Your website has a second audience. See what it sees.
Enter a URL, and Botlight checks which of seven answer engines — ChatGPT, Google AI, Claude, Perplexity, Copilot, Meta AI, Apple Intelligence — are allowed to fetch and cite your page, shows you the exact text a crawler receives, and generates the changes that are missing.
Botlight's first scan of its own homepage came back harsh: 283 words in the raw HTML and no structured data — a citability tool that was barely citable itself. What this page ships today comes straight from that report: JSON-LD that names the app and its maker, head tags a citation card can be built from, an llms.txt and a robots.txt that calls every fetch agent by name. No screenshots, no polished numbers — the example report below runs live, the moment you open it.
Four areas, weighted. The weakest one sets the tone of the report — and if the gate is closed, the headline number is capped instead of averaged into something comfortable.
The full detail — every check, every weight, the score cap and the crawler register with vendor sources — is on the methodology page.
Everything a scan shows is free and complete. Each step arrives the moment it becomes useful, not before — and exactly one thing costs money.
Free · no account
Still free · one confirmation click
Revenue Shield Report · €29 one-time · the only paid thing here
Arrives by email in minutes. Terms and Right of Withdrawal — English translation; the German AGB are the binding version.
The distinction almost every file gets wrong: Training crawlers collect pages to train a model — blocking them costs you no visibility. Index crawlers build the retrieval index an assistant cites from — block one, and you cannot be cited, no matter how good the page is. Live fetch agents open your page the moment someone asks — block them, and the link the assistant wanted to show is dead.
Which token belongs to which system — all seven answer engines, fifteen more crawlers and every vendor source — is on the methodology page. Botlight reads your robots.txt the way the specification prescribes (RFC 9309), not the way it looks.
A scan is a snapshot, and the line that decides it moves without anyone noticing: someone updates a plugin, an agency adds a rule, nothing breaks — you just aren't citable any more. The Watcher is the daily re-check that comes with the newsletter. Once a day the same checks run against the page you scanned, and you only hear from me when an engine verdict flips or the score moves by five points or more. Every one of those emails carries a link that switches it off again.
Because an average would hide exactly the thing that matters. When five of seven answer engines are not allowed to fetch your page, the cleanest markup cannot make up for it — so the headline number is pushed down to what the closed gate allows, and the report says by how much.
It means you are allowed to be cited. Botlight measures permission and legibility — both verifiable from the outside. How often an assistant actually names you cannot be measured from out here, and a tool that claims otherwise is selling you an invented number.
No. The report lives in your browser tab and is gone when you close it. There is no database behind this page. If you request the report by email, the address goes into a newsletter list with a confirmation click first — until that click, nothing is stored.
Everything a scan shows is free and complete — the verdicts, the Bot View, every generated fix. The first scan of the day runs without an address; your email address lifts the counter and unlocks the Site Scan across twelve pages and the daily Watcher; that costs nothing either. The one paid thing is the Revenue Shield Report, €29 one-time: it turns the revenue options the free report names into generated decisions — ready-to-use robots.txt postures, an RSL licence file, the TollBit route, a tracking setup for AI referrals. You will find it in your report, clearly marked as the only thing for sale.
The technical questions — does blocking GPTBot take me out of ChatGPT, why Grok has no tile, whether llms.txt helps — are answered on the methodology page. Botlight comes from Oliver Welsch, is free, needs no account and runs every check live, the moment you ask.
Send a competitor through the same seven answer engines. What's compared is what each page lets a crawler take away — not who gets cited more often. Nobody can measure that from the outside, and anyone claiming otherwise is guessing.
No screenshots, no rendering, no JavaScript — this is the page as a crawler receives it. Everything a model knows about you comes from here.
When an assistant takes over the comparison, your product data is your shelf space.
Generated from this page, with one rule behind every line: a value either comes
from your page or is marked REPLACE_WITH_REAL_… — never invented.
The label on each fix says whether it's ready to paste or needs your values
first.
A single page tells you about a single page. The useful sentence sounds different: the same thing is missing on eleven pages out of twelve. Botlight walks your sitemap, checks up to twelve pages with the same checks and sends the pattern to your inbox.
The image is drawn when someone loads it and cached for a day — so the number follows your website, not the moment you copied this. If a scan fails, the badge shows a dash — never a number nobody measured. The link leads to a scan of your own domain, fresh for everyone who clicks.
One click to go.
The email is on its way to , and in your inbox it looks like this:
It's usually there within a minute. If not, check your promotions or spam folder — nothing is stored until you click the link inside.