Bot Gate: scrapers out, Google in
A geo and behaviour filter with a human check. It stops bulk automated scraping, real visitors pass with one click, and search engines and AI search do not even notice it. We run it on our own 17 sites; we set up the same for yours.
Before blocking, scrapers took on average 10,281 pages and about 78 MB a day from the blocked countries alone. In the last 14 days we refused 11,912 requests (check pages: 9,393); refused requests that looked like real visitors: 0. Human gate: 7,541 checks, 15 passed. Updated: 2026-10-22 — refreshed daily.
How it works
1. Country and network
At request time, in memory, we look up the country and network of the IP address (open, public-domain IP database, refreshed weekly). You set the block list; EU/EEA countries cannot be blocked (EU Regulation 2018/302).
2. Behaviour
From any country we catch clients that look automated: claims to be a browser but sends no language setting; headless or automated browser; outdated browser on an old protocol; script identity (curl, python…).
3. Human gate or block
Per country: gate (a one-click check, ~2 seconds) or full block. The check is open source (Cap) and runs on our own server: a short computation in the browser plus browser signals. Data-centre networks are always blocked.
4. Rate limit afterwards too
After passing: a 24-hour security cookie, but at most a human-paced number of pages per hour. Even if a machine gets through, it cannot download in bulk.
What never gets a gate
This matters most: search ranking (SEO), visibility in AI search (GEO) and link sharing keep working unchanged. Before go-live we tested 40 search, AI and preview robots × 116 pages (4,640 requests): every one got the same response as a regular visitor, none was gated.
- Search: Google (all of its crawlers), Bing, DuckDuckGo, Apple, Yandex, Seznam, Qwant, Mojeek
- AI search and assistants: ChatGPT/OpenAI, Claude, Perplexity, Mistral, Meta, Amazon, Common Crawl
- Link previews: Facebook, X, LinkedIn, Slack, Telegram, WhatsApp, Discord
- Monitors and SEO tools: UptimeRobot, Lighthouse, Ahrefs, Semrush
- For everyone: robots.txt, sitemap, llms.txt, images, API, embeds
What you see in the dashboard
- Refusals per day, hour, site and country
- Bandwidth saved per site
- “Could be human?” — possible false blocks flagged separately
- Edit blocked countries; per country gate ↔ full block
- The last 50 refusals (without IP addresses)
Privacy and law
We do not log IP addresses, only country and category. The gate cookie is for security, so no consent is needed. No third party receives data and no external script is loaded. EU/EEA visitors are never excluded by country. Legal basis: legitimate interest (GDPR Art. 6(1)(f)).
Honest about the limits
No check stops a determined robot that imitates a human perfectly. The goal is to make bulk scraping expensive and slow: the computation costs machine time, automated browsers are filtered, and the rate limit holds back any machine that gets through. Country filtering does not catch traffic via VPNs or residential proxies; behaviour filtering partly covers that.
Who is it for?
Content sites, web shops, database and price pages where scraping loads the server or copies the data. We set it up in front of your nginx server (rule generator + human gate + dashboard), or run it on our EU servers.
Price: individual quote based on traffic and setup — write to info@tilezza.eu.
Does it hurt Google rankings?
No. Search, AI search and preview robots never get a gate, and robots.txt, sitemap and llms.txt are always reachable. We verified this with 4,640 robot requests before go-live.
What happens to a real visitor from a blocked country?
In gate mode, one click and about 2 seconds, then 24 hours of free browsing. In full-block mode the content is not reachable — that is why it is selectable per country.
Is a cookie banner needed?
No: the gate cookie is strictly necessary for security and does not track.