We design content architecture around entities and intent, implement structured data and measure the effect over a defined period. AI answers draw on the same web content and the same Search systems, with one further condition — the answer has to stand as a self-contained unit. Just the work and the numbers.
Send a URL, the business goal and your single most important page. You get an assessment of the problem, the scope of work and a rough price — before any engagement.
A reply arrives by email within a few working days.
Request an assessmentA prioritised backlog of findings with an owner, a completion criterion and a source for each figure. Per URL, not in general.
Hard work. A hard product.
Request an auditTopical map, entities, micro-intents, content, semantic HTML, internal links and JSON-LD — as one measurable system.
No templates. No shortcuts.
Request a buildOrganises content around entities, attributes, relations and user intent — not around repeating a keyword.
How the architecture is designed →A machine-readable description of what is already visible. It adds no facts the site cannot confirm.
Rules for a cohesive JSON-LD graph →Connects image content, textual context and entity data so a system can tell what an image shows.
A practical guide and a test →A site does not become citable by adding special “AI texts”. What is needed is technical availability, clearly defined entities, direct answers, verifiable claims and a structure a system can reliably extract information from.
AI search connects indexing, retrieval of relevant parts of a document and the generation of an answer.
02 · displayAI Overviews assemble an answer from several retrieved sources; clear structure eases extraction but guarantees no citation.
03 · accessAvailability to LLM crawlers is controlled through robots.txt, HTTP responses and stable canonical URLs.
04 · answerThe direct answer comes first, then evidence, conditions, exceptions and clearly separated facts, estimates and recommendations.
URLs, goals, data, sources and constraints.
Topical map, entities, micro-intents and internal links.
Visible content, semantic HTML and a connected JSON-LD graph.
A baseline period, events and a rule for a responsible conclusion.
The taxi.co.rs case study separates publicly verifiable architecture from data supplied by the project owner. No result is attributed to a single schema change.
| Metric | Value | Evidence type | Note |
|---|---|---|---|
| City hubs | 23 | Public | Visible in the site navigation; checked 15 August 2026. |
| Clicks | 4,139 | Owner data (GSC) | 12 June – 14 August 2026, alongside 182,041 impressions. |
| AI impressions | 26,512 | Owner data (GSC) | 14.6% of all impressions; not independently verified. |
| Fare pages | National + local | Public | Fares kept separate per city; checked 15 August 2026. |
A URL, the business goal, your single most important page and how you currently measure success. Passwords and account access are not needed.
Not for the free assessment. Access matters later, when a baseline period has to be set, performance analysed or the effect of an implementation verified.
Both models are possible: finished content and code, or a precise backlog for your existing team. Scope is defined only after reviewing the site and its constraints.
There is no responsible universal timeframe. It depends on the size of the change, the state of the site, competition, re-indexing and a measurement period long enough to draw a conclusion.
The first step is a free assessment. No obligation, no account access and no sales pressure.