# Features & details > Features, honestly listed: AI findability (GEO), sub-200 ms in the load test, data sovereignty, three languages at the core — and measuring whether it works. btlabs Core · Substance ## btlabs Core: get found, maintain with AI — and see whether it works. Every function in detail. No sales talk, no hidden asterisks: every function btlabs Core brings with it is listed here. btlabs Core is the digital foundation from Berger+Team in Bolzano — readable for people, dockable for AI assistants. So you can check what you get before deciding, instead of finding out afterwards. Index **01**Found by AI**02**On Google**03**Measuring**04**Chat upkeep**05**AI assistance**06**Multilingual**07**Editing**08**Speed**09**Data sovereignty**10**Security**11**Operations**12**Enquiries**13**Media**14**Search**15**Interfaces 2 ### Channels: Google & AI answers [SEO and GEO →](https://btlabs.dev/en/seo-and-geo) 3 ### Languages at the core [Multilingual website →](https://btlabs.dev/en/multilingual-website) 4 ### Levels of measurement — down to AI mentions [Measuring visibility →](https://btlabs.dev/en/measuring-visibility) <200 ms ### Response time in the load test [Found on Google →](https://btlabs.dev/en/found-on-google) 0 ### Cookie banners out of the box [When a cookie banner isn't needed →](https://btlabs.dev/en/posts/why-a-cookie-banner-is-often-unnecessary) 100 % ### yours — content and data ## Eight questions, one answer. Who keeps it current? Who owns it? Does it pay off? And in two years? — The answers in 56 seconds, before you get into the details below. Text on screen, sound optional. 01 · Found by AI ## So your business can show up in the AI answer — with the right facts. People looking for a recommendation today often ask ChatGPT or Perplexity directly — and get the answer without opening a website. So btlabs Core ships a clean, [machine-readable](https://btlabs.dev/en/glossary/machine-readable-website) version of your business: facts complete and free of contradictions, retrievals visible. Whether an AI names you is its own decision — but what it finds at your place is correct. llms.txt Text ``` # Your business — readable by AI in one pass name: Example Joinery location: Bolzano · South Tyrol services: - Custom furniture - Interior fit-out - Repair & service opening_hours: Mon–Fri 8–17 contact: info@example-joinery.example languages: de · it · en ``` If someone asks an AI for “joiner in Brixen”, it can read exactly this — services, hours, contact — instead of guessing. 01 ### A table of contents for AI systems (llms.txt) An assistant does not have to crawl your whole website: the [llms.txt](https://btlabs.dev/en/glossary/llms-txt) names your business, services and most important pages — compact, plus a detailed version. 02 ### Every page as plain text [Language models](https://btlabs.dev/en/glossary/llm-large-language-model) read your statement instead of the packaging: your pages have a tidy text version without menus and layout — and every page declares the way to it technically. 03 ### Your machine-readable brand profile (brand.json) One source instead of twelve directory entries that all claim something different: company name, contact routes, service area, services and tone of voice sit structured in one place. 04 ### FAQ as a question-and-answer list "Are you open at the weekend?" can get a reliable answer instead of a guessed one: your FAQ ship additionally as a pure question-answer file — a format assistants can read directly. 05 ### MCP: the interface for AI assistants A defined entry point leads assistant systems to your public data — pages, posts, jobs, team, FAQ — as ordered data instead of copied screen text. Searching and fetching the contact route works for an assistant right away. 06 ### Profiles where machines look (.well-known) Whoever searches for you by machine does not give up halfway: the technical descriptions sit at the storage points where automated systems expect them. 07 ### Citing yes, training no (TDM reservation) Your usage reservation is stored machine-readably at the places where serious providers check — based on EU law, in writing and dated. 08 ### Three switches you set yourself In answers yes, in training no, attribution required: you decide separately what AI systems may do with your content — with a click, not a lawyer. 09 ### Curated bot lists (robots.txt & ai.txt) Assistant and search systems are welcome, pure scrapers are told no (a signal, not a technical block) — and we keep the lists current. You do not have to track which new robot is making the rounds. 10 ### Preview services stay allowed When someone shares your link, the preview stays intact: those services are exempt from all AI switches — otherwise your link would break exactly when someone is promoting you. 11 ### A contact route for reports (security.txt) If someone finds a problem, they know where to report it: one file names the reporting route, a second one the people behind the website. 02 · Found on Google ## Search engines understand your business — instead of seeing a page with text on it. Nobody can seriously promise rankings. What btlabs Core does: create the conditions for Google & Co. to understand your business unambiguously — automatically, without anyone having to think of it. 01 ### Your company data as structured data Google can recognise you as a business, not just a page with text: business, opening hours, service area, services, FAQ, jobs and team are additionally output as [JSON-LD](https://btlabs.dev/en/glossary/structured-data) following Schema.org — the language search engines and AI systems understand without guessing. 02 ### Only what is visible gets marked up What gets marked up is what is also visible on your website — that's how the system is built. That is exactly where most who mean too well fail — and lose the enhanced display. 03 ### Title and description per page and language You decide which sentence represents your page in the result list — with a preview before you save. So you choose what a searcher reads first, not chance. 04 ### Readable addresses, 301 redirect included Rename a page and the system creates the [301 redirect](https://btlabs.dev/en/glossary/301-redirect) by itself. Last year's business card still leads to the destination. 