How we think about search when an AI writes the answers

Search is changing faster than at any time since Google appeared. AI Overviews summarise the answer before the first link. ChatGPT, Gemini, Perplexity or Copilot answer complex questions citing a handful of sources. Google Maps decides which businesses someone two hundred metres away gets to see. And, at the same time, the traditional search engine is still where a great many buying decisions are made.

There are two common reactions to this shift. The first is panic: "SEO is dead, we need to do something else". The second is a cosmetic makeover: renaming the same old services with new acronyms. We think both are wrong. What follows are the twelve principles we use to make decisions, for ourselves and for our clients.

Most SEO strategies start with a spreadsheet of keywords sorted by volume. We start with awkward questions: which products carry a margin? Which customers come back? What has to happen for someone to ask you for a quote? How much can you afford to pay to acquire a customer? Only with those answers does it make sense to decide which searches to pursue. Ten visits from people with buying intent are worth more than a thousand browsers, and a plan that ignores this optimises the wrong metric.

Your customer doesn't think in "channels". They ask an assistant for the best option, search your name on Google to check you exist, look at your reviews on the map and come back to your site from an email. Optimising each touchpoint separately, with different teams and metrics, produces contradictory messages and duplicated effort. We design one strategy that accounts for all of those touchpoints and measures visibility in each, including the ones that don't generate a click.

Assistants that consult the web in real time retrieve pages from search indexes. If your site isn't crawled properly, isn't indexed or isn't understood, it won't exist for them either. That's why we don't treat GEO as a replacement for SEO but as an additional layer: content structured to be easy to extract, clear entities, consistent structured data and citation tracking. Whoever abandons the foundation to chase the novelty ends up with neither.

A language model summarising ten similar pages has no reason to cite any one of them in particular. It cites the one that offers something the others don't: original data, a more precise explanation, first-hand experience, a useful comparison table, an unambiguous definition. We write to that standard. That means fewer pieces and more work per piece, interviews with the people in your company who genuinely know the subject, and editing that removes everything the reader could already find elsewhere.

A template that generates thousands of duplicate URLs, JavaScript that hides the main content, a migration without redirects or a page that takes six seconds to load on a phone aren't "IT issues". They are losses of visibility and conversions that compound month after month. We treat a site's technical health as a marketing asset, with issues prioritised by business impact rather than by the severity an automated tool assigns them.

A report full of green charts can hide a business that isn't growing. Before optimising anything, we make sure analytics records what matters: forms submitted, calls, bookings, sales, and where they come from. Then we report in that order: business results first, then the indicators that anticipate them (visibility, clicks, presence in AI answers) and, finally, the work done. When something can't be measured precisely, we say so instead of inventing an attribution.

We've seen too many companies without access to their own Google Ads account, their Google Business Profile or even their domain. It's a form of lock-in that has nothing to do with service quality. With us, everything is created in or transferred to your company's name from day one, and we work with access you grant and can withdraw. We want you to keep working with us because it pays off, not because leaving is painful.

Buying links from blog networks, publishing hundreds of auto-generated pages without review, inflating reviews or hiding text for search engines may produce a temporary spike. It can also produce a manual action, a drop after an algorithm update or a legal problem under consumer protection rules. Our rule is simple: if you wouldn't be comfortable explaining it in public, we don't do it. And we don't guarantee rankings, because nobody can honestly do so.

Increasingly, visibility depends on the ecosystem understanding who you are: which company stands behind the site, what it's expert in, who writes for it, where it's mentioned. That's why we work on entity SEO, consistent company information across the web and mentions in relevant media through digital PR. These processes are slow and cumulative, and that is precisely why they're a lasting competitive advantage.

Any tool can generate an endless list of warnings. The value lies in deciding which five things to do first, who should do them and what each is expected to achieve. We order work by expected impact, effort and dependencies, and we're willing to leave things for later. Doing the important things well is worth more than half-doing everything possible.

Every proposal sets out scope, deliverables, timeline and fixed fees. Every month you receive a report that explains, in plain language, the work done and its effects. Every quarter we review the strategy with the data in front of us. If an experiment doesn't work, we tell you. If we think an investment isn't right for you, even one we would carry out ourselves, we tell you that too.

Good search work requires understanding each business in depth, and that doesn't scale with templates. We take on a number of projects that lets us know your sector, your site and your goals, with a dedicated contact who doesn't change every quarter. When a project needs a particular speciality, we bring it in with care; what we don't do is accept work we can't do well.

What this manifesto isn't

It isn't a list of promised results. It's a description of how we make decisions. Search will keep changing: new formats, new assistants and new ways of measuring will appear. We'll revisit this text whenever we learn something that warrants it. What we don't expect to change is the substance: understand the business, be the best answer, build on a solid technical foundation and be honest about what we know and what we don't.

To see how these principles translate into day-to-day work, read our methodology or find out who we are.