An AI answer appearing above the search results

Someone types your category into a search box and gets a paragraph. Well-written, confident, sourced from a handful of pages, and complete enough that they never click anything. If your business is not inside that paragraph, you were not considered — and you will not see it in your analytics, because the visit that did not happen leaves no trace.

This is the quietest change in search since mobile, and the reaction to it splits neatly into two camps: people declaring SEO dead, and people carrying on exactly as before. Both are wrong in the same way. The mechanism changed; the underlying requirement did not.

What actually changed

Generated answers are assembled from sources the model considers reliable on that question. That is a different selection problem from ranking ten blue links. Being on page one is no longer sufficient and, occasionally, no longer necessary — what matters is whether your page contains a clear, extractable, attributable answer to the specific question being asked.

The practical consequences are unfamiliar. Pages that rank well but bury their answer under six paragraphs of preamble get skipped. Pages that answer plainly in the first hundred words get quoted. Long pages covering ten adjacent topics do worse than short pages covering one, because the model is matching a question, not a topic.

Ranking was a competition for position. Being cited is a competition for clarity.

What to do about it

Answer one question per page, near the top

Find the real questions your buyers ask — the phrasing they use with a colleague, not the phrasing a keyword tool suggests — and give each one a page that answers it in the opening paragraph, then earns the rest of the length by going deeper than anyone else. The answer first, the credibility after.

Make the facts machine-readable

Prices, ranges, timeframes, what is included, who it is for. Stated plainly, in text, in a structure a parser can follow. Numbers hidden in an image or generated by a script are invisible to the thing deciding whether to cite you. This is also why vague pricing pages perform badly here: there is nothing to extract.

Be specific enough to be worth quoting

Generic advice is already in the model. It has read every article that says “focus on quality content”. What it does not have is your particular arithmetic, your process written out, the trade-off you would explain to a client in a meeting. Specificity is not just better writing — it is the thing that makes a source worth naming.

Give it something to trust

Named authors with real expertise, dates, sources, consistent details about who you are across the web. The signals that made a site credible to a human reviewer are broadly the signals that make it a safe citation.

This is what the SEO and AI visibility work is now

Not ranking reports. Whether your business is present, accurate and quotable at the moment someone asks a question about your category — in the classic results and in the generated answer above them.

See how we approach AI visibility

What this does to how you measure

If part of your audience now gets their answer without visiting, sessions become a worse proxy for demand than they used to be. Two things become more useful. First, branded search volume — people who saw you in an answer and came looking specifically for you. Second, the honest conversation with new enquiries about how they first heard of you, which is unfashionable, unscalable and frequently more accurate than the analytics.

None of this makes the work optional. It makes it different: fewer, better pages, each one built to be the clearest available answer to a question a buyer actually asks. That was always the best version of the job. It is now the only version that works.