We build websites that search engines can index and answer engines can quote — and we do the process work behind them, so a digital change actually holds.
Canonicals, redirects, sitemaps, status codes and page speed. This is the layer where most sites lose traffic silently — a canonical pointing at a page that no longer exists costs nothing to fix and can undo months of accumulated signal.
JSON-LD that declares what your business is, where it is, what it sells and what it answers. This is the layer almost everyone skips, and it is the one that decides whether a machine understands you or merely stores your text.
Content shaped so it can be quoted: self-contained passages, questions answered in the first sentence, and FAQ blocks that match the visible page. Being the cited answer is a different job from ranking third in a list of links.
The process work behind the site: what is recorded, what is re-typed between systems, what still runs on a spreadsheet. We change one process at a time, because programmes that try to change everything at once are the ones that stall.
A website can be well designed, fast and full of good writing, and still fail to state the one thing a machine needs: what the organisation actually is. That statement lives in structured data — a block of JSON-LD in the page source declaring the entity type, the address, the opening hours, the services, the questions the page answers.
We audited three Portuguese websites side by side in 2026: a national hotel chain, one of
the country's largest radio stations, and a two-person consultancy. None of the three
blocked AI crawlers. None of the first two declared what they were — the hotel chain had
no Hotel or LodgingBusiness markup, and the radio station had no
RadioStation or BroadcastService. Both shipped the default graph
their CMS plugin generated, which described them as generic organisations with a contact
form.
This is the pattern we keep finding. Plumbing gets attention because it breaks visibly and because the CMS handles much of it. The semantic layer — the part that says what the entity is — gets skipped, and it is both the cheapest to fix and the one that weighs most for answer engines and for models learning about entities.
Classic SEO optimises for a position in a list of blue links. Answer engine optimisation optimises for being the passage a system quotes when it composes an answer. The two share their technical foundations — crawlable HTML, correct canonicals, valid structured data — but they reward different writing.
A passage that gets quoted is self-contained. It states its fact in the first sentence, does not depend on the paragraph above it, and does not resolve pronouns against something three sections earlier. Writing that way costs nothing and helps human readers too, since most of them are skimming rather than reading in order.
There is a deflationary point worth making here. Much of what is sold as AEO or GEO is technical SEO under a newer name. The most common problem we find — a canonical pointing at a page that returns 404 — is a classic error from 2010. What changed is not the fundamentals but the number of things reading them: a search crawler, half a dozen AI lab crawlers, and agents fetching pages in real time on a user's behalf.
For a small or mid-sized company, digital transformation is not something you buy. It is changing how one process runs, then the next. Which information gets recorded, which gets re-typed by a person moving between two systems, which decisions wait on someone opening a spreadsheet. The website is usually where the change becomes visible to customers, but the work sits behind it.
We scope digital transformation one process at a time, with a defined outcome for each. Programmes that set out to change everything simultaneously are the ones that stall, because they cannot show a result before the budget or the patience runs out. A first process that visibly works changes what the next decision should be — and that is worth more than a roadmap.
Where AI helps, it is usually in the seams: reading a document so a person does not have to re-type it, drafting a reply that a person approves, checking one system against another. Where it does not help, we say so, because an AI project attached to a process that nobody has mapped is a demonstration, not a change.
Two cases where the change happened on the client’s own website — and what got shorter was the distance between the visitor’s intent and the first useful answer.
An insurer replaced its «Call Me» button with an assistant that talks, gathers what the quote needs and returns a result in the same interaction — at 11pm or on a Sunday.
Delivered project. No conversion figures published.
Read the caseEight properties across three regions replaced the email quote request with a seven-step journey that returns a proposal to the client and an already-qualified lead to the sales team.
Delivered project. Catalogue figures, September 2026.
Read the caseA property development that answers by itself over email, with AWS Lambda and an LLM using the project’s real knowledge — with nothing for the buyer to install.
Delivered project. No conversion figures published.
Read the caseStructured data is a block of machine-readable code, in a format called JSON-LD, that states plainly what a page is about: that this organisation is a hotel, that this page answers six specific questions, that the business opens at 09:30. Search engines and answer engines read it directly instead of inferring meaning from the visible text. Most Portuguese business websites either omit it or ship whatever their CMS plugin generates by default, which usually fails to declare what the company actually is.
Answer engine optimisation, sometimes called AEO or GEO, is the work of being cited as the answer by a system that reads your page and summarises it — ChatGPT with search, Perplexity, Google AI Overviews. Classic SEO optimises for a position in a list of links. AEO optimises for being the passage that gets quoted. The technical foundations are shared: both need crawlable HTML, correct canonicals and clean structured data. The content differs: an answer engine rewards self-contained passages that state a fact in the first sentence.
That is your decision and we implement whichever you choose. Blocking crawlers such as GPTBot, ClaudeBot or CCBot keeps your content out of AI training and out of AI answers. Allowing them is what makes citation possible. In our experience auditing Portuguese sites, most allow them without ever having made the choice deliberately — the door is open by default rather than by decision.
For a small or mid-sized company, digital transformation is not a platform purchase. It is changing how one process runs, then the next: what gets recorded, who does what by hand today, where information is re-typed between systems. The website is usually where it becomes visible, but the work sits behind it. We scope it one process at a time, because programmes that try to change everything at once are the ones that stall.
Technical fixes register within days to weeks: a broken canonical, a missing sitemap or a page that returns the wrong status code are corrected and re-crawled quickly. Ranking for competitive terms takes months and depends on authority, which is built by other people linking to you and cannot be bought honestly. We separate the two when we scope, so you know which part of the work has a fast payback and which does not.
Yes, and we also build without a CMS when a site does not need one. A five-page company site that changes twice a year does not need a database, a plugin stack and a monthly update cycle. A site with a blog, several editors and frequent publishing does. We recommend based on how often the content actually changes and who will be editing it.