05 ### An always-current sitemap What is published is in it — drafts and expired job ads drop out automatically. Nobody maintains a list, and nobody forgets it. 06 ### New content reported immediately (IndexNow) With [IndexNow](https://btlabs.dev/en/glossary/indexnow) switched on, every published address goes straight to the participating search engines — Bing among them. Google finds it via the always-current [sitemap](https://btlabs.dev/en/glossary/sitemap) . 07 ### One main address instead of three near-identical pages If content is reachable via several routes, the preferred address is declared. The effect gathers on one page instead of spreading out. 08 ### You decide what enters search The internal price list or the half-finished promo page stays out — one switch per page, without anyone touching the tech. 09 ### Job ads Google Jobs understands Your jobs are maintained with an expiry date and output machine-readably; expired ones disappear on their own. Applicants only see what is open — and nobody tidies up the careers page. 10 ### A proper preview when shared (Open Graph) When a happy customer drops your link into a group, it looks good: title, description and preview image are prepared as [Open Graph](https://btlabs.dev/en/glossary/open-graph) data and refreshed when you swap the image. 11 ### Your blog as a feed, per language Trade portals, distributors and readers can pick up your posts automatically — without you submitting anything anywhere. 12 ### Clean structure, linkable sections Headings, breadcrumbs and directly addressable sections: you send a customer exactly the part that concerns them — instead of "scroll down a bit". 13 ### Every page checked before release How will it appear in search, what is still missing? An empty description gets noticed before publishing — not months later over coffee. 03 · Seeing whether it works ## You see whether it works — instead of believing it. The question every business owner asks second: how do I know this is doing anything? btlabs Core answers on four levels: who visits the site. What people find through search. Which AI systems read along. And: are you named in AI answers, even when nobody clicks. The numbers come from different sources and are deliberately not added into one pretty total. visibility-report.txt Report ``` # AI visibility — monthly snapshot question: “Who makes custom furniture in Brixen?” chatgpt → recommended · rank 1 perplexity → mentioned · rank 3 gemini → not yet named --- ai visits (30 days): 142 · verified visitors from AI answers: 38 ``` This is what measuring looks like: the monitor asks on its own, the answers get scored — you see in black and white which AI recommends you in this sample and which doesn't name you. 01 ### Your visitor statistics, without cookies The analysis runs on your own server — no external analytics service, no cookie for the visitor. You see what is going on without handing your customers' data to an ad corporation. 02 ### The actions that matter Counted is not just the visit: whether someone truly read, whether they clicked your phone number or e-mail address. You see whether someone tapped your number — not just that they were there. 03 ### Numbers for every single page Views, enquiries, clicks from search, positions and AI referrals come together per page. You see which service page works and which merely takes up space. Also in conversation: "Which page brought the most enquiries last month?" 04 ### Where your visitors come from Direct, search, other websites, campaign — cleanly separated. So you see whether the flyer, the ad or the search actually brings people. 05 ### Visitors from AI answers, recognised separately The known AI tools are tracked as their own source. You see people who found you through an assistant — as a lower bound, since some apps don't pass on where visitors came from. 06 ### AI accesses counted separately — and verified Which AI systems read your content, how often and which pages, runs separately from human visitors — and, where providers publish their address lists, accesses are checked for really coming from them. Otherwise you celebrate a success a scraper produced under a false name. 07 ### Assistant or scraper — sorted properly Every machine visitor is classified by kind and intent: [AI assistant](https://btlabs.dev/en/glossary/ai-agent) , classic search, [SEO](https://btlabs.dev/en/glossary/seo-search-engine-optimisation) tool, preview service. Requests from data centres do not count as AI interest — without that separation the numbers would look many times better than they are. 08 ### An AI's path through your website Consecutive accesses are joined into a path: you see whether an AI skims the homepage or works through your services. 09 ### Is the AI preparation being used? For every machine-readable file, retrievals are counted. That proves whether the files get read — instead of assuming it. 10 ### Google Search Console connected Once your account is connected, the system fetches your search data at night: the real queries, pages, countries — on first run back about a year. Also in conversation: "Which terms do people find me with right now — and does the page text match?" 11 ### Bing Webmaster Tools connected Also nightly, with queries and referring links. Bing feeds the answers of a major assistant system — that view is worth more today than it used to be. 12 ### The citation monitor: being named without a click If an AI recommends your business without anyone opening your site, no data point exists at your end. So the monitor asks by itself: your test questions go to several large language models regularly, and the evaluation shows whether your business is named — from "not at all" to "first recommendation". Switchable on, set up on request. 13 ### Who you are really up against in AI answers From the same answers, it derives which other providers are named. Often they are not the ones you suspect around town. 14 ### What stands in the way on your website The whole site is checked daily: orphaned pages, dead ends, broken links, pages competing for the same query. Every page gets a grade — the lever a business without an ad budget can move itself. 15 ### The quick wins Queries where you sit just behind the first result page are singled out. Exactly there, an afternoon of work can beat a new campaign. 