The Silence After Launch

Freddy shipped in March. By June the analytics dashboard shows forty visits a week, thirty of them Freddy's author checking that the site is still up. This chapter is about why that happens to good products, and what the first real measurement should be.

The problem

Nobody tells you that your site is invisible. There is no error, no bounced email, no red build. Search engines simply have nothing compelling to rank, or cannot fetch what you published, or fetched it once in April and have not returned. Answer engines compose replies about “the best small notes apps” from four sources, none of which is you. All of this is silent by design: the systems have no obligation to explain themselves, and they do not.

So the failure mode is not that people do the wrong work. It is that people do no work, for months, because nothing has raised its hand. And when they finally do act, they act on folklore.

The hard way

The traditional response to silence is to guess, in roughly this order:

• Write more content. Five blog posts appear. Nothing changes, because the problem was never volume.
• Sprinkle keywords. The landing page now says “notes app” eleven times and reads like a ransom note.
• Buy a subscription to an SEO suite. It reports 143 issues, 130 of which are “image is missing alt text” on the icons in your footer, and none of which is the actual cause.
• Ask an LLM to “optimise this page”. It rewrites your copy into beige marketing prose and cannot tell you whether anything improved, because it never measured anything.

What every one of these has in common: no measurement before, no measurement after, and no way to tell a cause from a symptom. Six weeks later the traffic graph is identical and the conclusion is “SEO is a scam” — which is the wrong lesson from a real experience.

Two audiences, not one

Before measuring anything, it is worth being clear about who you are being measured by, because since 2024 the answer is two different readers with two different failure modes.

The engine that ranks links. It crawls your page, ranks it against others, and shows a human a blue link to click. Everything in classical SEO — titles, descriptions, headings, sitemaps, internal links — serves that chain.

The engine that composes answers. Someone asks ChatGPT, Claude, Perplexity or Google's AI answers a question; the engine retrieves sources and writes a reply in its own voice; your page is either one of those sources or it is not. No click, no ranking, no list. The work of being chosen there is called GEO — generative engine optimization — and it overlaps with SEO without being the same thing.

A page can rank third and be cited by nobody. A page can be cited by three engines and sit on page four of Google. They read differently, they are blocked differently (chapter 6), and they fail differently. This course works on both and says which one each step is for.

What a score is, and what it is not

Sjá gives a page a score out of 100 and a grade from A to F. Before you see your first one, it is worth being precise about what that number means, because the number is where most people's understanding goes wrong.

The score is a weighted sum of checks that each have a defensible reason to exist. It is not a prediction of your ranking. Nobody outside Google can predict your ranking, and any tool that claims a number correlates with position is selling you a correlation it cannot show. What the score measures is eligibility: whether the things that make a page rankable and citable are present, correct and reachable.

That distinction matters in both directions. A 95 does not guarantee traffic — you can be technically perfect and answer a question nobody asks. A 40 does, however, reliably explain why traffic is absent, because at 40 something concrete is missing: a title, a description, a canonical, an entity, permission for the crawler.

The shape of the work ahead

The next thirteen chapters follow one arc, and each step exists because the step after it is meaningless without it:

1. Measure one page honestly (ch. 3) — you cannot fix what you have not seen.
2. See what a model extracts from it (ch. 4) — because you have been writing for humans and a keyword, not for an extractor.
3. Fix that page and prove the fix (ch. 5) — a change you did not re-measure is a belief, not a result.
4. Let the robots in (ch. 6) — the most common cause of total invisibility, and the one nobody checks.
5. Say who you are (ch. 7) — an engine that cannot resolve you to an entity will not cite you as a source.
6. Do all of it at site scale (ch. 8–9).
7. Ask the engines directly (ch. 10) and check whether their crawlers actually came (ch. 11).
8. Keep it (ch. 12–14), because optimisation decays the moment somebody ships a redesign.

Do this now

Before installing anything, write down two things. They take five minutes and they make every later chapter sharper.

One: the question. Not the keyword — the sentence a human would actually type or say. For Freddy it is not “notes app”; it is “what is a good notes app that keeps my notes as plain files I own?” That sentence is what you will put to the answer engines in chapter 10, and it is the thing your page must genuinely answer.

Two: the honest baseline. Today's weekly visits, today's impressions if you have Search Console, and today's date. Write them in a file. In chapter 14 you will want a number to compare against, and memory is a flattering witness.

What you learned

• Invisibility is silent. No system is obliged to tell you why it ignored your page, and none of them do.
• There are two audiences now — the engine that ranks links and the engine that composes answers. They fail differently.
• A score measures eligibility, not ranking. A high one is no promise; a low one is a reliable explanation.
• A category that measured nothing is left out rather than scored 100 — inflated numbers hide exactly the gaps you needed named.
• Write the question and the baseline down before you change anything.