We do not sell MS-Tools. It is not a product, it has no price, and we do not yet know whether it will ever be released. It is the tool we work with, and it monitors 22 client websites every day, the ones on a maintenance plan.
We are writing about it because what a tool does says more about how a company works than any marketing line.
Three problems
Anyone who maintains other people’s websites runs into the same three, over and over.
You cannot see. You do not know a site went down until the client phones you. You do not know a certificate is expiring until the red warning appears in every visitor’s browser.
You cannot show your work. The client asks “what exactly do you do for me every month?” and the answer is verbal. There is no measurable history, only impressions.
You run out of time. Checking one site by hand, speed, SEO, SSL, broken links, WordPress, is half an hour of work. Times 22 sites, every month, that is eleven hours nobody has.
MS-Tools turns all three into automation.
Eight checks on every scan
Each scan runs eight independent checks and keeps a full history, so that there is a comparison rather than a snapshot:
- Performance via Google PageSpeed Insights, mobile and desktop, together with SEO, Accessibility and Best Practices. We store the individual audits with their estimated savings in milliseconds and bytes, not just the final score.
- SEO and content: titles, meta, heading structure, structured data.
- Infrastructure files: robots.txt, sitemap, llms.txt.
- SSL with a real TLS probe: days to expiry, certificate chain.
- Technologies: detection of CMS, framework, CDN and plugins from HTML and headers.
- WordPress audit for WordPress installations.
- Broken links, via crawl.
- Environmental footprint, through the Green Web Foundation.
And a ninth that stands apart: traffic read from the server’s AWStats. Real unique visitors per month, bots excluded, read straight off the server. No sampling, no third-party estimate. An external tool only sees what the analytics script hands it; we see what actually reached the server.
Out of all this comes a health score per site, graded A+ to F on a 0 to 100 scale, a security score based on seven security headers, and per-site SLA targets that light up the moment a metric falls outside them.
The watchman, and the watchman’s watchman
Uptime monitoring runs every 5 minutes. The details are what matter.
Anti-flap logic. One failure is not enough. An immediate retry follows, and a site is only marked down after two consecutive failed cycles. A momentary network hiccup does not fire an alarm, because an alarm that cries wolf stops being read.
Two channels, with escalation. Email and Telegram, immediately. If the site stays down, it re-alerts at 1 hour, at 6 and at 24, so the problem does not get lost in the flow of the day.
A dead-man switch. If the monitoring system itself stops, we find out. The watchman watches the watchman.
And the point that decides whether any of the above is worth anything: we tested it. We took sites down deliberately, in controlled tests, to see whether and how fast the alert arrives. It arrived on Telegram immediately. A monitoring system you have never tested is not a system, it is a wish.
The SSL expiry warning fires when a certificate drops below 14 days, once a day so it does not become noise. It has already fired several times, always in time. It is the kind of problem the client never sees, precisely because somebody saw it first.
The part no conventional tool has
When SEO started turning into GEO, we went looking for a tool that measures whether a brand appears inside the answers of answer engines. There was none. So we built one.
The GEO panel in MS-Tools answers one question: when a prospect asks an answer engine, do you show up?
Per site we define the visibility variables: brand, alternative spellings, sector, regions, competitors, target language. From those we generate target questions, and they are deliberately blind: phrased the way someone who does not know you would phrase them. Not “what does company X do”, but “who builds websites in Athens”.
The questions run through Perplexity Sonar, GPT, Claude and Gemini, with the cost per call visible before anyone presses run. For each question we record whether the brand was mentioned, whether it was cited as a source, in what position, and with what share of voice against competitors. One click shows the full answer, with brand mentions highlighted and the sources as links.
The GEO score combines technical readiness at 35% and actual quality of appearance in the answers at 65%. Because technical readiness is a precondition, not a result.
The check runs automatically every week for the sites that have the service, and keeps a history of every run.
And the other side: who is reading you
Prompt-testing answers “what do the engines say about you”. That leaves the other half: who visits your site, and what did they come for. We do not guess this, we read it from the server logs.
MS-Tools separates crawlers into four categories by purpose: answer engines that read in order to cite you, user searches where somebody asked something right then and the model came to look, model training that harvests text, and dual-use for the ones that do both, such as Applebot and bingbot. Those last ones are counted separately, precisely so they do not inflate the pure totals.
Here is what 180 days on a client website showed:
- 23,393 model training crawls. ClaudeBot leads with 5,836, followed by Amazonbot, Meta-ExternalAgent, Bytespider and GPTBot.
- 2,012 crawls from answer engines, meaning crawlers that read in order to cite. Mostly OAI-SearchBot, GoogleOther and PerplexityBot.
- 262 live user queries: somebody asked, and the model came to read the page right then.
- 67 referrals. Visitors who actually arrived from an answer engine.
- 9,461 dual-use, counted separately.
Sit with that ratio for a second: for every single visitor an answer engine sent, 349 crawls went to model training. Your content is read a great deal and returns little. That is the new economics of search, in numbers rather than opinions.
The panel also shows which engines read you at all: OpenAI, Anthropic, Perplexity and Google all pass through this site, and the first three also with live user queries.
And one clarification the tool itself makes, because this is where the mistakes happen: crawl presence is not a citation. Being read means you are eligible, not that you are mentioned. Whether and how you are mentioned is answered only by prompt-testing. A tool that told you “ChatGPT reads you, therefore you appear” would be lying to you.
GEO, in other words, is not a theory we write articles about. It is a measurement we run.
What the client sees
None of this is worth anything if it stays on an internal dashboard. So MS-Tools produces a branded PDF report, per site or consolidated across all of a client’s sites, covering current status, performance, security and the last month’s uptime. There is also a version written without technical jargon, for the client rather than for the developer.
There is also a visual diff between two scans, highlighting the pixels that changed. It catches what no metric shows, such as a layout that broke after an update and nobody noticed.
The technical identity, briefly
Next.js 15 with the App Router, React, Node 22. SQLite with WAL. Playwright and Chromium for screenshots and PDFs. It runs on our own server in Finland, wired into the MS-Logic ecosystem’s single sign-on. In production since May 2026.
Reliability by design: a daily online backup of the database with 14-day retention, automatic pruning and VACUUM, versioned schema migrations, and every cron job with its own logs and fail-safe behavior.
What this means for you
We do not assume, we measure. Every site we look after has a history of measurements, not impressions. When we say a site improved, there is a before and an after, with a date on it.
We learn about the problem first, not you. A check every 5 minutes, alerts on two channels, escalation if it persists, an SSL warning two weeks ahead.
We build our own tools. When there was no way to measure visibility in answer engines, we made one. Whatever we do for ourselves, we know how to do for you.
We are ready for what comes next. SEO is becoming GEO. We are not predicting it, we are already measuring it.
The numbers in our case studies, such as mswriter.eu from F to A and tech-hom.gr from D to A, did not come from an estimate. They came from here.
If you want to see where your website stands today, get in touch.