16 ### "Not retrievable" instead of a zero If a value cannot be fetched, that is what it says — never a zero that looks like a result. Every metric carries its source. That is the difference between an analysis and a gut feeling with a chart around it. 17 ### Enquiries and goals as a file Everything gathered can be exported as a table — for your CRM or the annual review. 04 · Run your site by chat ## You say it — instead of clicking through menus. Your website can be connected to an AI assistant that works with the whole system: your numbers, your pages, your enquiries, your languages. Reading, judging and acting come together in one conversation. The time saved is not that one task goes faster — it is that tasks get done at all, because the hurdle drops from "I would have to sit down for this" to "I will just say it". your AI assistant \~ MCP You Set the summer opening hours: Saturdays until 5 pm. Core Done — saved as a **draft** in DE·IT·EN. Shall I put it **live** right away? You Yes, go live. Core Published — correct everywhere: page, sitemap, machine files. MCP · authorised logged live in DE·IT·EN 01 ### An assistant that actually pitches in It reads and edits content without anyone logging into the editing system. For a business without a marketing department, that is the difference between "when I find the time" and "done". 02 ### Analytics in conversation Visitor numbers, the real queries from Google and Bing, enquiries per page, AI accesses: "Which pages brought the most enquiries last month?" — and you get the answer instead of four open tools. 03 ### Have it judge "Go through my service pages and tell me what is missing or contradictory." — "Which pages are orphaned?" The assistant knows your content and your numbers and lays one over the other. 04 ### Create, change, schedule — by saying so "Create a bathroom renovation service page and hang it into the main navigation." — "Translate the post into Italian and store it as a draft." — "Schedule it for the start of next month." 05 ### You grant what it may do With an access key, separately for every area and every action; a sign-in by consent gets full access. For external access, the default is: read and write, but not delete. 06 ### Every access is recorded Even pure reading goes into the log — with tool and time. No blind trust required. 07 ### Secrets stay secret Credentials, keys and internal notes are automatically redacted in chat access. The assistant sees your content, not your keys. 08 ### Sign in by consent Common AI chats dock on via the [Model Context Protocol](https://btlabs.dev/en/glossary/mcp-model-context-protocol) (MCP) with the usual consent flow — instead of a key being passed around by mail. 09 ### It does not guess, it looks The assistant searches your pages, posts, FAQ, jobs and team before answering. So it relies on your information rather than guesswork. 10 ### It stays inside your brand design Colours, typography and spacing can be queried. What it produces does not come out garish. 11 ### Look first, then publish Every draft can be checked as a preview before it goes live — and built-in brakes slow down a runaway assistant. 12 ### Taking in images The assistant fetches images from public addresses and stores them with alt text, title and attribution — labelled properly instead of uploaded bare. 13 ### It reports what it notices Contradictions, tone deviations, gaps and translation problems land in an inbox a human reviews — the AI does not change your house rules itself. 14 ### Third-party assistants: public data only An AI service can answer customer questions correctly without access to your system — it gets the public data, nothing else. 05 · AI assistance built in ## The AI does the legwork — you decide. Artificial intelligence takes over the editorial grunt work here. It runs on your business's own AI account, follows your brand and language rules — and none of it replaces release by a human. 01 ### Translation proposals in your tone Whole documents are translated with formatting intact, your terminology and your tone of voice. An afternoon of translation work becomes reviewing and polishing. 02 ### Titles and short descriptions From an existing text come proposals per language — preventing the empty required fields that otherwise pile up. Choosing stays with you. 03 ### Image descriptions nobody else writes With image descriptions switched on, every newly uploaded image gets a title, alt text and description in every language, plus the focal point for cropping. The grunt work almost everyone fails at — and it decides accessibility and image search. 04 ### Pre-filling fields A few keywords become a draft. Better than the empty form you sit in front of and never start. 05 ### One view across the whole website Checked every night: broken links, pages without an [inbound link](https://btlabs.dev/en/glossary/backlink) , weak search snippets and pages competing for the same keyword — plus the search data from Google and Bing. You see where to improve; changes happen only with your release. 06 ### Your house rules stay in shape The stored brand and language rules are checked for duplicates, contradictions and outdated bits. So the rulebook does not decay into a pile of notes. 07 ### Suggestions for what to link For a draft, fitting targets are proposed — pages nobody links to yet come first. One of the few levers that works without an ad budget. 08 ### Improving search snippets per page Your pages get reviewed and you receive concrete improvements to adopt — no automatic corrections behind your back. 09 ### The tone stays yours Your tonality, your preferred terms and the words that have no business at your place flow into every AI job. That way your texts are meant to sound like your business, not like a text generator. 10 ### The right model for each task Translating, image description and pre-filling can be set separately — a cheap model for the simple, a strong one for the hard. Switching without any engineering. 11 ### Your account, your usage, visible The AI key is stored encrypted and shown only masked; every call is recorded with its usage, and per-person limits protect against the accidental mass run. 12 ### Nothing goes live without release AI texts on pages and posts land in draft. For image texts, brand profile and menu items the rule is instead: only empty fields get filled, existing content is never overwritten. 