No account. No access to your systems: only what is already public is read.

waiting for an address 0/30
  1. Connection and speed
  2. Content structure
  3. Can AI understand you
  4. Agent files
  5. Security headers

    We keep the domain you checked and a daily rotating hash of your IP, so one person cannot exhaust the free scans for everyone. No account is created and no address is stored unless you ask us to send you something.

    UpAndFound / live

    Our own system, measuring a domain we run. These are real answers from the last 7 days, including the ones where an assistant named somebody else.

    Your website Technical checks 30 checks AI crawlers AI answers ChatGPTAI OverviewsPerplexityGemini

    Live system events

    1. No events in the last 7 days. Nothing is hidden here: when the engines are quiet, the stream is quiet.
    What is being checked

    30 checks, in five passes.

    Search engines and language models do not look at your site. They fetch text and files, and they give up quietly when something is missing. These are the five things they try, in the order the scan runs them.

    1. 01

      Connection and speed

      Does the page answer at all, fast enough, over a certificate that works.

    2. 02

      Content structure

      Is there a title, a heading, real text — the things a reader parses before deciding what this page is.

    3. 03

      Can AI understand you

      Are your facts stated in a form a machine reads without guessing: structured data, canonical URL, contact details.

    4. 04

      Agent files

      Do robots.txt and your sitemap let crawlers in, and do AI agents actually get through when they knock.

    5. 05

      Security headers

      The three headers that decide what may run on your page and how it is transported.

    Reading the result

    What the score counts.

    Not three scores in a row. The engine produces one, out of the same 30 checks every time, and each check is either fine, worth fixing, or broken — you see which, per line, as it runs.

    What it measures is whether a machine can fetch, parse and understand your page. What it does not measure is whether an AI then chooses to quote you. That depends on what you have written and who else has written it. We can tell you the door is open. We cannot promise anyone walks through it.

    What the score cannot tell you

    A clean site is not a cited site.

    Technical health and AI visibility look like the same thing and behave like two different ones. A site can pass every check on this page and still never appear in an answer, while a slower, messier competitor gets named every time. That is not a paradox. They are separate measurements, and almost nobody takes the second one.

    YOUR PAGE Readable any scanner measures this Cited almost nobody measures this ANSWER
    A diagram, not a measurement. The gap between the two gates is the part that goes unmeasured — the real numbers are in the table below.
    One measurement, taken on a real pair of sites
    SiteCited sourcesTimes named
    Competitor 5 5
    Customer 0 0

    The customer had the cleaner site of the two. Every technical check a scanner would run, they passed. The difference was not on their site at all, and no amount of tidying the HTML would have found it.

    How the second measurement works

    1. 01

      Real questions, not keywords

      We ask the assistants the questions your buyers actually type, in their language and market. A keyword has no answer; a question does.

    2. 02

      Three separate facts, never one blended score

      Did it search. Did it find your page. Did it name you. Each can fail on its own, and each has a different fix.

    3. 03

      The step where it breaks

      Knowing you are invisible is not useful. Knowing the assistant searched, found you, and still named someone else is a different problem than never being found at all.

    Up to 4 assistants, asked 3 times on the first day and then once a day in rotation, so a single odd answer never becomes your result.

    30 runs with no mention is not zero. It is 0 to 11 percent, and we print it that way.

    The words on your dashboard, and why each one is there

    Mention rate
    How often an AI answer names you at all. Being readable is not the same as being named.
    Citation rate
    How often your page is one of the sources behind the answer. This is the link a buyer can click.
    Citation position
    Cited second and cited fifth are different business results. Most tools record neither.
    Share of AI voice
    Of everything cited on your questions, how much of it is you. Tells you the size of the gap, not just that there is one.
    Source gap
    The places an AI already trusts on your topic where you do not exist yet. Each one is a task, not a complaint.
    Cited instead
    Who took the place you wanted. Named, with the exact answer they were cited in.
    Confidence interval
    How much the number is worth. A percentage from seven runs is noise with a decimal point.
    Engine coverage
    Which assistants were actually asked. A result from one engine is a result about one engine.

    Where we stop

    After the scan

    Checked daily. Emailed the day it breaks.

    The scan above stays free. What it cannot tell you is what happens tomorrow — and the failures that cost the most are the silent ones: a header changed during a deploy, a sitemap that quietly stopped updating.

    Free and paid

    What it costs.

    Free means free: 10 scans per domain, and up to 20 a day from one network. No account, no card, and the full result every time — nothing is blurred out or held back for an email address.

    Watch

    €29/month

    One site. All 30 checks daily, a liveness check every 15 minutes, and an email the day something changes. AI visibility measured on a weekly rotation.

    You get a number.

    Verification

    Studio

    €99/month

    Everything above, plus the measurement this product exists for: a baseline before your change, a dense 14-day window after it, a control group of questions you did not touch, and the raw answers behind every number.

    You get a verdict.

    Prices are in euros. Any VAT is added at checkout, at the rate your country requires; EU businesses with a VAT number are not charged it. Stripe determines this, not us.

    Check a page. It takes about thirty seconds.

    No account, no card, no email. If the result is dull, you have lost half a minute and learned your site is fine.

    Run the check