Is SEO over, or did the thing we measure change?

For two years now we've been hearing the same sentence in different voices: "SEO is dead." It has been wrong every time, and every time there was a correct instinct underneath it. The correct instinct is this: search isn't disappearing, clicks are.

What changed

A search result used to return a list. Tenth place got you few clicks, first place got you many, and the whole job was climbing that list.

A search result now increasingly returns an answer. Google writes the response at the top of the page in an AI Overview. The user takes the answer and leaves without clicking anything. The industry calls this "zero-click" and its share has been climbing for years.

A second shift sits on top of it: people no longer type certain questions into a search box at all. They ask ChatGPT, Gemini or Perplexity directly. There's no list there either — there's an answer, and that answer names three or four brands.

Both cases lead to the same place: being first is not the same as being in the answer.

Why "SEO is dead" is wrong

It isn't dead, because every one of these new answers still feeds on an index. The model isn't talking out of thin air; it's reading from infrastructure that crawls pages. If you're not in the index, you're not in the answer.

One detail surprises most people: ChatGPT's web search uses the Bing index. There are sites with excellent Google rankings and no Bing presence at all, and those sites don't appear in AI answers. The basics of technical SEO — crawlability, sitemaps, clean structure — matter more than they used to. Only the prize changed. You used to win a position. Now you win the right to be in the answer.

The real change: the unit of measurement

The unit of classic SEO was position. One number, one direction, everyone understood it.

There is no position in an AI answer. Ask the same question twice and you get two different answers. You're named in one and absent from the other. That forces the unit to be frequency: the same question is asked several times, and you count how often you appear. Deciding from a single measurement is like flipping a coin once and drawing conclusions about the currency.

The second difference: in classic search you type a keyword, to an AI you form a sentence. Not "menu management software" but "which menu management software should I pick for a small café?" Same intent, different content requirements. A page optimised for the keyword can't answer the sentence.

Three doors into the answer

When a brand is missing from AI answers, the reason is almost always one of three. Each can be measured on its own, and each has a different fix.

Door one: access. Your robots.txt may be blocking AI crawlers. Sites that allow Googlebot while shutting out GPTBot, ClaudeBot and PerplexityBot are extremely common, and most did it by accident. It's the dullest cause, the most frequent one, and the fix is a single file.

Door two: the shape of your content. If a model is going to form a sentence, it needs material: comparisons, pricing logic, who it suits, where the limits are. A page that only lists features doesn't provide it. Pages that state their limitations plainly get cited more often than pages that claim perfection, because the model is trying to compare.

Door three: what other people write about you. Models build answers mostly from review platforms, listing sites and industry articles. Absent there, the model doesn't know you. Present but stale, and it repeats a price you stopped charging years ago.

That third door is the most frustrating, because you don't control it. But it is measurable: you can see which sources an answer draws on, and count how many of them are yours.

So does the old work go in the bin

No. Your rankings still matter, because:

• a meaningful share of AI Overview sources are picked from page one anyway
• zero-click doesn't happen on every query; transactional searches still click
• getting indexed is the precondition for both worlds

What changed is that ranking is no longer sufficient on its own. You now need to know where you stand on two axes, and those axes can move independently. Ranking well organically while never appearing in AI answers, or the reverse, are both real situations.

You can't manage what you don't measure

The practical conclusion: if you only track rankings, you're seeing half the picture. The other half — whether the AI mentions you, which sources it feeds on, how your competitor got in there — is currently measured by almost nobody.

That's why we built Scooutly. We put the same question set to five language models and to Google AI Overview, then translate the same intent into a search keyword and measure it in Google's organic results. Alongside it we put your competitor's pricing, packaging, reviews and ads — because the answer to "why do they show up and I don't" usually isn't in the content, it's in the positioning.

We don't ask the model for the score; we compute it from an open formula and keep the evidence behind every finding. A number you can't trace won't make anyone act on it.

---

The SEO era isn't ending. The era where it could be told with a single number is.