06 · Multilingual at the core ## German, Italian, English — equal from the ground up. Not bolted on afterwards. Switch live ### Multilingual at the core. Non-negotiable in South Tyrol — which is why multilingualism is part of the foundation. Every page, every AI file and every machine response exists per language, each with its own web address. DEITEN routing /de 200 btlabs.core/de /it 200 btlabs.core/it /en 200 btlabs.core/en · llms each language gets **its own file** [More about the multilingual website →](https://btlabs.dev/en/multilingual-website) German, Italian and English are equal language versions with their own addresses and their own SEO — not a bolted-on translation. Non-negotiable in South Tyrol, a head start in the DACH region. 01 ### Three languages, one source Every language has its own texts, addresses and search snippets — but everything sits in the same data base. One website with three versions instead of three websites with triple upkeep. 02 ### Own addresses per language Your Italian page has an Italian address — not the German one with a code behind it. The customer you send the link to notices. 03 ### Search engines see what belongs together (hreflang) The language versions are marked as alternatives of each other. So your Italian page does not count as a copy of the German one. 04 ### Search snippets per language The Italian version does not have to adopt the German phrasing — it can hit the terms actually searched for in the Italian market. Often a different word, not just a translation. 05 ### You see where something is missing For every piece of content you can tell whether a language version is complete, half done or empty. The gap catches your eye — not your Italian customer on the phone. Also in conversation: "What is missing in Italian?" 06 ### A copy is not a translation If the Italian field holds the same German text, that shows as its own state — instead of being waved through as "done". 07 ### Rather nothing than German on the Italian page If a page is not translated, it is deliberately not reachable in the other language. No Italian customer lands on a German page and thinks the business sloppy. 08 ### The admin interface is trilingual too Whoever maintains content works in their language — down to every field label, help text and error message. No English "Submit" under an Italian form. 09 ### Even the error page The "page not found" page is a designed page per language — a typo in an address leads on to your services instead of into a bare error message. 10 ### Clean translation handovers, no layout chaos All translatable text parts can be exported cleanly and played back just as cleanly. A translation agency can work with it without wrecking the formatting. 11 ### Whole batches at once Translate several documents in one pass, formatting intact, using your terminology. Also in conversation: "Translate the new service pages into Italian and store them as drafts." 12 ### The search knows inflections Whoever searches "renovation" also finds "renovations" — in all three languages. 07 · Content & editing ## So your website is still right two years in. A good editing interface needs clear states, data you maintain only once, and guards that show mistakes before the customer sees them. Most of it runs in the background — the best guard is the one you never notice. 01 ### Around 20 areas for your content Pages, posts, FAQ, glossary, job ads, team, testimonials, media, menus, text modules, notice banners, company data, enquiries, mailing list, campaign links, redirects — each area with what belongs to it. 02 ### Twenty blocks your team builds pages from Text, image, card grid, gallery, form, search field, automatic lists, team, testimonials and more — even mid-text. Your team assembles new pages itself, without calling us. 03 ### You see where a block is used Before you change something, you know what you are touching. That spares you the unpleasant discovery three pages later. 04 ### Preview in three device sizes Changes appear immediately in the real design — for phone, tablet and screen. A display glitch catches your eye, not your customer's. 05 ### Draft, release, schedule You prepare in peace and publish deliberately. Content can publish itself and withdraw itself: the seasonal promotion goes live on time without anyone getting up early. 06 ### Bring back earlier versions Older versions of every piece of content are kept, and the system keeps saving along while you edit. An accidentally overwritten text is no longer a drama. 07 ### Trash What you delete in the interface lands in the trash first and can be brought back — deleting via chat, by contrast, is final. 08 ### Who changed what Logins, changes and external access are logged. When it is "that wasn't me", there is an answer. 09 ### Company data maintained once Address, phone, opening hours, tax number sit in one place and appear everywhere automatically — in e-mails and forms too. New number? Change one field, not search thirty pages. 10 ### Text modules and notice banners Maintained once, identical everywhere — and notice banners with start and end date: entering the company holidays takes minutes, and the banner removes itself. 11 ### Assemble menus instead of programming them You build the navigation yourself, with your own labels per language — and what belongs on top, you drag to the top. 12 ### Explain terms once Your glossary links itself in body text automatically, synonyms included. That helps the customer who does not know the term — and shows search engines and AI systems what you stand for. 13 ### One answer, three places of effect Your FAQ sit on the website and are ready, machine-readable, for search engines and AI assistants. Written once, used three times. Also in conversation: "Add the three questions customers ask me on the phone." 14 ### Your people with a face People with role, short bio and qualification, linked to their posts. Expertise gets a name — that builds trust and counts for search engines too. 15 ### Testimonials maintained cleanly References are captured in a structured way and carried into the machine-readable data. 16 ### Trace which advertising worked For ads, flyers or social posts you assemble trackable campaign links. Afterwards you know which action brought enquiries — and which only cost money. 