Citablio is now live. It answers a narrow, practical question: when someone asks ChatGPT search, Claude, Perplexity, Google AI Overviews or Copilot about your field, can those systems technically reach and read your pages at all?
That question sits underneath the whole "AI visibility" conversation, and it is the part that can actually be measured. So that is the part we built.
Honesty First: Evidence, Not a Promise
There is a lot of noise in this space. "Generative engine optimisation" is often sold as a new discipline with secret levers, and the most prominent tools track brand mentions by sending prompts to AI assistants — useful, but noisy, because the same question asked twice rarely returns the same list.
Citablio starts from a stricter position. It runs deterministic checks against your own site and reports what an AI crawler can and cannot access. It never tells you that you will be cited, ranked or recommended — nobody outside those companies can honestly say that. A clean report means the doors are open. It does not mean anyone will walk through them.
The Mistake We Keep Finding: Training Bots vs. Search Bots
Many sites added "block the AI bots" rules to their robots.txt over the last two years. The intention is reasonable: keep content out of model training. The problem is that AI companies run different crawlers for different jobs, and a blanket rule rarely distinguishes them.
- Training crawlers — for example GPTBot, ClaudeBot, CCBot, and the Google-Extended control token — collect data for future models. Blocking them is a legitimate licensing choice, and Citablio reports it as information, never as a problem.
- Search and retrieval crawlers — for example OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot and Bingbot — build the indexes that answer engines search at question time. OpenAI's own documentation says sites that block OAI-SearchBot will not be shown in ChatGPT search answers.
When we scanned real, well-known sites during development, this was the most common serious finding: a publisher opts out of training, and in the same stroke disallows the search crawler that would have let it be linked as a source. Citablio flags exactly which bot is blocked, by which robots.txt rule, and gives a fix that opens only that bot — while telling you to keep the path rules you still need, because a crawler with its own group ignores the User-agent: * rules entirely.
What Citablio Checks
The scan has nine modules, weighted by how much evidence there is that each one affects AI search. The weights are published on the site and add up to 100.
- AI crawler access (25) — robots.txt parsed to the RFC 9309 rules for 19 AI crawlers, search and training kept apart.
- Render parity (20) — whether the real content is present in the HTML the server sends. Measurements published by Vercel found that the major AI crawlers from OpenAI, Anthropic and Perplexity download pages but do not execute JavaScript, so a client-rendered product description is simply invisible to them.
- Indexability (15) — noindex, nosnippet and max-snippet (which decide whether text can be quoted in AI Overviews), canonical tags, redirect chains and the sitemap.
- Edge bot blocking (10) — whether your CDN or firewall refuses requests carrying a search crawler's user agent even when robots.txt allows it.
- Content structure (10), citability signals (8), structured data (7) and IndexNow (3) — hygiene that keeps pages clear and attributable.
- llms.txt (2) — reported, but deliberately almost weightless.
That last weight is a statement. Google's guidance on AI features says no special files or markup are required, and an Ahrefs study of llms.txt files found that almost none of them were ever requested by AI crawlers. Adding one is harmless; its absence is not a reason to worry, and we will not grade you down for it. Structured data gets the same careful treatment: it keeps your entity information unambiguous, but we do not claim it earns citations.
What Real Sites Taught Us
A scanner that is wrong confidently is worse than no scanner, so we tested Citablio against live sites before launch, and several of them changed the engine:
- Ordinary pages embed reCAPTCHA settings and Cloudflare scripts, so a naive "bot wall" detector mistook two major news homepages for challenge pages. A page now counts as a bot wall only if it actually looks like one.
- Some sites use anti-bot services that answer the same request differently from one minute to the next. When that happens, Citablio says the result is inconclusive and names the protection — instead of inventing a score.
- A homepage canonical that redirects to a regional edition is untidy, not broken, and is now reported that way.
- A deep crawl should not score worse just because it looked at more pages, so page-level findings are averaged across the pages scanned.
Fetching arbitrary URLs is also a security surface. Before launch the scanner went through several rounds of security review, and every request is pinned to an address that has been checked as public, so a hostile site cannot redirect the scanner into private networks.
How It Works for You
The free check needs no account: enter a URL and you get the crawler-access, render and indexability findings for that page, with an exposure band rather than a grade — because we have not verified that the site is yours.
Once you verify ownership (DNS, file or meta tag), Citablio crawls the site, gives an A–F readiness score, and keeps monitoring it. The alert we expect to matter most is the simplest one: someone added a rule that blocks a search crawler. Reports are PDFs with a SHA-256 fingerprint, and anyone can confirm on the public verification page that a report is genuine and unaltered.
Plans start with the free check, then Insight at €19/month, Visibility at €59/month and Agency at €199/month, plus a one-off AI Readiness Report for €99 with no subscription.
Where It's Going
Deterministic checks tell you whether the doors are open. The next question is whether anyone actually walks through them, so we are working on an optional citation probe: a limited, repeated set of questions sent to answer engines to measure how often your site is cited. We will present it the same way we present everything else — as a sampled measurement with its limits stated, never as a ranking.
Why This Matters
More and more first questions are asked to an assistant rather than typed into a search box. The businesses that show up in those answers will not be the ones that bought a trick; they will be the ones whose pages were reachable, readable and clearly attributable when the crawler came by. Citablio makes that part checkable, and keeps it honest.
