Free AI export diagnosis
Start a check-up with nothing but a website URL. The server really crawls your site and its key pages, then a deterministic rule engine returns 3–8 key findings — each carrying evidence, source and timestamp. Verifiable, not model-generated.
Nine dimensions, checked deterministically
This is not a model writing an opinion about your site. Every dimension has reviewable checks, and each conclusion resolves to four things: rule id, evidence, source and timestamp.
GEO
Whether generative engines can reach you, read you, and are willing to cite you. Structured data, citable summaries and crawl access are the core here.
SEO
Index coverage on classic search: whether robots and sitemap agree, whether canonical leaves duplicates, and whether internal links carry weight to landing pages.
Website
The site itself: reachability, load behaviour, mobile layout, and whether the language entries are complete.
Search
How search demand lines up with your pages — what buyers query versus what your pages actually say.
Media
Whether what your site claims also appears, consistently and discoverably, on external platforms.
Content
Coverage of real buyer questions: for every question a buyer actually asks, is there a page that answers it head on.
Market
Fit between target markets and buyers — whether the market choice is right and the buyer profile is clear.
Acquisition
How your current channels map to your target customers, and which link is missing from the mix.
Conversion
The path from visitor to enquiry: where the entry points are, where the form stalls, and whether follow-up leaves a trace.
Every finding carries evidence: rule id, evidence, source URL and timestamp. Only with all four is a conclusion valid — which is what keeps advice from becoming plausible-sounding but unverifiable.
Every conclusion carries twelve fields
A diagnosis is not a paragraph of commentary but a structured, checkable conclusion. A conclusion missing fields never reaches the report — evidence and confidence above all.
Decision
What this finding says: one problem, one opportunity or one recommendation, in a sentence.
Evidence
The checkable fact behind it: the page, field, response header or excerpt that was captured. No evidence, no finding.
Confidence
How reliable the finding is and where it is weak. Low confidence is stated outright, not hidden.
Expected impact
What fixing it is expected to change, and on what basis. We make no effect claims without a basis.
Risk level
How risky the action is: what breaks if it is done wrong, and whether it can be undone.
Action
Concrete enough to act on — which file, which setting — rather than directional advice.
Approval required
For high-risk actions the system only proposes and pauses, until a human approves.
Reason code
Which rule produced the finding, so repeated hits on the same cause can be counted.
Source version
Which crawl batch this finding is based on, so re-tests compare like with like.
Model version
The model version involved. If the model changes and findings change, the report shows it.
Prompt version
Which prompt version was used. Prompts are versioned assets, not text edited on the fly.
Run ID
Every run is traceable: all findings from one run share a single identifier.
Market recommendations must come from actual diagnosis evidence. The shape of the conclusion is "this company is currently best placed to enter: Country A, Country B, Country C". Country placeholders are used here rather than three country names, because which three must be derived from evidence; without enough evidence we leave it blank and never pass sample data off as a real result.
High-risk actions are never taken by AI on its own. For legal signing, payments, changes to core commission rules, deletion of core data, changes to customer ownership, irreversible external commitments, permission changes and bulk sending, the system only proposes and pauses for human approval.
Check first, then decide on the full diagnosis
The check-up exists so you can see the problems before committing budget. It is not a trial version of the full diagnosis.
Free check-up
- Starts from a single website URL
- The server really crawls the site and its key pages
- Returns 3–8 key findings with evidence, source and timestamp
- Gives an overview reading across the nine dimensions
Full diagnosis
- All findings across the nine dimensions, each with evidence and source
- P0 / P1 / P2 priority ordering, so it is clear what to fix first
- AI search visibility matrix: buyer questions × languages × repeated sampling, with competitor comparison and raw answer excerpts
- An actionable schedule, same-baseline re-tests and a report version number
The check-up tells you what is wrong; the full diagnosis tells you what to fix first, and how you will know whether it improved.
One required field — the rest sharpen the report
The website URL is the only required field. Everything else turns generic conclusions into conclusions about you.
| Field | Requirement | What it is used for |
|---|---|---|
| Website URL | Required | The crawl entry point — the only required field |
| Company name | Optional | Report attribution and entity check |
| Industry | Optional | Sets the industry baseline so the comparison is fair |
| Target markets | Optional | Multiple countries or regions, comma separated |
| Target customer type | Optional | Sourcing buyer, distributor, end brand, and so on |
| Current acquisition channels | Optional | Comma separated, for example LinkedIn, trade shows, B2B platforms |
| Product highlights / main products | Optional | Judges whether content assets match buyer questions |
| Competitor URLs | Optional, up to 5 | Used for comparison, to see which item the gap sits in |
The diagnosis type is chosen here too: free check-up (3–8 key findings) or full diagnosis (login required, nine dimensions).
Four steps to a working growth loop
The diagnosis is not the destination. It hands off to execution and tracking, otherwise the report sits in a folder.
Free check-up
Enter the website URL; the server crawls it and returns 3–8 key findings.
Full diagnosis
After login, unlock all findings across the nine dimensions, priorities and a schedule.
Delivery project
Once a service is confirmed, a delivery project and default task list are generated, with owners.
CRM follow-through
Opportunity boards and activity timelines keep it moving; re-tests compare on the same baseline as the original conclusions.
State what can be measured; never invent the rest
The credibility of a diagnosis comes from admitting what it does not know.
Crawl discipline
Crawling respects robots.txt. On encountering a captcha, a login wall or rate limiting it stops immediately — no bypass, no retry. Where data is missing, the report says "not obtained" or "cannot be verified" rather than filling the gap with inference.
What is not promised
No fixed ranking or recommendation placement on any search engine or AI platform is promised. Rankings, indexing time and traffic depend on third-party platform rules and competition, not on a single provider's decision.
One thing to be clear about first: the diagnosis reads publicly crawlable information, structured data and the fields you enter. Where a dimension has no external data source, the report states "not obtained" instead of producing a plausible-looking number.
Run a free check-up first
All it needs is a website URL. Results carry evidence, source and timestamp, and can be verified finding by finding.