17 ### Internal knowledge stays internal Price lists, procedures and notes sit apart from public content, searchable only for the authorised. The knowledge that otherwise vanishes in mail attachments stays findable. 18 ### Important pages are protected from accidents Search, FAQ overview, thank-you page and error page can be designed but not accidentally deleted. Nobody kills the contact route by mistake. 19 ### The Italian mandatory details are provided for VAT ID, tax number, chamber of commerce, REA number, SDI code and the certified e-mail address have their own fields. For a business in Italy that is mandatory — and a foreign word in most systems. 20 ### The quiet helpers in the background Addresses generated cleanly, redirects on renaming automatic, internal references surviving address changes, broken links detected, formatting leftovers cleaned. Exactly the mistakes that never get tidied up on grown websites. 21 ### What you do not need, you do not see Blog, jobs, FAQ, testimonials, glossary can be switched on and off individually — with everything that belongs to them. Three roles separate cleanly: administrator, editor, reader. 08 · Speed ## The page is simply there — even with twenty people on it at once. Speed is not a number you write down once — it is the result of several decisions. And it is the first impression: whoever waits three seconds for your page does not call. 01 ### Finished pages instead of rebuilding per click Pages are delivered ready-made instead of being assembled on every visit. Even with many visitors at once, delivery stays fast — in the load test (August 2026) up to 40 concurrent requests. 02 ### Changes visible immediately When you save, the affected page is fresh at once — the rest of the site catches up in the background without visitors waiting. No "wait until the cache clears", no nightly run. 03 ### A net underneath Several caching layers interlock. If something jams, the page keeps serving the last good version. 04 ### Measured, not claimed In the load test (August 2026, one cached page, up to 40 simultaneous requests) the median response time stayed under 200 ms. How fast our site reaches you, you can check yourself: open it and compare it with yours. 05 ### No grown tower of extensions The public website does without the stack of interdependent add-on modules you otherwise collect over the years. Less ballast means: faster, fewer update conflicts, less that can break. 06 ### Fonts from the device, not a third-party server Out of the box, the fonts the visitor's device already has are used — no loading, no flash, no connection to a font service. A brand typeface is possible and then planned deliberately. 07 ### Images load when needed What sits far down loads when someone scrolls there. That makes the first impression far faster — on a phone, in the valley, with one bar of signal. 08 ### Updates without an error page During an update, requests are held briefly instead of rejected with an error page. Your visitors see no error page; during the short switch-over a page at most takes a little longer to load. 09 · Data sovereignty & privacy ## Your data belongs to you — and your visitors' data to nobody else. The way we deliver, your website sets no cookies for visitors, loads nothing from third-party servers and contains not a single ad tracker. That is why it needs no consent window in normal operation — our reasoned position on the delivered state, not legal advice. 01 ### No third-party trackers — none is in there No ad tracker, no ad pixel, no external analytics service. There is nothing to switch off, because nothing is built in. 02 ### No cookie for your visitors A single cookie exists — for logging into the editing system. That is the technical basis for needing no consent window in normal operation. If you later embed third-party content — video, booking widget, review widget — you need one; we tell you beforehand which embed triggers it. 03 ### Nothing is loaded from third-party servers No scripts, no fonts, no images from outside — no exceptions. 04 ### No IP addresses in statistics and customer data There sits a regularly rotating code from which nothing can be traced back. The technical server log keeps addresses for a few weeks — that too is stated openly in the privacy text. 05 ### A deletion concept that actually exists For 16 kinds of data it is set down in writing how long they stay and why — and the clean-up runs automatically at night. In an audit that is the first question, and here it is answered. 06 ### Error logs without personal data Before anything is logged, personal details and credentials are removed. Diagnosis yes, data collection no. 07 ### Legal texts in three languages included Imprint, privacy policy and terms as maintained pages, in the format for an Italian SME, with third parties named openly — a scaffold that gets adapted to your business. 08 ### Your own server, your own installation No shared host, no database shared with strangers' websites. If something goes wrong elsewhere, that is not your problem. 09 ### Customer data and backups in the EU Operation and backups sit in European data centres. The AI text work runs via your own AI account — with released website content, never with your visitors' data. 10 ### Your content belongs to you Texts, images and company data are assigned to your business, and the core system hangs on no ongoing usage lock-in — and content, media and settings can be exported in full at any time. Built once, not rented forever. 11 ### No fee per login No recurring payment per person. Everyone who should maintain something gets their own account — instead of sharing one for cost reasons. 10 · Security & access ## Security and operations run on their own — every night. Backups, security updates, patches — automatic. You take care of business. btlabs-core \~ night-run \[02:00\] OK backup → encrypted · **multi-copy** \[02:14\] OK os-updates → applied \[02:31\] OK containers → freshly rebuilt \[02:47\] OK cve-scan → no hits \[03:02\] DONE runs on its own — **you sleep** 01 ### Encrypted connection The connection is protected, the usual browser protections are set, and the server gives out no technical information. 