Local on Windows · Developed by CoreWeb Studio
Obserya audits a site on your own machine and links each finding to the affected page and the value measured there. The result is an action plan for both the client and the implementation team.
The application interface is currently in German. Screenshots on this page show it as it is.
Obserya is website audit software built by CoreWeb Studio for Windows. It crawls a site locally, scores technical SEO and AI readiness on two separate axes, and backs every finding with the affected page and the value measured there. From that it produces a prioritised action plan, a PDF client report and technical exports.
Why Obserya exists
An audit score describes a state, but it assigns no responsibility. Which page is affected, and who tackles what first, is usually worked out by hand after the crawl. Obserya brings that step into the analysis.
Overall score
90/100 does not mean flawless
The overall score summarises the site’s condition. Findings show which specific issues still need attention despite a strong score.
Finding
The finding names the problem and where it occurs
It describes the concrete issue and leads to the affected URLs.
Evidence
Evidence makes the finding verifiable
The measured values, the target and the page context show why the finding holds.
Action
Prioritised and ready to hand over
Effort, owner and next step prepare the finding for implementation.
Obserya does not replace an enterprise crawler. It closes the gap between the crawl result and implementation, where signals turn into prioritised work packages.
How Obserya works
Each step carries the previous step’s data forward — so any finding traces back to its source.
The site is captured on your own machine.
Obserya fetches and parses the delivered HTML while following internal links. The result is a full page inventory with metadata, status codes and an internal link graph.
A legal notice is not judged like a service page.
Every page is assigned a role. A missing meta description weighs differently on a utility page than on a service page.
Two axes, scored separately.
Technical SEO and on-page signals on one side, AI readiness and GEO signals on the other — with no third number blending the two.
Every finding carries its proof.
A finding consists of the affected page, the issue and the value measured there. Where no reliable occurrence can be established, Obserya says so explicitly.
Order comes from effort, impact and evidence.
Findings that can be implemented directly stay separate from those that need verifying. Both get an owner and a first step.
The client report and the technical exports come out of the same audit state.
The client report explains the state without SEO vocabulary; the exports pass the same data to the implementation team.
Detail from “How Obserya Works” — the view follows the step on the left.
Select a step.
Full workflow
How Obserya Works: six steps from the local crawl through to reports and exports, with the local-first architecture and optional LLM path below.
Select a step.
What Obserya checks
Technical SEO describes the foundation for classic search; AI readiness describes how well the same page reads as an answer source.
Why there is no third score
A page can be technically sound and still be hard to use as an answer source. An average would hide that difference.
Visibility foundation for classic search
Local SEOoptional
Local signals run as an optional block at project level; the standalone analysis area arrives with the beta cut.
Usefulness as an answer source
Findings and evidence
Six findings from a demo audit. The technical key appears the same way in the exports.
Select a finding
Finding types and actions match the application; quantities come from the demo audit.
Technical SEO · Internal linking
Links lead nowhere: an empty target, a bare hash, or a script call.
Evidence quality and urgency are two different things
A finding whose evidence is only partly conclusive is filed under “verify first” rather than in the immediate list.
Prioritisation and drilldowns
Obserya ranks findings by impact, effort and evidence quality, and shows its working.
Score class
Where the measured value sits on the scale.
A project scoring 91 falls into the “Good” class.
Action status
Whether evidenced findings are still open.
That same project can still have confirmed actions outstanding.
Both appear separately in the report and are easily confused.
The labels come from the report and describe how directly a finding can be acted on — they are not deadlines.
Low effort
This week
Well evidenced and actionable without further checks.
More effort
This month
Clear in substance, heavier to implement.
Check before acting
Verify first
Evidenced, but worth confirming manually once.
Bar width matches the value range: “Very good” covers five points, “Significant work needed” covers sixty-one.
Both scores, all areas and the overall status.
One analysis area with its value and findings.
The issue with its cause, action and affected pages.
The specific URL with the value measured there.
Every level stays navigable.
Local-first architecture
Crawling, analysis, scoring and report generation all run on the device. The audit data sits in a database of its own there.
The crawl of the target site needs network access.
Leaves the device only when triggered
The audit core stays on the device
From page capture through to the finished report.
