Parent technology brand

The technology behind the assessment

Genyra builds the applied-intelligence platform behind IXSEO, including the analysis pipeline, structured evidence model and interpretation layer.

The role

What a platform is for in a service business

The difficulty with expert assessment is not producing a good one. It is producing a good one every time, on a website nobody has seen before, without the quality depending on which day of the week it was started. Consistency is the thing consultancies struggle with, and it is almost entirely a tooling problem.

Genyra exists to remove that variability from the parts where it does not belong. Every analysis gathers the same evidence in the same way, applies the same rules to it, and presents findings in the same structure. What varies is the judgement — which is where variation is the point, because judgement is responding to a business rather than following a checklist.

The second reason is discipline. Separating fact-gathering from interpretation forces a distinction that is easy to blur: what was measured, what was inferred, and what somebody concluded. When those three are handled by the same process they tend to be presented with the same confidence, and that is how analysis becomes plausible rather than accurate.

Components

What Genyra contributes

Four layers, each with a boundary the others cannot cross.

Collection

A guarded fetcher that re-checks every redirect hop against a DNS-level guard
Byte caps, timeouts and content-type limits on every request
Passive markup parsing with no script execution
Public configuration lookups across DNS, mail and transport
Third-party adapters for measured performance data

Evidence model

A single structured object describing everything observed
Explicit distinction between measured, inferred and absent
Recorded failures rather than silent omissions
Nothing carried forward that was not established

Interpretation

Rule-based interpretation that always runs
Model-assisted interpretation where configured
Strict schema validation of all model output
A filter that removes claims IXSEO will not make
Score clamping so passive analysis cannot claim perfection

Presentation

One finding structure applied everywhere
Severity and confidence carried on every finding
Public and verified views built from the same evidence
Sensitive detail withheld rather than hidden

Division of labour

What code establishes, and what it must not

The most important design decision in the platform is also the simplest: a language model is never asked a question that code can answer. Whether a canonical tag exists, how many H1 elements a page has, whether SPF is published, what Lighthouse measured — these are determinate. Asking a model produces an answer that is usually right, which is considerably worse than one that is always right.

What a model does well is the part code cannot approach: reading a home page and forming a view on whether a stranger would understand what the business sells. Judging whether content answers a question or restates it. Deciding which of thirty valid findings a managing director should read first.

Keeping that line sharp is what makes the interpretation trustworthy. When a finding says something was found, code found it. When a finding says something matters, that is a judgement — and it is labelled as one.

Determined by code

Presence of titles, descriptions and canonicals
Heading structure and validity
Robots and sitemap availability
Structured data types declared
SPF, DKIM and DMARC presence
Measured performance and response states
Security header presence

Interpreted with judgement

Whether the site explains what it sells
Whether content is useful or merely present
Entity clarity and generative readiness
Trust and brand perception
Priority order and commercial framing
The executive summary

Constraints

What the interpretation layer is prevented from saying

Model output passes a filter before anything reaches a report. These claims are removed regardless of how confidently they were produced.

Guaranteed rankings

No system can promise a search position, and a report that implies one is worse than useless — it sets an expectation the work cannot meet.

Guaranteed AI citation

No generative system publishes its retrieval criteria. Claiming a change will produce citations is speculation presented as method.

Guaranteed traffic growth

Traffic depends on demand, competition and factors outside the website entirely. Improvement can be argued for; it cannot be promised.

Unevidenced penalties

Declaring a site penalised without manual-action evidence is a common and damaging guess. Declines have many causes and most are not penalties.

Exploitability

Passive analysis observes indicators. Stating that something is exploitable requires testing, which requires authorisation we have not been given.

Breach implies compromise

An address appearing in a third-party breach says something about that third party. Presenting it as evidence of active compromise is alarmist and wrong.

Common questions

Is Genyra a product I can buy separately?

Not through IXSEO. Genyra is the technology brand behind the service rather than something sold alongside it. What you engage is IXSEO — the assessment, the interpretation and the people. The platform is how that gets done consistently rather than a separate line item.

Does an AI write my report?

No. A language model contributes interpretation — priority ordering, business framing, the executive summary — working only from structured evidence that code has already established. It never determines whether a title exists, whether a page can be crawled or what the DNS publishes, because those are facts a model would only be guessing at. Its output is validated against a strict schema, filtered against a list of claims we will not make, and reviewed by a consultant before it reaches a paid report.

What happens if the AI is unavailable?

The analysis still runs and still produces genuine findings. Rule-based interpretation always executes and establishes the factual findings and the scores; the model adds an interpretive layer on top of it. This is deliberate: a service that produced nothing when a third-party API had an outage would be badly designed, and building the deterministic path first is also what stops the model quietly becoming the source of truth.

Is my website data used to train models?

No. Evidence gathered during an analysis is sent to the interpretation provider for the purpose of producing your report and nothing else. Raw fetched content is held only as long as the analysis needs it and is discarded on a short retention window, which is set out in the data retention policy.

Genyra is the parent technology brand for IXSEO. Elite Security Europe and Intelis contribute specialist expertise to specific parts of the service and are separate from it.

IntelligenceHow It WorksThe eight stages of an analysis, and which are automated, interpreted or reviewed.PolicyAI DisclosureExactly where AI is used, what it is constrained from doing, and how output is reviewed.How we workMethodologyWhat counts as evidence and how confidence levels are assigned to findings.

See the platform produce something

A free snapshot runs the full collection and interpretation pipeline against your own website.

Analyse your website Read the methodology