02 ### Two-factor sign-in Editorial accounts can be secured with a one-time code and recovery codes. A guessed password alone is then no longer enough — we recommend it for every account with publishing rights. 03 ### Short passwords are not accepted at all The measure with the best effort-to-effect ratio — and the one missing most often. 04 ### After several failed attempts, that is it Account temporarily locked, attempts per connection limited. Automated guessing runs into the void. 05 ### Brakes where it gets expensive Almost all access routes have graduated limits, depending on how costly an operation is — which slows down abusive mass access at application level. 06 ### No planted actions Every changing request is checked for really coming from your website — the standard protection against a foreign page triggering something in your name. 07 ### Everything that comes in gets cleaned External content and uploaded files are stripped of possible malicious code in several layers — the logo someone sends you does not become a way in. 08 ### Credentials encrypted The accounts for connected services sit encrypted and are shown only masked. 09 ### Permissions apply at the interfaces too What an assistant sees, what a public retrieval delivers and what a role in the interface may do are regulated separately. Three layers instead of one switch. 10 ### Misconfigured means: does not start On boot, the system checks the critical settings and refuses to start when in doubt. A wrong setting becomes a start-up error — not a data leak in operation. 11 ### Automated scans run into the void The scan patterns actually observed are rejected at the entrance before they cost computing time. Such scans hit every public website all day — at yours, the common ones end at the door. On the network level, whatever is not needed is closed. 12 ### Checked weekly against known vulnerabilities All components in use are matched against the public vulnerability databases; severe findings are reported and closed. And before every go-live, security tests run — a flaw shows up while building, not at your customers'. 11 · Operations & reliability ## If something does happen, there is a plan. Reliability means: changes run controlled, faults become visible, data stays restorable. That takes technology — and it takes clarity about who does what. 01 ### Backed up every night, encrypted Database, images and settings — the database backup encrypted. Nobody has to think of it. 02 ### Backups nobody can delete The last thirty days of backups are locked against deletion at the storage provider — even someone with captured credentials cannot reach them. Plus monthly archives and an image archive that only grows. 03 ### A copy outside the data centre A copy is pulled out daily; missed runs are caught up. If the cloud provider fails, that is not the end. 04 ### Restoring is rehearsed For playing backups back there is a dedicated test procedure against a throwaway database. A backup that has never been restored is not a backup. 05 ### Clear target values for the emergency Target: back online within two hours, at most one day of data loss (up to three days if the whole cloud provider fails) — several emergency scenarios are rehearsed with a handbook. In an emergency, nobody improvises. 06 ### A checking chain runs before every go-live Coding style, trilingualism, database discipline, the machine-readable files — plus an extensive automated test suite. What does not boot fully and answer does not go live at all. 07 ### After the update, someone looks The most important addresses are queried in live operation — including the ones nobody ever looks at. Not just the homepage. 08 ### Watched from outside and inside An external monitor checks reachability continuously; website, database and statistics additionally check themselves. A hung service is restarted before anyone calls. 09 ### Security updates run automatically The restart happens at night and only when needed. The components in use are renewed weekly, with automatic rollback — if a fix exists for a vulnerability in a third-party component, it arrives with the next renewal instead of lying around for years. 10 ### Back to yesterday, in minutes Earlier versions stay available; the running one can never be deleted. A rollback is a normal procedure, not a rescue operation. 11 ### Daily accessibility check The website is checked automatically against [WCAG](https://btlabs.dev/en/glossary/web-accessibility) 2.1 AA — the machine-checkable part, daily, with a report. For the new obligations, a foundation to build on. 12 · Enquiries & contact ## The moment your website pays for itself. An enquiry is the point where a visitor becomes a customer. So this is about the whole route — from the filled-in form to the documented consent. 01 ### Build forms without programming From ready field types you assemble what you need: text, e-mail, number, select, checkbox. A call-back form stands in ten minutes — without calling anyone. 02 ### Errors explained while filling in What is missing or wrong appears right at the field. The visitor corrects it themselves — and you chase fewer incomplete enquiries by phone. 03 ### Spam protection without picture puzzles Common machine spam is rejected without a real customer having to click traffic lights — and a brake kicks in during spam waves. Your inbox does not flood. 04 ### Every enquiry is kept in the system Each one sits in the system regardless of whether the mail arrived — and can be exported as a table. No order vanishes because an inbox was full or the mail landed in spam. 05 ### You see where the enquiry came from Campaign, source page, country and time are stored along. That makes it provable which channel actually brings you customers — and the supplied privacy text describes exactly that. 06 ### Notification and confirmation of receipt Two mails are prepared — one to you, one to the prospect, each in three languages, with company data inserted automatically. The prospect hears within a minute that their enquiry arrived. 