Audit data stays on the device
Crawl and page data, findings, evidence and project settings.
No service in between
The core workflow needs no account and no API key.
Sending data out is a decision
Whether anything leaves is decided by the user, per task.
Everything inside the large block runs in the application. Only the dashed path at the bottom right leaves the device.
Full view
Obserya System Architecture: input signals, the local application with its ingestion, analysis and decision layers, local audit data, outputs and the optional LLM path.
Select a section.
Optional LLM assistance
Crawling, analysis, scoring and reports all work without a language model, which is what keeps results reproducible and defensible in a client meeting. Optional LLM assistance comes in only when it is explicitly triggered, and only for one clearly bounded task.
The strength lies in the local, reproducible audit core. A language model helps where a task is clearly bounded — and nowhere else.
Three stages run locally, only the last one leaves the device.
The core runs without it
The same audit state produces the same result through the same rules.
One task, one prompt
A selected task produces a structured prompt with context.
Only when triggered
Without that step, no audit content leaves the device.
Additive, never scoring
No language model changes scores or core findings.
Client reports and exports
The client report explains the state in plain language and separates confirmed work from open checks.
Audit state
Pages, findings, evidence and scores.
Prioritisation
Impact, effort and evidence quality set the order.
Report pages
Summary judgement, status and action plan.
Output
PDF for the client, plus HTML, CSV, JSON, Markdown and an Obsidian export for the work ahead.
Scroll sideways · open a page for the full view
The same findings, keys and evidence — once as a report to read, once as data to work with.
For the client
For the implementation team
Where it fits
A working tool for recurring audits. Six situations from practice.
Taking on a new client site
A new site enters the project. The first audit provides the basis for the proposal and effort estimate.
Create the project, crawl locally, run the audit.
A PDF for the client meeting and a CSV file for the effort estimate.
Built for
The focus is on small and mid-sized agencies with recurring audit work.
Development status
Obserya grows out of day-to-day audit work at CoreWeb Studio and is tested against real client projects. The current product core already works. The next features arrive with the beta cut; further confirmed modules follow step by step through to the full release.
These features are being finished now and are due with the beta cut.
In development now and scheduled for the beta cut.
After the beta cut, Obserya gains features that connect crawl data, content and search data.
These features are a committed part of the product plan and will be delivered step by step after the beta, through to the full release.
Common questions
These answers describe the current development status.
Website audit software built by CoreWeb Studio for Windows. It crawls locally, scores technical SEO and AI readiness separately, and backs every finding with the affected page and the value found. Output: an action plan, a PDF client report and technical exports.
Yes. Crawl, analysis, scoring and report generation run in the desktop application; the audit data stays in a local database. The crawl of the target site needs network access. Optional LLM assistance is used only when explicitly triggered, and only for one selected task.
No, those three steps are rule-based. Optional LLM assistance applies only to a selected task and changes neither scores nor core findings.
Technical SEO covers seven core areas: indexing and crawlability, document structure, meta and snippet basics, content basics, internal linking, structured data, and images and page elements; Local SEO runs alongside as an option. AI readiness scores six dimensions on its own axis, from answer clarity through to context density.
A finding names the affected pages and the value found — for a broken link, the target and anchor text. Where evidence only partly holds up, Obserya flags it for verification.
A client report as PDF, the same report as shareable HTML, a CSV of the findings and a technical CSV of the redirects observed during the crawl, JSON as a raw export, Markdown, and an Obsidian-compatible export.
Primarily small and mid-sized web, SEO and digital agencies handing audit results to clients and implementation teams. Freelancers and in-house teams use it in the same way.
The product core works: a full local HTML crawl with page inventory and link graph, technical SEO across seven core areas, AI readiness across six dimensions, evidence-based findings, a PDF client report and technical exports. The beta cut adds full JavaScript rendering during crawling, delivery as an installer, project data exchange between installations, Local SEO as an analysis area in its own right, and an optional LLM review view. Confirmed for the full release: Google Search Console data, cannibalisation analysis, keyword clustering, content briefs, and migration and redirect mapping.
Obserya comes out of day-to-day audit work at CoreWeb Studio — where crawl data has to turn into something a client will read and a team can work through.