07 ### A thank-you page that leads on Instead of a grey confirmation line comes the next sensible step: directions, references or the phone number for the urgent case. 08 ### Mail sending your way Either via a specialised sending service — then, once set up, you also learn when a mail does not arrive — or via your own mail server, with a test send. 09 ### A mailing list that stands up to an audit Sign-ups run through genuine [double opt-in](https://btlabs.dev/en/glossary/double-opt-in) : time-limited confirmation link, documented history, automatic deletion without confirmation. If there is ever a dispute, you can prove consent existed. 10 ### Unsubscribe, access, deletion — without you For list subscribers, all three routes run via a personal link, no account needed. Data-subject rights handle themselves without anyone at your end lifting a finger. 11 ### Old data disappears by itself Form enquiries are deleted after half a year, unconfirmed sign-ups after a few days. Data minimisation happens — instead of staying a good intention. 12 ### Your business systems can help themselves Enquiries and forms can be fetched by your ERP or CRM — via open routes, no island. 13 · Media & images ## Images that load fast — and don't give away where a photo was taken. Images are the most common reason for slow websites — and the most common source of data published unintentionally. Both get handled in the background here, without you having to think of it. 01 ### Everything in one place Images, videos and documents sit centrally — one library instead of five folders on three computers and a USB stick in a drawer. 02 ### Every image in the right size From one photo, several sizes are generated automatically, plus a crop for sharing. A phone loads the small version, a screen the large one — nobody loads the original straight from the camera. 03 ### No more cut-off heads You define what matters in the image — all crops follow automatically. 04 ### Readable file names "IMG\_4711.jpg" becomes a name you understand. That helps in image search and keeps the library searchable two years on. 05 ### Location and camera data get stripped Capture location, camera model and the like are removed from photos automatically. Your home address does not hide in the company photo just because it was taken at home. 06 ### Supplied graphics are checked Logos and graphics are cleaned of possible malicious code before being stored. A logo someone sends you does not become a way in. 07 ### Image rights under control Author and licence are maintained on the image and can be exported together. When someone asks, you do not dig through old invoices. Also in conversation: "Fetch the image from this address, label it and enter the author." 08 ### AI images stay recognisable as such If an image carries the simple machine-readable label as AI-generated, it is preserved (a cryptographically signed provenance record does not survive processing) — relevant for the transparency duties of the European [AI Act](https://btlabs.dev/en/glossary/ai-act-eu-ai-regulation) . 09 ### Icons for browser tab and home screen From one template come all needed sizes — a detail almost always missing on grown websites, and one everyone notices anyway. 14 · On-site search ## Visitors find what they came for — instead of leaving. From about thirty subpages on, your own search decides whether a visitor finds what they are looking for — or looks for it elsewhere. 01 ### Built-in full-text search It searches pages, posts, jobs, glossary and FAQ; drafts stay out. No external search service, no data transfer, no monthly fee. 02 ### Even with typos If the normal search returns next to nothing, a typo-tolerant search over the titles kicks in — "renovaton" still leads to the target. 03 ### You see beforehand whether it fits Every hit comes with an excerpt and highlighted matches. 04 ### You learn what your customers search for The search terms are recorded anonymously. That is the most honest topic list there is — your customers tell you there which service is still missing from the website. 05 ### Internal search kept separate The internal knowledge store has its own search, open only to the authorised, strictly separated from the public area. 15 · Interfaces & channels ## One data base — more than one website. Because content and presentation are separated, more than a website can grow from the same base — and your other systems do not maintain anything twice. 01 ### Your other systems can help themselves Released content is available via open interfaces — ERP, CRM or an app fetch it. Your opening hours are not maintained twice in three systems. 02 ### You see whether the access points are used What is offered publicly shows up in the analysis. So you can tell whether the preparation gets used — or was merely well meant. 03 ### Campaign pages from the same base A promo page draws on existing company data and blocks instead of being maintained separately. It stands in hours instead of weeks — and the contact data is right automatically. 04 ### Take your settings with you Configurations can be exported and imported — for a test environment or a second presence, say. 05 ### Social profiles with one paste Paste a profile address, the platform is recognised, the right icon set. The bar in the footer is a paste job, not a design task. 06 ### A design of your own, not a bought template Colours, typography and spacing are a system of their own. Your website does not look like ten thousand others — and an AI assistant can query it so that what it produces stays inside your design. 07 ### Prepared for more Further brand, location or topic sites can technically draw on the same base — planned as its own project, carried by the same foundation. **Extendable on request:** not standard today, but buildable to fit the foundation: a complete **shop and inventory module** with payment integration · a protected **client and tenant area** (exclusive pricing, documents, portals) · a **live price calculator** · a “next available slot” display with booking · an **app or second location** on the same base · deeper integration with **inventory and customer management** · dedicated **location pages for branches** — each with local findability on Google and in AI. **And looking ahead:** your foundation is built as a service interface — the next step is AI assistants that, with your approval, don't just read but act: checking availability, requesting an appointment, preparing an inquiry. The foundation for that is already in your system, via the MCP interface. Tell us what you need — we'll build it. ## Further reading. - [· What is already built in, while others are still debating · The knowledge about AI visibility is freely available. It is almost nowhere delivered: of 8,636 measured domains, 14.1% serve an llms.txt and 0.1% a machine-readable catalogue. On the gap between talking and building.](https://btlabs.dev/en/posts/what-is-already-built-in) - [· "I Don't Know What Our Website Actually Does For Us" — The Visibility Cockpit · Gut feeling instead of numbers: a visibility cockpit shows website visits, AI-bot reads, Google clicks and inquiries in one place — a made-up before/after example.](https://btlabs.dev/en/posts/visibility-cockpit-what-does-my-website-bring) - [· Why an AI cannot represent your brand · An AI can quote your text. It cannot reproduce how you look — it does not know. Of 18 design-system domains, none publishes its design tokens at a discoverable address.](https://btlabs.dev/en/posts/machine-readable-brand-design-tokens) FAQ ## Common questions about the substance. **Can I really maintain content myself?** Yes. Block system with live preview — texts, images, whole pages without code, with draft/published status. And if you like: straight from your chat or AI system (MCP). **What happens if I break something?** Almost nothing is final: every change is saved as a version — compare and restore with one click. There are always a few exceptions, but for day-to-day editing you're covered. **Can I restore what I've deleted?** In the editing system yes — deleted items go to the trash and can be brought back. Through the machine interface, meaning in conversation with an AI assistant, deletion is final. That is exactly why this one permission can be switched off per access; for external access, write-without-delete is the sensible default. **Do I have to take care of maintenance, backups or security?** No. EU hosting, isolated per partner — backups and security patches run automatically in the background. Larger system updates we sort out together when they come up. **Where does my form data end up?** In your own database — not with third-party providers. Exportable for your CRM or fetchable from there via open interfaces; a direct, automatic hand-over to the CRM is a later expansion stage. Double opt-in applies to the newsletter list, not to form enquiries. **Do I get the same website you have?** The same foundation, not the same site. You start with around twenty standard blocks your team assembles pages from. Our own website uses considerably more, a good half of them purpose-built — those come out of project work and are not part of the standard. We say so because screenshots of our site would otherwise promise more than the package holds. **Can I get out again if I want to switch?** Yes. Content, media and core settings can be exported in full at any time, in open formats. To be honest about it: an export is a complete copy of your data — not a website that runs elsewhere straight away. Rebuilding it with the new provider is work you have to plan for. But your content is yours, and it comes in a form anyone can work with. Let's talk straight ## Sounds like your foundation? Let's talk straight — no sales patter. We look at your business and tell you honestly what makes sense. - Non-binding & no obligation - Honest assessment, no pitch - Reply usually within 24 hours Free · 30 minutes · directly with Florian [Discuss your project →](https://btlabs.dev/en/contact#cta)[Call: 30 min, free](https://calendly.com/florian%5Fberger) Facts ## Everything you need — at a glance. What btlabs Core can do, how to reach us directly, what it costs, and why you can trust us — the hard facts, compact and ready to expand. Show allCollapse all Business Berger+Team d. Florian Berger Rating [5.0 on Google](https://maps.google.com/?cid=17283123306740815377) Address Reichrieglerweg 17/3 · I-39100 Bozen (BZ) · Italy – South Tyrol Email [info@berger.team](mailto:info@berger.team) Phone · WhatsApp [+393461033556](tel:+393461033556) Languages Deutsch Italiano English **1 btlabs Core found by humans & AI** Solution Tailored foundation on headless architecture — a base that can power your website, shop & AI agents alike. Findable Prepared for Google and AI answers — GEO out of the box (llms.txt, MCP), under 200 ms median in the load test (August 2026). Automation AI translates, writes meta-texts & image descriptions, and reviews content — DE/IT/EN, via your own AI account. Integration MCP interface, REST & GraphQL — open interfaces for CRM, ERP & co. **2 Direct Contact 1 business day** Available Email, phone, WhatsApp, video call Contact Florian Berger Response time usually within 24 hours **3 Service Area DACH region, Italy** On-site South Tyrol — Bolzano and surroundings Remote Italy & entire DACH region — well-practiced Languages [Deutsch](https://btlabs.dev/de) · [Italiano](https://btlabs.dev/it) · [English](https://btlabs.dev/en) **4 Availability & Capacity limited** Currently **We have openings — but only a few.** We deliberately take on only a few projects at a time. Start after a straight-talk conversation and clear roadmap **5 Price Transparency €€-€€€** Projects Classification **€€-€€€** · All prices excluding VAT. Model Investment — **no subscription for the core system** Ownership Your data and content belong to you — full export, anytime **6 Experience & Trust since 2019** Collective Berger+Team — since **2019** Experience **20+ years** in branding, web & marketing Testimonials [Reviews on Google](https://maps.google.com/?cid=17283123306740815377) · [Partner voices on the home page](https://btlabs.dev/en#stimmen) **7 Data Protection & Tech cookieless** Hosting EU, isolated Tracking Cookieless, no third-party trackers — **no cookie banner** Security AES-256, 2FA, daily encrypted backups --- Source: https://btlabs.dev/en/features Languages: [de](https://btlabs.dev/llms/de/features) · [it](https://btlabs.dev/llms/it/funzionalita) See also: [llms.txt](https://btlabs.dev/llms.txt) · [ai.txt (Policy)](https://btlabs.dev/ai.txt) · [identity.json](https://btlabs.dev/identity.json)